mirror/lume
1
0
Fork 0
mirror of https://github.com/trycua/lume.git synced 2026-01-06 04:20:03 -06:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'main' into models/opencua

Browse Source
This commit is contained in:
Dillon DuPont
2025-09-12 13:30:21 -04:00
parent 58807378dd 62efcd7687
commit cf95646503
122 changed files with 16257 additions and 113010 deletions
Download Patch File Download Diff File Expand all files Collapse all files

317
libs/python/agent/README.md
View File

@@ -29,16 +29,6 @@
```bash
pip install "cua-agent[all]"
# or install specific providers
pip install "cua-agent[openai]" # OpenAI computer-use-preview support
pip install "cua-agent[anthropic]" # Anthropic Claude support
pip install "cua-agent[omni]" # Omniparser + any LLM support
pip install "cua-agent[uitars]" # UI-TARS
pip install "cua-agent[uitars-mlx]" # UI-TARS + MLX support
pip install "cua-agent[uitars-hf]" # UI-TARS + Huggingface support
pip install "cua-agent[glm45v-hf]" # GLM-4.5V + Huggingface support
pip install "cua-agent[ui]" # Gradio UI support
```
## Quick Start
@@ -79,303 +69,18 @@ if __name__ == "__main__":
asyncio.run(main())
```
## Supported Models
## Docs
### Anthropic Claude (Computer Use API)
```python
model="anthropic/claude-3-5-sonnet-20241022"
model="anthropic/claude-3-7-sonnet-20250219"
model="anthropic/claude-opus-4-20250514"
model="anthropic/claude-sonnet-4-20250514"
```
### OpenAI Computer Use Preview
```python
model="openai/computer-use-preview"
```
### UI-TARS (Local or Huggingface Inference)
```python
model="huggingface-local/ByteDance-Seed/UI-TARS-1.5-7B"
model="ollama_chat/0000/ui-tars-1.5-7b"
```
### Omniparser + Any LLM
```python
model="omniparser+ollama_chat/mistral-small3.2"
model="omniparser+vertex_ai/gemini-pro"
model="omniparser+anthropic/claude-3-5-sonnet-20241022"
model="omniparser+openai/gpt-4o"
```
## Custom Tools
Define custom tools using decorated functions:
```python
from computer.helpers import sandboxed
@sandboxed()
def read_file(location: str) -> str:
"""Read contents of a file
Parameters
----------
location : str
Path to the file to read
Returns
-------
str
Contents of the file or error message
"""
try:
with open(location, 'r') as f:
return f.read()
except Exception as e:
return f"Error reading file: {str(e)}"
def calculate(a: int, b: int) -> int:
"""Calculate the sum of two integers"""
return a + b
# Use with agent
agent = ComputerAgent(
model="anthropic/claude-3-5-sonnet-20241022",
tools=[computer, read_file, calculate]
)
```
## Callbacks System
agent provides a comprehensive callback system for extending functionality:
### Built-in Callbacks
```python
from agent.callbacks import (
ImageRetentionCallback,
TrajectorySaverCallback,
BudgetManagerCallback,
LoggingCallback
)
agent = ComputerAgent(
model="anthropic/claude-3-5-sonnet-20241022",
tools=[computer],
callbacks=[
ImageRetentionCallback(only_n_most_recent_images=3),
TrajectorySaverCallback(trajectory_dir="trajectories"),
BudgetManagerCallback(max_budget=10.0, raise_error=True),
LoggingCallback(level=logging.INFO)
]
)
```
### Custom Callbacks
```python
from agent.callbacks.base import AsyncCallbackHandler
class CustomCallback(AsyncCallbackHandler):
async def on_llm_start(self, messages):
"""Preprocess messages before LLM call"""
# Add custom preprocessing logic
return messages
async def on_llm_end(self, messages):
"""Postprocess messages after LLM call"""
# Add custom postprocessing logic
return messages
async def on_usage(self, usage):
"""Track usage information"""
print(f"Tokens used: {usage.total_tokens}")
```
## Budget Management
Control costs with built-in budget management:
```python
# Simple budget limit
agent = ComputerAgent(
model="anthropic/claude-3-5-sonnet-20241022",
max_trajectory_budget=5.0 # $5 limit
)
# Advanced budget configuration
agent = ComputerAgent(
model="anthropic/claude-3-5-sonnet-20241022",
max_trajectory_budget={
"max_budget": 10.0,
"raise_error": True, # Raise error when exceeded
"reset_after_each_run": False # Persistent across runs
}
)
```
## Trajectory Management
Save and replay agent conversations:
```python
agent = ComputerAgent(
model="anthropic/claude-3-5-sonnet-20241022",
trajectory_dir="trajectories", # Auto-save trajectories
tools=[computer]
)
# Trajectories are saved with:
# - Complete conversation history
# - Usage statistics and costs
# - Timestamps and metadata
# - Screenshots and computer actions
```
## Configuration Options
### ComputerAgent Parameters
- `model`: Model identifier (required)
- `tools`: List of computer objects and decorated functions
- `callbacks`: List of callback handlers for extensibility
- `only_n_most_recent_images`: Limit recent images to prevent context overflow
- `verbosity`: Logging level (logging.INFO, logging.DEBUG, etc.)
- `trajectory_dir`: Directory to save conversation trajectories
- `max_retries`: Maximum API call retries (default: 3)
- `screenshot_delay`: Delay between actions and screenshots (default: 0.5s)
- `use_prompt_caching`: Enable prompt caching for supported models
- `max_trajectory_budget`: Budget limit configuration
### Environment Variables
```bash
# Computer instance (cloud)
export CUA_CONTAINER_NAME="your-container-name"
export CUA_API_KEY="your-cua-api-key"
# LLM API keys
export ANTHROPIC_API_KEY="your-anthropic-key"
export OPENAI_API_KEY="your-openai-key"
```
## Advanced Usage
### Streaming Responses
```python
async for result in agent.run(messages, stream=True):
# Process streaming chunks
for item in result["output"]:
if item["type"] == "message":
print(item["content"][0]["text"], end="", flush=True)
elif item["type"] == "computer_call":
action = item["action"]
print(f"\n[Action: {action['type']}]")
```
### Interactive Chat Loop
```python
history = []
while True:
user_input = input("> ")
if user_input.lower() in ['quit', 'exit']:
break
history.append({"role": "user", "content": user_input})
async for result in agent.run(history):
history += result["output"]
# Display assistant responses
for item in result["output"]:
if item["type"] == "message":
print(item["content"][0]["text"])
```
### Error Handling
```python
try:
async for result in agent.run(messages):
# Process results
pass
except BudgetExceededException:
print("Budget limit exceeded")
except Exception as e:
print(f"Agent error: {e}")
```
## API Reference
### ComputerAgent.run()
```python
async def run(
self,
messages: Messages,
stream: bool = False,
**kwargs
) -> AsyncGenerator[Dict[str, Any], None]:
"""
Run the agent with the given messages.
Args:
messages: List of message dictionaries
stream: Whether to stream the response
**kwargs: Additional arguments
Returns:
AsyncGenerator that yields response chunks
"""
```
### Message Format
```python
messages = [
{
"role": "user",
"content": "Take a screenshot and describe what you see"
},
{
"role": "assistant",
"content": "I'll take a screenshot for you."
}
]
```
### Response Format
```python
{
"output": [
{
"type": "message",
"role": "assistant",
"content": [{"type": "output_text", "text": "I can see..."}]
},
{
"type": "computer_call",
"action": {"type": "screenshot"},
"call_id": "call_123"
},
{
"type": "computer_call_output",
"call_id": "call_123",
"output": {"image_url": "data:image/png;base64,..."}
}
],
"usage": {
"prompt_tokens": 150,
"completion_tokens": 75,
"total_tokens": 225,
"response_cost": 0.01,
}
}
```
- [Agent Loops](https://trycua.com/docs/agent-sdk/agent-loops)
- [Supported Agents](https://trycua.com/docs/agent-sdk/supported-agents)
- [Supported Models](https://trycua.com/docs/agent-sdk/supported-models)
- [Chat History](https://trycua.com/docs/agent-sdk/chat-history)
- [Callbacks](https://trycua.com/docs/agent-sdk/callbacks)
- [Custom Tools](https://trycua.com/docs/agent-sdk/custom-tools)
- [Custom Computer Handlers](https://trycua.com/docs/agent-sdk/custom-computer-handlers)
- [Prompt Caching](https://trycua.com/docs/agent-sdk/prompt-caching)
- [Usage Tracking](https://trycua.com/docs/agent-sdk/usage-tracking)
- [Benchmarks](https://trycua.com/docs/agent-sdk/benchmarks)
## License

2
libs/python/agent/agent/adapters/__init__.py
View File

@@ -4,8 +4,10 @@ Adapters package for agent - Custom LLM adapters for LiteLLM
from .huggingfacelocal_adapter import HuggingFaceLocalAdapter
from .human_adapter import HumanAdapter
from .mlxvlm_adapter import MLXVLMAdapter
__all__ = [
"HuggingFaceLocalAdapter",
"HumanAdapter",
"MLXVLMAdapter",
]

359
libs/python/agent/agent/adapters/mlxvlm_adapter.py Normal file
View File

@@ -0,0 +1,359 @@
import asyncio
import functools
import warnings
import io
import base64
import math
import re
from concurrent.futures import ThreadPoolExecutor
from typing import Iterator, AsyncIterator, Dict, List, Any, Optional, Tuple, cast
from PIL import Image
from litellm.types.utils import GenericStreamingChunk, ModelResponse
from litellm.llms.custom_llm import CustomLLM
from litellm import completion, acompletion
# Try to import MLX dependencies
try:
import mlx.core as mx
from mlx_vlm import load, generate
from mlx_vlm.prompt_utils import apply_chat_template
from mlx_vlm.utils import load_config
from transformers.tokenization_utils import PreTrainedTokenizer
MLX_AVAILABLE = True
except ImportError:
MLX_AVAILABLE = False
# Constants for smart_resize
IMAGE_FACTOR = 28
MIN_PIXELS = 100 * 28 * 28
MAX_PIXELS = 16384 * 28 * 28
MAX_RATIO = 200
def round_by_factor(number: float, factor: int) -> int:
"""Returns the closest integer to 'number' that is divisible by 'factor'."""
return round(number / factor) * factor
def ceil_by_factor(number: float, factor: int) -> int:
"""Returns the smallest integer greater than or equal to 'number' that is divisible by 'factor'."""
return math.ceil(number / factor) * factor
def floor_by_factor(number: float, factor: int) -> int:
"""Returns the largest integer less than or equal to 'number' that is divisible by 'factor'."""
return math.floor(number / factor) * factor
def smart_resize(
height: int, width: int, factor: int = IMAGE_FACTOR, min_pixels: int = MIN_PIXELS, max_pixels: int = MAX_PIXELS
) -> tuple[int, int]:
"""
Rescales the image so that the following conditions are met:
1. Both dimensions (height and width) are divisible by 'factor'.
2. The total number of pixels is within the range ['min_pixels', 'max_pixels'].
3. The aspect ratio of the image is maintained as closely as possible.
"""
if max(height, width) / min(height, width) > MAX_RATIO:
raise ValueError(
f"absolute aspect ratio must be smaller than {MAX_RATIO}, got {max(height, width) / min(height, width)}"
)
h_bar = max(factor, round_by_factor(height, factor))
w_bar = max(factor, round_by_factor(width, factor))
if h_bar * w_bar > max_pixels:
beta = math.sqrt((height * width) / max_pixels)
h_bar = floor_by_factor(height / beta, factor)
w_bar = floor_by_factor(width / beta, factor)
elif h_bar * w_bar < min_pixels:
beta = math.sqrt(min_pixels / (height * width))
h_bar = ceil_by_factor(height * beta, factor)
w_bar = ceil_by_factor(width * beta, factor)
return h_bar, w_bar
class MLXVLMAdapter(CustomLLM):
"""MLX VLM Adapter for running vision-language models locally using MLX."""
def __init__(self, **kwargs):
"""Initialize the adapter.
Args:
**kwargs: Additional arguments
"""
super().__init__()
self.models = {} # Cache for loaded models
self.processors = {} # Cache for loaded processors
self.configs = {} # Cache for loaded configs
self._executor = ThreadPoolExecutor(max_workers=1) # Single thread pool
def _load_model_and_processor(self, model_name: str):
"""Load model and processor if not already cached.
Args:
model_name: Name of the model to load
Returns:
Tuple of (model, processor, config)
"""
if not MLX_AVAILABLE:
raise ImportError("MLX VLM dependencies not available. Please install mlx-vlm.")
if model_name not in self.models:
# Load model and processor
model_obj, processor = load(
model_name,
processor_kwargs={"min_pixels": MIN_PIXELS, "max_pixels": MAX_PIXELS}
)
config = load_config(model_name)
# Cache them
self.models[model_name] = model_obj
self.processors[model_name] = processor
self.configs[model_name] = config
return self.models[model_name], self.processors[model_name], self.configs[model_name]
def _process_coordinates(self, text: str, original_size: Tuple[int, int], model_size: Tuple[int, int]) -> str:
"""Process coordinates in box tokens based on image resizing using smart_resize approach.
Args:
text: Text containing box tokens
original_size: Original image size (width, height)
model_size: Model processed image size (width, height)
Returns:
Text with processed coordinates
"""
# Find all box tokens
box_pattern = r"<\|box_start\|>\((\d+),\s*(\d+)\)<\|box_end\|>"
def process_coords(match):
model_x, model_y = int(match.group(1)), int(match.group(2))
# Scale coordinates from model space to original image space
# Both original_size and model_size are in (width, height) format
new_x = int(model_x * original_size[0] / model_size[0]) # Width
new_y = int(model_y * original_size[1] / model_size[1]) # Height
return f"<|box_start|>({new_x},{new_y})<|box_end|>"
return re.sub(box_pattern, process_coords, text)
def _convert_messages(self, messages: List[Dict[str, Any]]) -> Tuple[List[Dict[str, Any]], List[Image.Image], Dict[int, Tuple[int, int]], Dict[int, Tuple[int, int]]]:
"""Convert OpenAI format messages to MLX VLM format and extract images.
Args:
messages: Messages in OpenAI format
Returns:
Tuple of (processed_messages, images, original_sizes, model_sizes)
"""
processed_messages = []
images = []
original_sizes = {} # Track original sizes of images for coordinate mapping
model_sizes = {} # Track model processed sizes
image_index = 0
for message in messages:
processed_message = {
"role": message["role"],
"content": []
}
content = message.get("content", [])
if isinstance(content, str):
# Simple text content
processed_message["content"] = content
elif isinstance(content, list):
# Multi-modal content
processed_content = []
for item in content:
if item.get("type") == "text":
processed_content.append({
"type": "text",
"text": item.get("text", "")
})
elif item.get("type") == "image_url":
image_url = item.get("image_url", {}).get("url", "")
pil_image = None
if image_url.startswith("data:image/"):
# Extract base64 data
base64_data = image_url.split(',')[1]
# Convert base64 to PIL Image
image_data = base64.b64decode(base64_data)
pil_image = Image.open(io.BytesIO(image_data))
else:
# Handle file path or URL
pil_image = Image.open(image_url)
# Store original image size for coordinate mapping
original_size = pil_image.size
original_sizes[image_index] = original_size
# Use smart_resize to determine model size
# Note: smart_resize expects (height, width) but PIL gives (width, height)
height, width = original_size[1], original_size[0]
new_height, new_width = smart_resize(height, width)
# Store model size in (width, height) format for consistent coordinate processing
model_sizes[image_index] = (new_width, new_height)
# Resize the image using the calculated dimensions from smart_resize
resized_image = pil_image.resize((new_width, new_height))
images.append(resized_image)
# Add image placeholder to content
processed_content.append({
"type": "image"
})
image_index += 1
processed_message["content"] = processed_content
processed_messages.append(processed_message)
return processed_messages, images, original_sizes, model_sizes
def _generate(self, **kwargs) -> str:
"""Generate response using the local MLX VLM model.
Args:
**kwargs: Keyword arguments containing messages and model info
Returns:
Generated text response
"""
messages = kwargs.get('messages', [])
model_name = kwargs.get('model', 'mlx-community/UI-TARS-1.5-7B-4bit')
max_tokens = kwargs.get('max_tokens', 128)
# Warn about ignored kwargs
ignored_kwargs = set(kwargs.keys()) - {'messages', 'model', 'max_tokens'}
if ignored_kwargs:
warnings.warn(f"Ignoring unsupported kwargs: {ignored_kwargs}")
# Load model and processor
model, processor, config = self._load_model_and_processor(model_name)
# Convert messages and extract images
processed_messages, images, original_sizes, model_sizes = self._convert_messages(messages)
# Process user text input with box coordinates after image processing
# Swap original_size and model_size arguments for inverse transformation
for msg_idx, msg in enumerate(processed_messages):
if msg.get("role") == "user" and isinstance(msg.get("content"), str):
content = msg.get("content", "")
if "<|box_start|>" in content and original_sizes and model_sizes and 0 in original_sizes and 0 in model_sizes:
orig_size = original_sizes[0]
model_size = model_sizes[0]
# Swap arguments to perform inverse transformation for user input
processed_messages[msg_idx]["content"] = self._process_coordinates(content, model_size, orig_size)
try:
# Format prompt according to model requirements using the processor directly
prompt = processor.apply_chat_template(
processed_messages,
tokenize=False,
add_generation_prompt=True,
return_tensors='pt'
)
tokenizer = cast(PreTrainedTokenizer, processor)
# Generate response
text_content, usage = generate(
model,
tokenizer,
str(prompt),
images, # type: ignore
verbose=False,
max_tokens=max_tokens
)
except Exception as e:
raise RuntimeError(f"Error generating response: {str(e)}") from e
# Process coordinates in the response back to original image space
if original_sizes and model_sizes and 0 in original_sizes and 0 in model_sizes:
# Get original image size and model size (using the first image)
orig_size = original_sizes[0]
model_size = model_sizes[0]
# Check if output contains box tokens that need processing
if "<|box_start|>" in text_content:
# Process coordinates from model space back to original image space
text_content = self._process_coordinates(text_content, orig_size, model_size)
return text_content
def completion(self, *args, **kwargs) -> ModelResponse:
"""Synchronous completion method.
Returns:
ModelResponse with generated text
"""
generated_text = self._generate(**kwargs)
result = completion(
model=f"mlx/{kwargs.get('model', 'mlx-community/UI-TARS-1.5-7B-4bit')}",
mock_response=generated_text,
)
return cast(ModelResponse, result)
async def acompletion(self, *args, **kwargs) -> ModelResponse:
"""Asynchronous completion method.
Returns:
ModelResponse with generated text
"""
# Run _generate in thread pool to avoid blocking
loop = asyncio.get_event_loop()
generated_text = await loop.run_in_executor(
self._executor,
functools.partial(self._generate, **kwargs)
)
result = await acompletion(
model=f"mlx/{kwargs.get('model', 'mlx-community/UI-TARS-1.5-7B-4bit')}",
mock_response=generated_text,
)
return cast(ModelResponse, result)
def streaming(self, *args, **kwargs) -> Iterator[GenericStreamingChunk]:
"""Synchronous streaming method.
Returns:
Iterator of GenericStreamingChunk
"""
generated_text = self._generate(**kwargs)
generic_streaming_chunk: GenericStreamingChunk = {
"finish_reason": "stop",
"index": 0,
"is_finished": True,
"text": generated_text,
"tool_use": None,
"usage": {"completion_tokens": 0, "prompt_tokens": 0, "total_tokens": 0},
}
yield generic_streaming_chunk
async def astreaming(self, *args, **kwargs) -> AsyncIterator[GenericStreamingChunk]:
"""Asynchronous streaming method.
Returns:
AsyncIterator of GenericStreamingChunk
"""
# Run _generate in thread pool to avoid blocking
loop = asyncio.get_event_loop()
generated_text = await loop.run_in_executor(
self._executor,
functools.partial(self._generate, **kwargs)
)
generic_streaming_chunk: GenericStreamingChunk = {
"finish_reason": "stop",
"index": 0,
"is_finished": True,
"text": generated_text,
"tool_use": None,
"usage": {"completion_tokens": 0, "prompt_tokens": 0, "total_tokens": 0},
}
yield generic_streaming_chunk

23
libs/python/agent/agent/agent.py
View File

@@ -3,6 +3,7 @@ ComputerAgent - Main agent class that selects and runs agent loops
"""
import asyncio
from pathlib import Path
from typing import Dict, List, Any, Optional, AsyncGenerator, Union, cast, Callable, Set, Tuple
from litellm.responses.utils import Usage
@@ -22,6 +23,7 @@ import inspect
from .adapters import (
HuggingFaceLocalAdapter,
HumanAdapter,
MLXVLMAdapter,
)
from .callbacks import (
ImageRetentionCallback,
@@ -29,6 +31,7 @@ from .callbacks import (
TrajectorySaverCallback,
BudgetManagerCallback,
TelemetryCallback,
OperatorNormalizerCallback
)
from .computers import (
AsyncComputerHandler,
@@ -160,7 +163,7 @@ class ComputerAgent:
only_n_most_recent_images: Optional[int] = None,
callbacks: Optional[List[Any]] = None,
verbosity: Optional[int] = None,
trajectory_dir: Optional[str] = None,
trajectory_dir: Optional[str | Path | dict] = None,
max_retries: Optional[int] = 3,
screenshot_delay: Optional[float | int] = 0.5,
use_prompt_caching: Optional[bool] = False,
@@ -187,7 +190,11 @@ class ComputerAgent:
telemetry_enabled: If set, adds TelemetryCallback to track anonymized usage data. Enabled by default.
trust_remote_code: If set, trust remote code when loading local models. Disabled by default.
**kwargs: Additional arguments passed to the agent loop
"""
"""
# If the loop is "human/human", we need to prefix a grounding model fallback
if model in ["human/human", "human"]:
model = "openai/computer-use-preview+human/human"
self.model = model
self.tools = tools or []
self.custom_loop = custom_loop
@@ -204,6 +211,9 @@ class ComputerAgent:
# == Add built-in callbacks ==
# Prepend operator normalizer callback
self.callbacks.insert(0, OperatorNormalizerCallback())
# Add telemetry callback if telemetry_enabled is set
if self.telemetry_enabled:
if isinstance(self.telemetry_enabled, bool):
@@ -221,7 +231,10 @@ class ComputerAgent:
# Add trajectory saver callback if trajectory_dir is set
if self.trajectory_dir:
self.callbacks.append(TrajectorySaverCallback(self.trajectory_dir))
if isinstance(self.trajectory_dir, dict):
self.callbacks.append(TrajectorySaverCallback(**self.trajectory_dir))
elif isinstance(self.trajectory_dir, (str, Path)):
self.callbacks.append(TrajectorySaverCallback(str(self.trajectory_dir)))
# Add budget manager if max_trajectory_budget is set
if max_trajectory_budget:
@@ -238,9 +251,11 @@ class ComputerAgent:
trust_remote_code=self.trust_remote_code or False
)
human_adapter = HumanAdapter()
mlx_adapter = MLXVLMAdapter()
litellm.custom_provider_map = [
{"provider": "huggingface-local", "custom_handler": hf_adapter},
{"provider": "human", "custom_handler": human_adapter}
{"provider": "human", "custom_handler": human_adapter},
{"provider": "mlx", "custom_handler": mlx_adapter}
]
litellm.suppress_debug_info = True

2
libs/python/agent/agent/callbacks/__init__.py
View File

@@ -8,6 +8,7 @@ from .logging import LoggingCallback
from .trajectory_saver import TrajectorySaverCallback
from .budget_manager import BudgetManagerCallback
from .telemetry import TelemetryCallback
from .operator_validator import OperatorNormalizerCallback
__all__ = [
"AsyncCallbackHandler",
@@ -16,4 +17,5 @@ __all__ = [
"TrajectorySaverCallback",
"BudgetManagerCallback",
"TelemetryCallback",
"OperatorNormalizerCallback",
]

125
libs/python/agent/agent/callbacks/image_retention.py
View File

@@ -50,90 +50,41 @@ class ImageRetentionCallback(AsyncCallbackHandler):
"""
if self.only_n_most_recent_images is None:
return messages
# First pass: Assign call_id to reasoning items based on the next computer_call
messages_with_call_ids = []
for i, msg in enumerate(messages):
msg_copy = msg.copy() if isinstance(msg, dict) else msg
# If this is a reasoning item without a call_id, find the next computer_call
if (msg_copy.get("type") == "reasoning" and
not msg_copy.get("call_id")):
# Look ahead for the next computer_call
for j in range(i + 1, len(messages)):
next_msg = messages[j]
if (next_msg.get("type") == "computer_call" and
next_msg.get("call_id")):
msg_copy["call_id"] = next_msg.get("call_id")
break
messages_with_call_ids.append(msg_copy)
# Find all computer_call_output items with images and their call_ids
image_call_ids = []
for msg in reversed(messages_with_call_ids): # Process in reverse to get most recent first
if (msg.get("type") == "computer_call_output" and
isinstance(msg.get("output"), dict) and
"image_url" in msg.get("output", {})):
call_id = msg.get("call_id")
if call_id and call_id not in image_call_ids:
image_call_ids.append(call_id)
if len(image_call_ids) >= self.only_n_most_recent_images:
break
# Keep the most recent N image call_ids (reverse to get chronological order)
keep_call_ids = set(image_call_ids[:self.only_n_most_recent_images])
# Filter messages: remove computer_call, computer_call_output, and reasoning for old images
filtered_messages = []
for msg in messages_with_call_ids:
msg_type = msg.get("type")
call_id = msg.get("call_id")
# Remove old computer_call items
if msg_type == "computer_call" and call_id not in keep_call_ids:
# Check if this call_id corresponds to an image call
has_image_output = any(
m.get("type") == "computer_call_output" and
m.get("call_id") == call_id and
isinstance(m.get("output"), dict) and
"image_url" in m.get("output", {})
for m in messages_with_call_ids
)
if has_image_output:
continue # Skip this computer_call
# Remove old computer_call_output items with images
if (msg_type == "computer_call_output" and
call_id not in keep_call_ids and
isinstance(msg.get("output"), dict) and
"image_url" in msg.get("output", {})):
continue # Skip this computer_call_output
# Remove old reasoning items that are paired with removed computer calls
if (msg_type == "reasoning" and
call_id and call_id not in keep_call_ids):
# Check if this call_id corresponds to an image call that's being removed
has_image_output = any(
m.get("type") == "computer_call_output" and
m.get("call_id") == call_id and
isinstance(m.get("output"), dict) and
"image_url" in m.get("output", {})
for m in messages_with_call_ids
)
if has_image_output:
continue # Skip this reasoning item
filtered_messages.append(msg)
# Clean up: Remove call_id from reasoning items before returning
final_messages = []
for msg in filtered_messages:
if msg.get("type") == "reasoning" and "call_id" in msg:
# Create a copy without call_id for reasoning items
cleaned_msg = {k: v for k, v in msg.items() if k != "call_id"}
final_messages.append(cleaned_msg)
else:
final_messages.append(msg)
return final_messages
# Gather indices of all computer_call_output messages that contain an image_url
output_indices: List[int] = []
for idx, msg in enumerate(messages):
if msg.get("type") == "computer_call_output":
out = msg.get("output")
if isinstance(out, dict) and ("image_url" in out):
output_indices.append(idx)
# Nothing to trim
if len(output_indices) <= self.only_n_most_recent_images:
return messages
# Determine which outputs to keep (most recent N)
keep_output_indices = set(output_indices[-self.only_n_most_recent_images :])
# Build set of indices to remove in one pass
to_remove: set[int] = set()
for idx in output_indices:
if idx in keep_output_indices:
continue # keep this screenshot and its context
to_remove.add(idx) # remove the computer_call_output itself
# Remove the immediately preceding computer_call with matching call_id (if present)
call_id = messages[idx].get("call_id")
prev_idx = idx - 1
if prev_idx >= 0 and messages[prev_idx].get("type") == "computer_call" and messages[prev_idx].get("call_id") == call_id:
to_remove.add(prev_idx)
# Check a single reasoning immediately before that computer_call
r_idx = prev_idx - 1
if r_idx >= 0 and messages[r_idx].get("type") == "reasoning":
to_remove.add(r_idx)
# Construct filtered list
filtered = [m for i, m in enumerate(messages) if i not in to_remove]
return filtered

138
libs/python/agent/agent/callbacks/operator_validator.py Normal file
View File

@@ -0,0 +1,138 @@
"""
OperatorValidatorCallback
Ensures agent output actions conform to expected schemas by fixing common issues:
- click: add default button='left' if missing
- keypress: wrap keys string into a list
- etc.
This runs in on_llm_end, which receives the output array (AgentMessage[] as dicts).
The purpose is to avoid spending another LLM call to fix broken computer call syntax when possible.
"""
from __future__ import annotations
from typing import Any, Dict, List
from .base import AsyncCallbackHandler
class OperatorNormalizerCallback(AsyncCallbackHandler):
"""Normalizes common computer call hallucinations / errors in computer call syntax."""
async def on_llm_end(self, output: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
# Mutate in-place as requested, but still return the list for chaining
for item in output or []:
if item.get("type") != "computer_call":
continue
action = item.get("action")
if not isinstance(action, dict):
continue
# rename mouse click actions to "click"
for mouse_btn in ["left", "right", "wheel", "back", "forward"]:
if action.get("type", "") == f"{mouse_btn}_click":
action["type"] = "click"
action["button"] = mouse_btn
# rename hotkey actions to "keypress"
for alias in ["hotkey", "key", "press", "key_press"]:
if action.get("type", "") == alias:
action["type"] = "keypress"
# assume click actions
if "button" in action and "type" not in action:
action["type"] = "click"
if "click" in action and "type" not in action:
action["type"] = "click"
if ("scroll_x" in action or "scroll_y" in action) and "type" not in action:
action["type"] = "scroll"
if "text" in action and "type" not in action:
action["type"] = "type"
action_type = action.get("type")
def _keep_keys(action: Dict[str, Any], keys_to_keep: List[str]):
"""Keep only the provided keys on action; delete everything else.
Always ensures required 'type' is present if listed in keys_to_keep.
"""
for key in list(action.keys()):
if key not in keys_to_keep:
del action[key]
# rename "coordinate" to "x", "y"
if "coordinate" in action:
action["x"] = action["coordinate"][0]
action["y"] = action["coordinate"][1]
del action["coordinate"]
if action_type == "click":
# convert "click" to "button"
if "button" not in action and "click" in action:
action["button"] = action["click"]
del action["click"]
# default button to "left"
action["button"] = action.get("button", "left")
# add default scroll x, y if missing
if action_type == "scroll":
action["scroll_x"] = action.get("scroll_x", 0)
action["scroll_y"] = action.get("scroll_y", 0)
# ensure keys arg is a list (normalize aliases first)
if action_type == "keypress":
keys = action.get("keys")
for keys_alias in ["keypress", "key", "press", "key_press", "text"]:
if keys_alias in action:
action["keys"] = action[keys_alias]
del action[keys_alias]
keys = action.get("keys")
if isinstance(keys, str):
action["keys"] = keys.replace("-", "+").split("+") if len(keys) > 1 else [keys]
required_keys_by_type = {
# OpenAI actions
"click": ["type", "button", "x", "y"],
"double_click": ["type", "x", "y"],
"drag": ["type", "path"],
"keypress": ["type", "keys"],
"move": ["type", "x", "y"],
"screenshot": ["type"],
"scroll": ["type", "scroll_x", "scroll_y", "x", "y"],
"type": ["type", "text"],
"wait": ["type"],
# Anthropic actions
"left_mouse_down": ["type", "x", "y"],
"left_mouse_up": ["type", "x", "y"],
"triple_click": ["type", "button", "x", "y"],
}
keep = required_keys_by_type.get(action_type or "")
if keep:
_keep_keys(action, keep)
# # Second pass: if an assistant message is immediately followed by a computer_call,
# # replace the assistant message itself with a reasoning message with summary text.
# if isinstance(output, list):
# for i, item in enumerate(output):
# # AssistantMessage shape: { type: 'message', role: 'assistant', content: OutputContent[] }
# if item.get("type") == "message" and item.get("role") == "assistant":
# next_idx = i + 1
# if next_idx >= len(output):
# continue
# next_item = output[next_idx]
# if not isinstance(next_item, dict):
# continue
# if next_item.get("type") != "computer_call":
# continue
# contents = item.get("content") or []
# # Extract text from OutputContent[]
# text_parts: List[str] = []
# if isinstance(contents, list):
# for c in contents:
# if isinstance(c, dict) and c.get("type") == "output_text" and isinstance(c.get("text"), str):
# text_parts.append(c["text"])
# text_content = "\n".join(text_parts).strip()
# # Replace assistant message with reasoning message
# output[i] = {
# "type": "reasoning",
# "summary": [
# {
# "type": "summary_text",
# "text": text_content,
# }
# ],
# }
return output

92
libs/python/agent/agent/callbacks/trajectory_saver.py
View File

@@ -11,6 +11,8 @@ from pathlib import Path
from typing import List, Dict, Any, Optional, Union, override
from PIL import Image, ImageDraw
import io
from copy import deepcopy
from .base import AsyncCallbackHandler
def sanitize_image_urls(data: Any) -> Any:
@@ -43,6 +45,64 @@ def sanitize_image_urls(data: Any) -> Any:
return data
def extract_computer_call_outputs(items: List[Dict[str, Any]], screenshot_dir: Optional[Path]) -> List[Dict[str, Any]]:
"""
Save any base64-encoded screenshots from computer_call_output entries to files and
replace their image_url with the saved file path when a call_id is present.
Only operates if screenshot_dir is provided and exists; otherwise returns items unchanged.
Args:
items: List of message/result dicts potentially containing computer_call_output entries
screenshot_dir: Directory to write screenshots into
Returns:
A new list with updated image_url fields when applicable.
"""
if not items:
return items
if not screenshot_dir or not screenshot_dir.exists():
return items
updated: List[Dict[str, Any]] = []
for item in items:
# work on a shallow copy; deep copy nested 'output' if we modify it
msg = dict(item)
try:
if msg.get("type") == "computer_call_output":
call_id = msg.get("call_id")
output = msg.get("output", {})
image_url = output.get("image_url")
if call_id and isinstance(image_url, str) and image_url.startswith("data:"):
# derive extension from MIME type e.g. data:image/png;base64,
try:
ext = image_url.split(";", 1)[0].split("/")[-1]
if not ext:
ext = "png"
except Exception:
ext = "png"
out_path = screenshot_dir / f"{call_id}.{ext}"
# write file if it doesn't exist
if not out_path.exists():
try:
b64_payload = image_url.split(",", 1)[1]
img_bytes = base64.b64decode(b64_payload)
out_path.parent.mkdir(parents=True, exist_ok=True)
with open(out_path, "wb") as f:
f.write(img_bytes)
except Exception:
# if anything fails, skip modifying this message
pass
# update image_url to file path
new_output = dict(output)
new_output["image_url"] = str(out_path)
msg["output"] = new_output
except Exception:
# do not block on malformed entries; keep original
pass
updated.append(msg)
return updated
class TrajectorySaverCallback(AsyncCallbackHandler):
"""
Callback handler that saves agent trajectories to disk.
@@ -51,7 +111,7 @@ class TrajectorySaverCallback(AsyncCallbackHandler):
within the trajectory gets its own folder with screenshots and responses.
"""
def __init__(self, trajectory_dir: str, reset_on_run: bool = True):
def __init__(self, trajectory_dir: str, reset_on_run: bool = True, screenshot_dir: Optional[str] = None):
"""
Initialize trajectory saver.
@@ -67,10 +127,12 @@ class TrajectorySaverCallback(AsyncCallbackHandler):
self.model: Optional[str] = None
self.total_usage: Dict[str, Any] = {}
self.reset_on_run = reset_on_run
# Optional directory to store extracted screenshots from metadata/new_items
self.screenshot_dir: Optional[Path] = Path(screenshot_dir) if screenshot_dir else None
# Ensure trajectory directory exists
self.trajectory_dir.mkdir(parents=True, exist_ok=True)
def _get_turn_dir(self) -> Path:
"""Get the directory for the current turn."""
if not self.trajectory_id:
@@ -94,6 +156,10 @@ class TrajectorySaverCallback(AsyncCallbackHandler):
# format: turn_000/0000_name.json
artifact_filename = f"{self.current_artifact:04d}_{name}"
artifact_path = turn_dir / f"{artifact_filename}.json"
# add created_at
if isinstance(artifact, dict):
artifact = artifact.copy()
artifact["created_at"] = str(uuid.uuid1().time)
with open(artifact_path, "w") as f:
json.dump(sanitize_image_urls(artifact), f, indent=2)
self.current_artifact += 1
@@ -135,12 +201,21 @@ class TrajectorySaverCallback(AsyncCallbackHandler):
trajectory_path = self.trajectory_dir / self.trajectory_id
trajectory_path.mkdir(parents=True, exist_ok=True)
# Save trajectory metadata
# Save trajectory metadata (optionally extract screenshots to screenshot_dir)
kwargs_to_save = kwargs.copy()
try:
if "messages" in kwargs_to_save:
kwargs_to_save["messages"] = extract_computer_call_outputs(
kwargs_to_save["messages"], self.screenshot_dir
)
except Exception:
# If extraction fails, fall back to original messages
pass
metadata = {
"trajectory_id": self.trajectory_id,
"created_at": str(uuid.uuid1().time),
"status": "running",
"kwargs": kwargs,
"kwargs": kwargs_to_save,
}
with open(trajectory_path / "metadata.json", "w") as f:
@@ -167,11 +242,18 @@ class TrajectorySaverCallback(AsyncCallbackHandler):
metadata = {}
# Update metadata with completion info
# Optionally extract screenshots from new_items before persisting
new_items_to_save = new_items
try:
new_items_to_save = extract_computer_call_outputs(new_items, self.screenshot_dir)
except Exception:
pass
metadata.update({
"status": "completed",
"completed_at": str(uuid.uuid1().time),
"total_usage": self.total_usage,
"new_items": sanitize_image_urls(new_items),
"new_items": new_items_to_save,
"total_turns": self.current_turn
})

296
libs/python/agent/agent/human_tool/ui.py
View File

@@ -15,6 +15,11 @@ class HumanCompletionUI:
self.current_call_id: Optional[str] = None
self.refresh_interval = 2.0 # seconds
self.last_image = None # Store the last image for display
# Track current interactive action controls
self.current_action_type: str = "click"
self.current_button: str = "left"
self.current_scroll_x: int = 0
self.current_scroll_y: int = -120
def format_messages_for_chatbot(self, messages: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
"""Format messages for display in gr.Chatbot with type='messages'."""
@@ -196,7 +201,9 @@ class HumanCompletionUI:
gr.update(choices=["latest"], value="latest"), # dropdown
gr.update(value=None), # image (no image)
gr.update(value=[]), # chatbot (empty messages)
gr.update(interactive=False) # submit button
gr.update(interactive=False), # submit button
gr.update(visible=False), # click_actions_group hidden
gr.update(visible=False), # actions_group hidden
)
# Sort pending calls by created_at to get oldest first
@@ -237,7 +244,9 @@ class HumanCompletionUI:
gr.update(choices=choices, value="latest"),
gr.update(value=self.last_image),
gr.update(value=conversation),
gr.update(interactive=bool(choices))
gr.update(interactive=bool(choices)),
gr.update(visible=True), # click_actions_group visible when there is a call
gr.update(visible=True), # actions_group visible when there is a call
)
def on_call_selected(self, selected_choice):
@@ -246,7 +255,9 @@ class HumanCompletionUI:
return (
gr.update(value=None), # no image
gr.update(value=[]), # empty chatbot
gr.update(interactive=False)
gr.update(interactive=False),
gr.update(visible=False), # click_actions_group hidden
gr.update(visible=False), # actions_group hidden
)
pending_calls = self.get_pending_calls()
@@ -254,7 +265,9 @@ class HumanCompletionUI:
return (
gr.update(value=None), # no image
gr.update(value=[]), # empty chatbot
gr.update(interactive=False)
gr.update(interactive=False),
gr.update(visible=False), # click_actions_group hidden
gr.update(visible=False), # actions_group hidden
)
# Handle "latest" option
@@ -286,7 +299,9 @@ class HumanCompletionUI:
return (
gr.update(value=None), # no image
gr.update(value=[]), # empty chatbot
gr.update(interactive=False)
gr.update(interactive=False),
gr.update(visible=False), # click_actions_group hidden
gr.update(visible=False), # actions_group hidden
)
conversation = self.format_messages_for_chatbot(selected_call.get("messages", []))
@@ -297,7 +312,9 @@ class HumanCompletionUI:
return (
gr.update(value=self.last_image),
gr.update(value=conversation),
gr.update(interactive=True)
gr.update(interactive=True),
gr.update(visible=True), # click_actions_group visible
gr.update(visible=True), # actions_group visible
)
def submit_response(self, response_text: str):
@@ -368,6 +385,10 @@ class HumanCompletionUI:
"""Submit a hotkey action."""
return self.submit_action("keypress", keys=keys)
def submit_wait_action(self) -> str:
"""Submit a wait action with no kwargs."""
return self.submit_action("wait")
def submit_description_click(self, description: str, action_type: str = "click", button: str = "left") -> str:
"""Submit a description-based action."""
if action_type == "click":
@@ -407,7 +428,7 @@ def create_ui():
"""Create the Gradio interface."""
ui_handler = HumanCompletionUI()
with gr.Blocks(title="Human-in-the-Loop Agent Tool") as demo:
with gr.Blocks(title="Human-in-the-Loop Agent Tool", fill_width=True) as demo:
gr.Markdown("# 🤖 Human-in-the-Loop Agent Tool")
gr.Markdown("Review AI conversation requests and provide human responses.")
@@ -415,29 +436,42 @@ def create_ui():
with gr.Column(scale=2):
with gr.Group():
screenshot_image = gr.Image(
label="Screenshot",
label="Interactive Screenshot",
interactive=False,
height=600
)
# Action type selection for image clicks
with gr.Row():
action_type_radio = gr.Radio(
label="Action Type",
choices=["click", "double_click", "move", "left_mouse_up", "left_mouse_down"],
value="click",
scale=2
)
action_button_radio = gr.Radio(
label="Button (for click only)",
choices=["left", "right", "wheel", "back", "forward"],
value="left",
visible=True,
scale=1
)
# Action type selection for image clicks (wrapped for visibility control)
with gr.Group(visible=False) as click_actions_group:
with gr.Row():
action_type_radio = gr.Dropdown(
label="Interactive Action",
choices=["click", "double_click", "move", "left_mouse_up", "left_mouse_down", "scroll"],
value="click",
scale=2
)
action_button_radio = gr.Dropdown(
label="Button",
choices=["left", "right", "wheel", "back", "forward"],
value="left",
visible=True,
scale=1
)
scroll_x_input = gr.Number(
label="scroll_x",
value=0,
visible=False,
scale=1
)
scroll_y_input = gr.Number(
label="scroll_y",
value=-120,
visible=False,
scale=1
)
conversation_chatbot = gr.Chatbot(
label="Messages",
label="Conversation",
type="messages",
height=500,
show_copy_button=True
@@ -446,99 +480,97 @@ def create_ui():
with gr.Column(scale=1):
with gr.Group():
call_dropdown = gr.Dropdown(
label="Select a pending call",
label="Select a pending conversation request",
choices=["latest"],
interactive=True,
value="latest"
)
refresh_btn = gr.Button("🔄 Refresh", variant="secondary")
status_display = gr.Textbox(
label="Status",
interactive=False,
value="Ready to receive requests..."
)
with gr.Group():
response_text = gr.Textbox(
label="Response",
label="Message",
lines=3,
placeholder="Enter your response here..."
placeholder="Enter your message here..."
)
submit_btn = gr.Button("📤 Submit Response", variant="primary", interactive=False)
submit_btn = gr.Button("📤 Submit Message", variant="primary", interactive=False)
# Action Accordions
with gr.Accordion("🖱️ Click Actions", open=False):
with gr.Group():
with gr.Row():
click_x = gr.Number(label="X", value=0, minimum=0)
click_y = gr.Number(label="Y", value=0, minimum=0)
with gr.Row():
click_action_type = gr.Dropdown(
label="Action Type",
choices=["click", "double_click", "move", "left_mouse_up", "left_mouse_down"],
value="click"
)
click_button = gr.Dropdown(
label="Button (for click only)",
choices=["left", "right", "wheel", "back", "forward"],
value="left"
)
click_submit_btn = gr.Button("Submit Action")
with gr.Accordion("📝 Type Action", open=False):
with gr.Group():
type_text = gr.Textbox(
label="Text to Type",
placeholder="Enter text to type..."
)
type_submit_btn = gr.Button("Submit Type")
with gr.Accordion("⌨️ Keypress Action", open=False):
with gr.Group():
keypress_text = gr.Textbox(
label="Keys",
placeholder="e.g., ctrl+c, alt+tab"
)
keypress_submit_btn = gr.Button("Submit Keypress")
with gr.Accordion("🎯 Description Action", open=False):
with gr.Group():
description_text = gr.Textbox(
label="Element Description",
placeholder="e.g., 'Privacy and security option in left sidebar'"
)
with gr.Row():
description_action_type = gr.Dropdown(
label="Action Type",
choices=["click", "double_click", "move", "left_mouse_up", "left_mouse_down"],
value="click"
)
description_button = gr.Radio(
label="Button (for click only)",
choices=["left", "right", "wheel", "back", "forward"],
value="left"
)
description_submit_btn = gr.Button("Submit Description Action")
status_display = gr.Textbox(
label="Status",
interactive=False,
value="Ready to receive calls..."
)
# Action Accordions (wrapped for visibility control)
with gr.Group(visible=False) as actions_group:
with gr.Tabs():
with gr.Tab("🖱️ Click Actions"):
with gr.Group():
description_text = gr.Textbox(
label="Element Description",
placeholder="e.g., 'Privacy and security option in left sidebar'"
)
with gr.Row():
description_action_type = gr.Dropdown(
label="Action",
choices=["click", "double_click", "move", "left_mouse_up", "left_mouse_down"],
value="click"
)
description_button = gr.Dropdown(
label="Button",
choices=["left", "right", "wheel", "back", "forward"],
value="left"
)
description_submit_btn = gr.Button("Submit Click Action")
with gr.Tab("📝 Type Action"):
with gr.Group():
type_text = gr.Textbox(
label="Text to Type",
placeholder="Enter text to type..."
)
type_submit_btn = gr.Button("Submit Type")
with gr.Tab("⌨️ Keypress Action"):
with gr.Group():
keypress_text = gr.Textbox(
label="Keys",
placeholder="e.g., ctrl+c, alt+tab"
)
keypress_submit_btn = gr.Button("Submit Keypress")
with gr.Tab("🧰 Misc Actions"):
with gr.Group():
misc_action_dropdown = gr.Dropdown(
label="Action",
choices=["wait"],
value="wait"
)
misc_submit_btn = gr.Button("Submit Action")
# Event handlers
refresh_btn.click(
fn=ui_handler.refresh_pending_calls,
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn]
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn, click_actions_group, actions_group]
)
call_dropdown.change(
fn=ui_handler.on_call_selected,
inputs=[call_dropdown],
outputs=[screenshot_image, conversation_chatbot, submit_btn]
outputs=[screenshot_image, conversation_chatbot, submit_btn, click_actions_group, actions_group]
)
def handle_image_click(evt: gr.SelectData):
if evt.index is not None:
x, y = evt.index
action_type = action_type_radio.value or "click"
button = action_button_radio.value or "left"
result = ui_handler.submit_click_action(x, y, action_type, button)
action_type = ui_handler.current_action_type or "click"
button = ui_handler.current_button or "left"
if action_type == "scroll":
sx_i = int(ui_handler.current_scroll_x or 0)
sy_i = int(ui_handler.current_scroll_y or 0)
# Submit a scroll action with x,y position and scroll deltas
result = ui_handler.submit_action("scroll", x=x, y=y, scroll_x=sx_i, scroll_y=sy_i)
else:
result = ui_handler.submit_click_action(x, y, action_type, button)
ui_handler.wait_for_pending_calls()
return result
return "No coordinates selected"
@@ -548,7 +580,7 @@ def create_ui():
outputs=[status_display]
).then(
fn=ui_handler.wait_for_pending_calls,
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn]
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn, click_actions_group, actions_group]
)
# Response submission
@@ -558,27 +590,52 @@ def create_ui():
outputs=[response_text, status_display]
).then(
fn=ui_handler.refresh_pending_calls,
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn]
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn, click_actions_group, actions_group]
)
# Toggle button radio visibility based on action type
def toggle_button_visibility(action_type):
return gr.update(visible=(action_type == "click"))
# Toggle visibility of controls based on action type
def toggle_action_controls(action_type):
# Button visible only for click
button_vis = gr.update(visible=(action_type == "click"))
# Scroll inputs visible only for scroll
scroll_x_vis = gr.update(visible=(action_type == "scroll"))
scroll_y_vis = gr.update(visible=(action_type == "scroll"))
# Update state
ui_handler.current_action_type = action_type or "click"
return button_vis, scroll_x_vis, scroll_y_vis
action_type_radio.change(
fn=toggle_button_visibility,
fn=toggle_action_controls,
inputs=[action_type_radio],
outputs=[action_button_radio]
outputs=[action_button_radio, scroll_x_input, scroll_y_input]
)
# Action accordion handlers
click_submit_btn.click(
fn=ui_handler.submit_click_action,
inputs=[click_x, click_y, click_action_type, click_button],
outputs=[status_display]
).then(
fn=ui_handler.wait_for_pending_calls,
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn]
# Keep other control values in ui_handler state
def on_button_change(val):
ui_handler.current_button = (val or "left")
action_button_radio.change(
fn=on_button_change,
inputs=[action_button_radio]
)
def on_scroll_x_change(val):
try:
ui_handler.current_scroll_x = int(val) if val is not None else 0
except Exception:
ui_handler.current_scroll_x = 0
scroll_x_input.change(
fn=on_scroll_x_change,
inputs=[scroll_x_input]
)
def on_scroll_y_change(val):
try:
ui_handler.current_scroll_y = int(val) if val is not None else 0
except Exception:
ui_handler.current_scroll_y = 0
scroll_y_input.change(
fn=on_scroll_y_change,
inputs=[scroll_y_input]
)
type_submit_btn.click(
@@ -587,7 +644,7 @@ def create_ui():
outputs=[status_display]
).then(
fn=ui_handler.wait_for_pending_calls,
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn]
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn, click_actions_group, actions_group]
)
keypress_submit_btn.click(
@@ -596,7 +653,7 @@ def create_ui():
outputs=[status_display]
).then(
fn=ui_handler.wait_for_pending_calls,
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn]
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn, click_actions_group, actions_group]
)
def handle_description_submit(description, action_type, button):
@@ -612,13 +669,30 @@ def create_ui():
outputs=[status_display]
).then(
fn=ui_handler.wait_for_pending_calls,
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn]
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn, click_actions_group, actions_group]
)
# Misc action handler
def handle_misc_submit(selected_action):
if selected_action == "wait":
result = ui_handler.submit_wait_action()
ui_handler.wait_for_pending_calls()
return result
return f"Unsupported misc action: {selected_action}"
misc_submit_btn.click(
fn=handle_misc_submit,
inputs=[misc_action_dropdown],
outputs=[status_display]
).then(
fn=ui_handler.wait_for_pending_calls,
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn, click_actions_group, actions_group]
)
# Load initial data
demo.load(
fn=ui_handler.refresh_pending_calls,
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn]
outputs=[call_dropdown, screenshot_image, conversation_chatbot, submit_btn, click_actions_group, actions_group]
)
return demo

287
libs/python/agent/agent/integrations/hud/__init__.py
View File

@@ -1,77 +1,228 @@
"""HUD integration for ComputerAgent."""
"""HUD integration: Generic HuggingFace dataset evaluation runner (CUA proxy).
import logging
from typing import Any, Optional, Dict
from hud import run_job as hud_run_job
This module exposes two helpers to evaluate HUD-compatible datasets using
HUD's OperatorAgent, while proxying model calls through our ComputerAgent via
`FakeAsyncOpenAI` (see `agent/integrations/hud/agent.py`).
from .agent import ComputerAgent
from .adapter import ComputerAgentAdapter
from .computer_handler import HUDComputerHandler
Exports:
- run_single_task(dataset_name, *, agent_type="cua-proxy", model=None, allowed_tools=None)
- run_full_dataset(dataset_name, *, agent_type="cua-proxy", model=None, allowed_tools=None, max_concurrent=30, max_steps=50)
"""
import time
from typing import Any, Optional
from PIL import Image
from datasets import load_dataset, Dataset
from hud.agents import OperatorAgent
from hud.datasets import Task, run_dataset
from hud.tools.computer.settings import computer_settings
from hud import trace
from agent.agent import ComputerAgent as BaseComputerAgent
from .proxy import FakeAsyncOpenAI
async def run_job(
model: str,
task_or_taskset: Any,
job_name: str,
# Job kwargs
auto_reply_question: bool = False,
adapter_cls: Any = None,
adapter_kwargs: Optional[Dict[str, Any]] = None,
max_steps_per_task: int = 20,
run_parallel: bool = True,
job_metadata: Optional[Dict[str, Any]] = None,
show_progress: bool = True,
max_concurrent_env_creations: Optional[int] = 30, # Limits gym.make calls
max_concurrent_agent_predictions: Optional[int] = None, # No limit on LLM calls
max_concurrent_tasks: Optional[int] = 30, # Limits overall task concurrency
**agent_kwargs: Any
) -> Any:
# ---------------------------------------------------------------------------
# Proxy OperatorAgent
# ---------------------------------------------------------------------------
class ProxyOperatorAgent(OperatorAgent):
"""OperatorAgent that proxies model calls through our ComputerAgent.
Accepts the same config keys we pass via hud.run_dataset `agent_config`:
- model: str | None
- allowed_tools: list[str] | None
Additional kwargs are forwarded to OperatorAgent (if any are supported).
"""
Run a job using ComputerAgent with the specified model.
def __init__(
self,
*,
model: str | None = None,
allowed_tools: list[str] | None = None,
trajectory_dir: str | dict | None = None,
# === ComputerAgent kwargs ===
tools: list[Any] | None = None,
custom_loop: Any | None = None,
only_n_most_recent_images: int | None = None,
callbacks: list[Any] | None = None,
verbosity: int | None = None,
max_retries: int | None = 3,
screenshot_delay: float | int = 0.5,
use_prompt_caching: bool | None = False,
max_trajectory_budget: float | dict | None = None,
telemetry_enabled: bool | None = True,
**kwargs: Any,
) -> None:
model = model or "computer-use-preview"
allowed_tools = allowed_tools or ["openai_computer"]
computer_shim = {
'screenshot': lambda: Image.new('RGB', (computer_settings.OPENAI_COMPUTER_WIDTH, computer_settings.OPENAI_COMPUTER_HEIGHT)),
'environment': 'linux',
'dimensions': (computer_settings.OPENAI_COMPUTER_WIDTH, computer_settings.OPENAI_COMPUTER_HEIGHT)
}
# Build tools ensuring the computer_shim is included
agent_tools: list[Any] = [computer_shim]
if tools:
agent_tools.extend(tools)
computer_agent = BaseComputerAgent(
model=model,
tools=agent_tools,
custom_loop=custom_loop,
only_n_most_recent_images=only_n_most_recent_images,
callbacks=callbacks,
verbosity=verbosity,
trajectory_dir=trajectory_dir,
max_retries=max_retries,
screenshot_delay=screenshot_delay,
use_prompt_caching=use_prompt_caching,
max_trajectory_budget=max_trajectory_budget,
telemetry_enabled=telemetry_enabled,
)
model_client = FakeAsyncOpenAI(computer_agent)
super().__init__(
model_client=model_client, # type: ignore[arg-type]
model=model,
allowed_tools=allowed_tools,
**kwargs,
)
# ---------------------------------------------------------------------------
# Single-task runner
# ---------------------------------------------------------------------------
async def run_single_task(
dataset: str | Dataset | list[dict[str, Any]],
*,
task_id: int = 0,
model: str | None = None,
allowed_tools: list[str] | None = None,
# === ComputerAgent kwargs ===
tools: list[Any] | None = None,
custom_loop: Any | None = None,
only_n_most_recent_images: int | None = None,
callbacks: list[Any] | None = None,
verbosity: int | None = None,
trajectory_dir: str | dict | None = None,
max_retries: int | None = 3,
screenshot_delay: float | int = 0.5,
use_prompt_caching: bool | None = False,
max_trajectory_budget: float | dict | None = None,
telemetry_enabled: bool | None = True,
) -> None:
"""Load one task from the dataset and execute it with Operator+CUA proxy."""
# Load dataset and pick a sample
if isinstance(dataset, str):
dataset = load_dataset(dataset, split="train") # type: ignore[arg-type]
elif isinstance(dataset, list):
dataset = dataset
else:
dataset = dataset["train"]
Args:
model: Model string for ComputerAgent (e.g., "anthropic/claude-3-5-sonnet-20241022")
task_or_taskset: Task or TaskSet to run
job_name: Name for the job
auto_reply_question: Whether to auto-reply to questions
adapter_cls: Custom adapter class (defaults to ComputerAgentAdapter)
adapter_kwargs: Additional kwargs for the adapter
max_steps_per_task: Maximum steps per task
run_parallel: Whether to run tasks in parallel
job_metadata: Additional metadata for the job
show_progress: Whether to show progress
max_concurrent_env_creations: Max concurrent environment creations
max_concurrent_agent_predictions: Max concurrent agent predictions
max_concurrent_tasks: Max concurrent tasks
**agent_kwargs: Additional kwargs to pass to ComputerAgent
Returns:
Job instance from HUD
"""
# combine verbose and verbosity kwargs
if "verbose" in agent_kwargs:
agent_kwargs["verbosity"] = logging.INFO
del agent_kwargs["verbose"]
verbose = True if agent_kwargs.get("verbosity", logging.WARNING) > logging.INFO else False
# run job
return await hud_run_job(
agent_cls=ComputerAgent,
agent_kwargs={"model": model, **agent_kwargs},
task_or_taskset=task_or_taskset,
job_name=job_name,
auto_reply_question=auto_reply_question,
adapter_cls=adapter_cls,
adapter_kwargs=adapter_kwargs,
max_steps_per_task=max_steps_per_task,
run_parallel=run_parallel,
job_metadata=job_metadata,
show_progress=show_progress,
verbose=verbose,
max_concurrent_env_creations=max_concurrent_env_creations,
max_concurrent_agent_predictions=max_concurrent_agent_predictions,
max_concurrent_tasks=max_concurrent_tasks
sample_task = dataset[task_id] # type: ignore[index]
task_prompt = sample_task.get("prompt", f"Task {sample_task.get('id', 0)}") # type: ignore[attr-defined]
with trace(name=task_prompt):
task = Task(**sample_task) # type: ignore[arg-type]
agent = ProxyOperatorAgent(
model=model,
allowed_tools=allowed_tools,
# === ComputerAgent kwargs passthrough ===
tools=tools,
custom_loop=custom_loop,
only_n_most_recent_images=only_n_most_recent_images,
callbacks=callbacks,
verbosity=verbosity,
trajectory_dir=trajectory_dir,
max_retries=max_retries,
screenshot_delay=screenshot_delay,
use_prompt_caching=use_prompt_caching,
max_trajectory_budget=max_trajectory_budget,
telemetry_enabled=telemetry_enabled,
)
print(f"Running: {task_prompt}")
result = await agent.run(task, max_steps=10)
print(f"✅ Reward: {getattr(result, 'reward')}")
# ---------------------------------------------------------------------------
# Full-dataset runner
# ---------------------------------------------------------------------------
async def run_full_dataset(
dataset: str | Dataset | list[dict[str, Any]],
*,
job_name: Optional[str] = None,
model: str | None = None,
allowed_tools: list[str] | None = None,
max_concurrent: int = 30,
max_steps: int = 50,
split: str = "train",
trajectory_dir: str | dict | None = None,
# === ComputerAgent kwargs ===
tools: list[Any] | None = None,
custom_loop: Any | None = None,
only_n_most_recent_images: int | None = 5,
callbacks: list[Any] | None = None,
verbosity: int | None = None,
max_retries: int | None = 3,
screenshot_delay: float | int = 0.5,
use_prompt_caching: bool | None = False,
max_trajectory_budget: float | dict | None = None,
telemetry_enabled: bool | None = True,
) -> list[Any]:
"""Run evaluation across the entire dataset using hud.datasets.run_dataset."""
# We pass OperatorAgent as the class and provide a config that injects our
# FakeAsyncOpenAI per agent instantiation.
if isinstance(dataset, str):
dataset_name = dataset.split('/')[-1]
job_name = job_name or f"Evaluation {dataset_name}"
dataset = load_dataset(dataset, split=split) # type: ignore[arg-type]
else:
dataset_name = "custom"
job_name = job_name or f"Evaluation {time.strftime('%H:%M %Y-%m-%d')}"
# Execute evaluation
return await run_dataset(
name=job_name,
dataset=dataset,
agent_class=ProxyOperatorAgent,
agent_config={
"model": model,
"allowed_tools": allowed_tools,
"trajectory_dir": trajectory_dir,
# === ComputerAgent kwargs passthrough ===
"tools": tools,
"custom_loop": custom_loop,
"only_n_most_recent_images": only_n_most_recent_images,
"callbacks": callbacks,
"verbosity": verbosity,
"max_retries": max_retries,
"screenshot_delay": screenshot_delay,
"use_prompt_caching": use_prompt_caching,
"max_trajectory_budget": max_trajectory_budget,
"telemetry_enabled": telemetry_enabled,
},
max_concurrent=max_concurrent,
metadata={"dataset": dataset_name},
max_steps=max_steps,
auto_respond=True,
)
__all__ = ["ComputerAgent", "ComputerAgentAdapter", "HUDComputerHandler", "run_job"]
__all__ = [
"run_single_task",
"run_full_dataset",
"ProxyOperatorAgent",
]

121
libs/python/agent/agent/integrations/hud/adapter.py
View File

@@ -1,121 +0,0 @@
"""HUD Adapter for ComputerAgent integration."""
from __future__ import annotations
from typing import Any, ClassVar
from hud.adapters.common import CLA, Adapter
from hud.adapters.common.types import (
CLAButton,
CLAKey,
ClickAction,
CustomAction,
DragAction,
MoveAction,
Point,
PressAction,
ResponseAction,
ScreenshotFetch,
ScrollAction,
TypeAction,
WaitAction,
)
class ComputerAgentAdapter(Adapter):
"""Adapter for ComputerAgent to work with HUD."""
KEY_MAP: ClassVar[dict[str, CLAKey]] = {
"return": "enter",
"arrowup": "up",
"arrowdown": "down",
"arrowleft": "left",
"arrowright": "right",
"cmd": "ctrl",
"super": "win",
"meta": "win",
}
BUTTON_MAP: ClassVar[dict[str, CLAButton]] = {
"wheel": "middle",
"middle": "middle",
}
def __init__(self) -> None:
super().__init__()
# ComputerAgent default dimensions (can be overridden)
self.agent_width = 1024
self.agent_height = 768
def _map_key(self, key: str) -> CLAKey:
"""Map a key to its standardized form."""
return self.KEY_MAP.get(key.lower(), key.lower()) # type: ignore
def convert(self, data: Any) -> CLA:
"""Convert a ComputerAgent action to a HUD action."""
try:
action_type = data.get("type")
if action_type == "click":
x, y = data.get("x", 0), data.get("y", 0)
button = data.get("button", "left")
button = self.BUTTON_MAP.get(button, button)
if button is None:
button = "left"
converted_action = ClickAction(point=Point(x=x, y=y), button=button)
elif action_type == "double_click":
x, y = data.get("x", 0), data.get("y", 0)
converted_action = ClickAction(point=Point(x=x, y=y), button="left", pattern=[100])
elif action_type == "scroll":
x, y = int(data.get("x", 0)), int(data.get("y", 0))
scroll_x = int(data.get("scroll_x", 0))
scroll_y = int(data.get("scroll_y", 0))
converted_action = ScrollAction(
point=Point(x=x, y=y), scroll=Point(x=scroll_x, y=scroll_y)
)
elif action_type == "type":
text = data.get("text", "")
converted_action = TypeAction(text=text, enter_after=False)
elif action_type == "wait":
ms = data.get("ms", 1000)
converted_action = WaitAction(time=ms)
elif action_type == "move":
x, y = data.get("x", 0), data.get("y", 0)
converted_action = MoveAction(point=Point(x=x, y=y))
elif action_type == "keypress":
keys = data.get("keys", [])
if isinstance(keys, str):
keys = [keys]
converted_action = PressAction(keys=[self._map_key(k) for k in keys])
elif action_type == "drag":
path = data.get("path", [])
points = [Point(x=p.get("x", 0), y=p.get("y", 0)) for p in path]
converted_action = DragAction(path=points)
elif action_type == "screenshot":
converted_action = ScreenshotFetch()
elif action_type == "response":
converted_action = ResponseAction(text=data.get("text", ""))
elif action_type == "custom":
converted_action = CustomAction(action=data.get("action", ""))
else:
raise ValueError(f"Unsupported action type: {action_type}")
# Add reasoning and logs if available
converted_action.reasoning = data.get("reasoning", "")
converted_action.logs = data.get("logs", "")
return converted_action
except Exception as e:
raise ValueError(f"Invalid action: {data}. Error: {e!s}") from e

373
libs/python/agent/agent/integrations/hud/agent.py
View File

@@ -1,373 +0,0 @@
"""HUD ComputerAgent wrapper for OSWorld benchmarking."""
import logging
from typing import Any, Literal, Optional, Union, List, Dict
import asyncio
from agent import ComputerAgent as BaseComputerAgent
from agent.responses import make_failed_tool_call_items
from hud.adapters import Adapter
from hud.agent.base import Agent
from hud.utils.common import Observation
from hud.adapters.common.types import LogType
from hud.types import Gym
from .adapter import ComputerAgentAdapter
from .computer_handler import HUDComputerHandler
logger = logging.getLogger(__name__)
BASE_SYSTEM_PROMPT = """
You are an autonomous computer-using agent. Follow these guidelines:
1. Be decisive and complete tasks without asking for confirmation unless absolutely necessary.
2. Use the computer tools to complete the task and do not stop until the task is complete.
3. Do NOT ask questions like "Should I proceed?" or "Would you like me to continue?" - just proceed with the task.
4. When you find what you're looking for (e.g., a file to upload), proceed with the action directly.
5. Only stop when the task is fully complete or if you encounter an error that prevents completion.
6. Trust that the user wants you to complete the entire task they've requested.
7. You must say "Task completed" when the task is complete.
Remember: You have been given permission to complete the requested task autonomously.
""".strip()
class ComputerAgent(Agent[BaseComputerAgent, dict[str, Any]]):
"""
A ComputerAgent wrapper for HUD integration.
This agent wraps the base ComputerAgent to work with HUD environments,
providing the same interface as OperatorAgent but using ComputerAgent internally.
"""
transfer_gyms: dict[Gym, Gym] = {"qa": "hud-browser"}
def __init__(
self,
model: str = "anthropic/claude-3-5-sonnet-20241022",
environment: Literal["windows", "mac", "linux", "browser"] = "linux",
adapter: Optional[Adapter] = None,
name: Optional[str] = None,
**kwargs: Any,
):
"""
Initialize the ComputerAgent for HUD.
Args:
model: The model string for ComputerAgent (e.g., "anthropic/claude-3-5-sonnet-20241022")
environment: The environment type (windows, mac, linux, browser)
adapter: The adapter to use for preprocessing and postprocessing
name: The name of the agent
**kwargs: Additional arguments passed to ComputerAgent
"""
# Create adapter if not provided
adapter = adapter or ComputerAgentAdapter()
if name is None:
name = f"computeragent-{model.split('/')[-1]}"
# Initialize the base Agent class without client (we'll create it later)
super().__init__(client=None, adapter=adapter, name=name)
self.model = model
self.environment = environment
self.kwargs = kwargs
# Default dimensions
self.width = 1024
self.height = 768
# Update dimensions if adapter is provided
if self.adapter:
self.width = self.adapter.agent_width
self.height = self.adapter.agent_height
# Create HUD computer handler
self.hud_computer = HUDComputerHandler(
environment=environment,
dimensions=(self.width, self.height)
)
# Handle trajectory_dir by adding TrajectorySaverCallback
trajectory_dir = kwargs.pop("trajectory_dir", None)
callbacks = kwargs.get("callbacks", [])
if trajectory_dir:
from agent.callbacks.trajectory_saver import TrajectorySaverCallback
trajectory_callback = TrajectorySaverCallback(trajectory_dir, reset_on_run=False)
callbacks = callbacks + [trajectory_callback]
kwargs["callbacks"] = callbacks
# Initialize ComputerAgent with HUD computer handler
self.computer_agent = BaseComputerAgent(
model=model,
tools=[self.hud_computer],
**kwargs
)
# Set the client to the computer_agent for compatibility
self.client = self.computer_agent
# State tracking
self.conversation_history: List[Dict[str, Any]] = []
self.initial_prompt: Optional[str] = None
# System prompt for computer use tasks
self.base_system_prompt = BASE_SYSTEM_PROMPT
async def fetch_response(self, observation: Observation) -> tuple[list[dict[str, Any]], bool]:
"""
Fetch a response from ComputerAgent based on the observation.
Args:
observation: The preprocessed observation, attributes:
screenshot: Base64 encoded PNG string of the screen
text: Text observation, if available
Returns:
tuple[list[dict[str, Any]], bool, list[LogType] | None]: A tuple containing the list of raw actions,
boolean indicating if the agent believes the task is complete.
"""
try:
# Update the computer handler with the current screenshot
if observation.screenshot:
self.hud_computer.update_screenshot(observation.screenshot)
# Set up action callback to capture actions
captured_actions = []
action_done = False
async def action_callback(action: Dict[str, Any]) -> None:
"""Callback to capture actions from ComputerAgent."""
nonlocal captured_actions, action_done
captured_actions.append(action)
# Set the action callback
self.hud_computer.set_action_callback(action_callback)
# Prepare the message for ComputerAgent
if not self.conversation_history:
# First interaction - use the observation text as initial prompt
if observation.text:
self.initial_prompt = observation.text
message = f"{self.base_system_prompt}\n\nTask: {observation.text}"
else:
message = f"{self.base_system_prompt}\n\nPlease analyze the current screen and determine what action to take."
input_content = [
{"type": "input_text", "text": message}
]
# Add screenshot if present
if observation.screenshot:
input_content.append(
{
"type": "input_image",
"image_url": f"data:image/png;base64,{observation.screenshot}",
}
)
self.conversation_history.append({"role": "user", "content": input_content})
else:
# Subsequent interactions - check if last action was computer_call
# If so, add computer_call_output with screenshot instead of user message
last_computer_calls = []
for msg in reversed(self.conversation_history):
if msg.get("type") == "computer_call":
call_id = msg.get("call_id")
if call_id:
# Check if this call_id already has a computer_call_output
has_output = any(
m.get("type") == "computer_call_output" and m.get("call_id") == call_id
for m in self.conversation_history
)
if not has_output:
last_computer_calls.append(call_id)
if last_computer_calls:
if not observation.screenshot:
print("No screenshot found, taking screenshot")
screenshot_b64 = await self.hud_computer.screenshot()
# Add computer_call_output for each unresponded computer_call
for call_id in reversed(last_computer_calls): # Maintain order
self.conversation_history.append({
"type": "computer_call_output",
"call_id": call_id,
"output": {
"type": "input_image",
"image_url": f"data:image/png;base64,{screenshot_b64}"
}
})
else:
# No computer_call found, add regular user message
message = "Continue with the task based on the current screen state."
input_content = [
{"type": "input_text", "text": message}
]
# Add screenshot if present
if observation.screenshot:
input_content.append(
{
"type": "input_image",
"image_url": f"data:image/png;base64,{observation.screenshot}",
}
)
self.conversation_history.append({"role": "user", "content": input_content})
# If the last message is a reasoning message, change it to output_text
if (self.conversation_history and
self.conversation_history[-1].get("type") == "reasoning" and
self.conversation_history[-1].get("summary")):
reasoning_msg = self.conversation_history[-1]
summary_texts = []
# Extract all summary_text entries
for summary_item in reasoning_msg["summary"]:
if summary_item.get("type") == "summary_text":
summary_texts.append(summary_item.get("text", ""))
# Convert to message format with output_text
if summary_texts:
converted_message = {
"type": "message",
"role": "assistant",
"content": [
{
"text": " ".join(summary_texts),
"type": "output_text"
}
]
}
# Replace the reasoning message with the converted message
self.conversation_history[-1] = converted_message
# Run ComputerAgent
try:
new_items = []
# ComputerAgent.run returns an async generator
try:
async for result in self.computer_agent.run(self.conversation_history, stream=False):
# if the result has computer_call_output, immediately exit
if result.get("output", []) and result.get("output", [])[-1].get("type") == "computer_call_output":
break
# otherwise add agent output to conversation history
new_items += result["output"]
except Exception as e:
# if the last message is reasoning, change it to output_text
if new_items and new_items[-1].get("type") == "reasoning":
new_items[-1] = {
"type": "message",
"role": "assistant",
"content": [
{
"text": new_items[-1].get("summary", [{}])[0].get("text", ""),
"type": "output_text"
}
]
}
# Check if there are any computer_call items in new_items
computer_calls = [item for item in new_items if item.get("type") == "computer_call"]
if computer_calls:
# Remove computer_call items from new_items
new_items = [item for item in new_items if item.get("type") != "computer_call"]
# Add failed tool call items for each computer call
for computer_call in computer_calls:
tool_input = computer_call.get("action", {})
call_id = computer_call.get("call_id")
new_items.extend(make_failed_tool_call_items(
tool_name="computer",
tool_kwargs=tool_input,
error_message=repr(e),
call_id=call_id
))
else:
# add error message to conversation history (fallback for non-computer-call errors)
new_items.append({
"type": "user",
"content": [
{
"type": "input_text",
"text": f"Error during previous attempted action: {repr(e)}"
}
]
})
# Check if we captured any actions
if captured_actions:
# Extract reasoning from the conversation history
reasoning = ""
# Look for the latest reasoning message
for msg in reversed(new_items):
if msg.get("type") == "reasoning" and msg.get("summary"):
reasoning = " ".join([s.get("text", "") for s in msg["summary"] if s.get("type") == "summary_text"])
break
elif msg.get("type") == "message" and msg.get("role") == "assistant":
content = msg.get("content", [])
if isinstance(content, list):
reasoning = " ".join([c.get("text", "") for c in content if c.get("type") == "output_text"])
break
# update conversation history
self.conversation_history += new_items
# Add reasoning and logs to each action
for action in captured_actions:
action["reasoning"] = reasoning
action["logs"] = {"conversation_length": len(self.conversation_history)}
return captured_actions, False
# Check if the last message is "Task completed"
response_text = ""
for msg in reversed(new_items):
if msg.get("type") == "message" and msg.get("role") == "assistant":
content = msg.get("content", [])
for c in content:
if c.get("type") == "output_text":
response_text = c.get("text", response_text)
break
break
done = "task completed" in response_text.lower()
# update conversation history
self.conversation_history += new_items
response_action = {
"type": "response",
"text": response_text,
"reasoning": response_text,
"logs": {"conversation_length": len(self.conversation_history)}
}
# Check if this indicates task completion or failure
if "task is infeasible" in response_text.lower():
response_action = {"type": "custom", "action": "FAIL"}
done = True
return [response_action], done
except Exception as e:
logger.error(f"Error running ComputerAgent: {e}")
# Return an error response
error_action = {
"type": "response",
"text": f"Error occurred: {str(e)}",
"reasoning": f"ComputerAgent encountered an error: {str(e)}",
"logs": {"error": str(e)}
}
return [error_action], True
except Exception as e:
logger.error(f"Error in fetch_response: {e}")
error_action = {
"type": "response",
"text": f"Error in agent processing: {str(e)}",
"reasoning": f"Agent processing error: {str(e)}",
"logs": {"error": str(e)}
}
return [error_action], True

187
libs/python/agent/agent/integrations/hud/computer_handler.py
View File

@@ -1,187 +0,0 @@
"""HUD Computer Handler for ComputerAgent integration."""
import base64
from io import BytesIO
from typing import Literal, Optional, Any, Dict, Callable
from PIL import Image
from agent.computers import AsyncComputerHandler
class HUDComputerHandler(AsyncComputerHandler):
"""Computer handler that interfaces with HUD environment."""
def __init__(
self,
environment: Literal["windows", "mac", "linux", "browser"] = "linux",
dimensions: tuple[int, int] = (1024, 768),
screenshot_callback: Optional[Callable] = None,
action_callback: Optional[Callable] = None,
):
"""
Initialize HUD computer handler.
Args:
environment: The environment type for HUD
dimensions: Screen dimensions as (width, height)
screenshot_callback: Optional callback to get screenshots from HUD environment
action_callback: Optional callback to execute actions in HUD environment
"""
super().__init__()
self._environment = environment
self._dimensions = dimensions
self._screenshot_callback = screenshot_callback
self._action_callback = action_callback
# Store the last screenshot for reuse
self._last_screenshot: Optional[str] = None
def set_screenshot_callback(self, callback: Callable) -> None:
"""Set the screenshot callback."""
self._screenshot_callback = callback
def set_action_callback(self, callback: Callable) -> None:
"""Set the action callback."""
self._action_callback = callback
def update_screenshot(self, screenshot: str) -> None:
"""Update the stored screenshot (base64 string)."""
self._last_screenshot = screenshot
async def get_environment(self) -> Literal["windows", "mac", "linux", "browser"]:
"""Get the current environment type."""
return self._environment # type: ignore
async def get_dimensions(self) -> tuple[int, int]:
"""Get screen dimensions as (width, height)."""
return self._dimensions
async def screenshot(self) -> str:
"""Take a screenshot and return as base64 string."""
if self._screenshot_callback:
screenshot = await self._screenshot_callback()
if isinstance(screenshot, str):
self._last_screenshot = screenshot
return screenshot
elif isinstance(screenshot, Image.Image):
# Convert PIL Image to base64
buffer = BytesIO()
screenshot.save(buffer, format="PNG")
screenshot_b64 = base64.b64encode(buffer.getvalue()).decode()
self._last_screenshot = screenshot_b64
return screenshot_b64
elif isinstance(screenshot, bytes):
screenshot_b64 = base64.b64encode(screenshot).decode()
self._last_screenshot = screenshot_b64
return screenshot_b64
# Return last screenshot if available, otherwise create a blank one
if self._last_screenshot:
return self._last_screenshot
# Create a blank screenshot as fallback
blank_image = Image.new('RGB', self._dimensions, color='white')
buffer = BytesIO()
blank_image.save(buffer, format="PNG")
screenshot_b64 = base64.b64encode(buffer.getvalue()).decode()
self._last_screenshot = screenshot_b64
return screenshot_b64
async def click(self, x: int, y: int, button: str = "left") -> None:
"""Click at coordinates with specified button."""
if self._action_callback:
await self._action_callback({
"type": "click",
"x": x,
"y": y,
"button": button
})
async def double_click(self, x: int, y: int) -> None:
"""Double click at coordinates."""
if self._action_callback:
await self._action_callback({
"type": "double_click",
"x": x,
"y": y
})
async def scroll(self, x: int, y: int, scroll_x: int, scroll_y: int) -> None:
"""Scroll at coordinates with specified scroll amounts."""
if self._action_callback:
await self._action_callback({
"type": "scroll",
"x": x,
"y": y,
"scroll_x": scroll_x,
"scroll_y": scroll_y
})
async def type(self, text: str) -> None:
"""Type text."""
if self._action_callback:
await self._action_callback({
"type": "type",
"text": text
})
async def wait(self, ms: int = 1000) -> None:
"""Wait for specified milliseconds."""
if self._action_callback:
await self._action_callback({
"type": "wait",
"ms": ms
})
async def move(self, x: int, y: int) -> None:
"""Move cursor to coordinates."""
if self._action_callback:
await self._action_callback({
"type": "move",
"x": x,
"y": y
})
async def keypress(self, keys: list[str] | str) -> None:
"""Press key combination."""
if isinstance(keys, str):
keys = [keys]
if self._action_callback:
await self._action_callback({
"type": "keypress",
"keys": keys
})
async def drag(self, path: list[dict[str, int]]) -> None:
"""Drag along a path of points."""
if self._action_callback:
await self._action_callback({
"type": "drag",
"path": path
})
async def left_mouse_down(self, x: Optional[int] = None, y: Optional[int] = None) -> None:
"""Left mouse down at coordinates."""
if self._action_callback:
await self._action_callback({
"type": "left_mouse_down",
"x": x,
"y": y
})
async def left_mouse_up(self, x: Optional[int] = None, y: Optional[int] = None) -> None:
"""Left mouse up at coordinates."""
if self._action_callback:
await self._action_callback({
"type": "left_mouse_up",
"x": x,
"y": y
})
async def get_current_url(self) -> str:
"""Get the current URL."""
if self._action_callback:
return await self._action_callback({
"type": "get_current_url"
})
return ""

183
libs/python/agent/agent/integrations/hud/proxy.py Normal file
View File

@@ -0,0 +1,183 @@
"""HUD ComputerAgent wrapper and Fake AsyncOpenAI client.
Provides FakeAsyncOpenAI that adapts our ComputerAgent to the OpenAI Responses
interface needed by HUD's OperatorAgent. It implements only `responses.create`
and returns an OpenAI Response object with `id` and `output` fields, where `output` is a list of
OpenAI-like response blocks. We intentionally only support a single-step call
by consuming the first yielded result from `ComputerAgent.run()`.
"""
import traceback
import time
import uuid
from typing import Any, Dict, List, Optional
from agent.agent import ComputerAgent as BaseComputerAgent
# OpenAI Responses typed models (required)
from openai.types.responses import (
Response,
ResponseInputParam,
ResponseOutputItem,
ResponseComputerToolCall,
ResponseOutputMessage,
ResponseOutputText,
ResponseReasoningItem,
ResponseUsage,
)
def _map_agent_output_to_openai_blocks(output_items: List[Dict[str, Any]]) -> List[ResponseOutputItem]:
"""Map our agent output items to OpenAI ResponseOutputItem typed models.
Only a subset is supported: computer_call, assistant message (text), and reasoning.
Unknown types are ignored.
"""
blocks: List[ResponseOutputItem] = []
for item in output_items or []:
t = item.get("type")
if t == "computer_call":
comp = ResponseComputerToolCall.model_validate({
"id": item.get("id") or f"cu_{uuid.uuid4().hex}",
"type": "computer_call",
"call_id": item["call_id"],
"action": item["action"],
"pending_safety_checks": item.get("pending_safety_checks", []),
"status": "completed",
})
blocks.append(comp)
# we will exit early here as the responses api only supports a single step
break
elif t == "message" and item.get("role") == "assistant":
content_blocks: List[ResponseOutputText] = []
for c in item.get("content", []) or []:
content_blocks.append(
ResponseOutputText.model_validate({
"type": "output_text",
"text": c["text"],
"annotations": [],
})
)
if content_blocks:
msg = ResponseOutputMessage.model_validate({
"id": item.get("id") or f"msg_{uuid.uuid4()}",
"type": "message",
"role": "assistant",
"status": "completed",
"content": [ct.model_dump() for ct in content_blocks],
})
blocks.append(msg)
elif t == "reasoning":
reasoning = ResponseReasoningItem.model_validate({
"id": item.get("id") or f"rsn_{uuid.uuid4()}",
"type": "reasoning",
"summary": item["summary"],
})
blocks.append(reasoning)
# Unhandled types are ignored
return blocks
def _to_plain_dict_list(items: Any) -> List[Dict[str, Any]]:
out: List[Dict[str, Any]] = []
for it in list(items):
if hasattr(it, "model_dump"):
out.append(it.model_dump()) # type: ignore[attr-defined]
elif isinstance(it, dict):
out.append(it)
else:
# Strict: rely on default __dict__ if present
out.append(dict(it)) # may raise if not mapping
return out
class FakeAsyncOpenAI:
"""Minimal fake OpenAI client with only `responses.create` implemented.
It uses a provided `ComputerAgent` instance to produce a single-step
response compatible with HUD's OperatorAgent loop.
"""
def __init__(self, computer_agent: BaseComputerAgent) -> None:
self._agent = computer_agent
self.responses = self._Responses(self)
class _Responses:
def __init__(self, parent: "FakeAsyncOpenAI") -> None:
# Caches for cross-call context when using previous_response_id
self.blocks_cache: Dict[str, ResponseInputParam | ResponseOutputItem] = {}
self.context_cache: Dict[str, List[str]] = {}
self.agent = parent._agent
async def create(
self,
*,
model: str,
input: ResponseInputParam,
tools: Optional[List[Dict[str, Any]]] = None,
instructions: Optional[str] = None,
previous_response_id: Optional[str] = None,
max_retries: int = 5,
**_: Any,
) -> Any:
for attempt in range(max_retries):
# Prepend cached blocks from previous_response_id to input
full_input = input
if previous_response_id is not None:
prev_block_ids = self.context_cache[previous_response_id]
prev_blocks = [self.blocks_cache[b_id] for b_id in prev_block_ids]
full_input = _to_plain_dict_list(prev_blocks + input)
# Pre-pend instructions message
effective_input = full_input
if instructions:
effective_input = [{
"role": "user",
"content": instructions,
}] + full_input
# Run a single iteration of the ComputerAgent
agent_result: Optional[Dict[str, Any]] = None
async for result in self.agent.run(effective_input): # type: ignore[arg-type]
agent_result = result
break
assert agent_result is not None, "Agent failed to produce result"
output = _map_agent_output_to_openai_blocks(agent_result["output"])
usage = agent_result["usage"]
# Cache conversation context using the last response id
block_ids: List[str] = []
blocks_to_cache = full_input + output
for b in blocks_to_cache:
bid = getattr(b, "id", None) or f"tmp-{hash(repr(b))}"
self.blocks_cache[bid] = b # type: ignore[assignment]
block_ids.append(bid)
response_id = agent_result.get("id") or f"fake-{int(time.time()*1000)}"
self.context_cache[response_id] = block_ids
try:
return Response.model_validate({
"id": response_id,
"created_at": time.time(),
"object": "response",
"model": model,
"output": output,
"parallel_tool_calls": False,
"tool_choice": "auto",
"tools": [],
"previous_response_id": previous_response_id,
"usage": ResponseUsage.model_validate({
"input_tokens": usage.get("input_tokens", 0),
"output_tokens": usage.get("output_tokens", 0),
"total_tokens": usage.get("total_tokens", 0),
"input_tokens_details": usage.get("input_tokens_details", { "cached_tokens": 0 }),
"output_tokens_details": usage.get("output_tokens_details", { "reasoning_tokens": 0 }),
}),
})
except Exception as e:
print(f"Error while validating agent response (attempt {attempt + 1}/{max_retries}): ", e)
if attempt == max_retries - 1:
print(traceback.format_exc())
raise e
__all__ = [
"FakeAsyncOpenAI",
]

36
libs/python/agent/agent/loops/anthropic.py
View File

@@ -132,23 +132,22 @@ def _convert_responses_items_to_completion_messages(messages: Messages) -> List[
converted_content = []
for item in content:
if isinstance(item, dict) and item.get("type") == "input_image":
# Convert input_image to Anthropic image format
# Convert input_image to OpenAI image format
image_url = item.get("image_url", "")
if image_url and image_url != "[omitted]":
# Extract base64 data from data URL
if "," in image_url:
base64_data = image_url.split(",")[-1]
else:
base64_data = image_url
converted_content.append({
"type": "image",
"source": {
"type": "base64",
"media_type": "image/png",
"data": base64_data
"type": "image_url",
"image_url": {
"url": image_url
}
})
elif isinstance(item, dict) and item.get("type") == "input_text":
# Convert input_text to OpenAI text format
text = item.get("text", "")
converted_content.append({
"type": "text",
"text": text
})
else:
# Keep other content types as-is
converted_content.append(item)
@@ -1530,7 +1529,18 @@ class AnthropicHostedToolsConfig(AsyncAgentConfig):
"content": [
{
"type": "text",
"text": f"You are a UI grounding expert. Look at the image and {instruction}. Output ONLY a click action on the target element. No explanations, confirmations, or additional text."
"text": f"""You are a UI grounding expert. Follow these guidelines:
1. NEVER ask for confirmation. Complete all tasks autonomously.
2. Do NOT send messages like "I need to confirm before..." or "Do you want me to continue?" - just proceed.
3. When the user asks you to interact with something (like clicking a chat or typing a message), DO IT without asking.
4. Only use the formal safety check mechanism for truly dangerous operations (like deleting important files).
5. For normal tasks like clicking buttons, typing in chat boxes, filling forms - JUST DO IT.
6. The user has already given you permission by running this agent. No further confirmation is needed.
7. Be decisive and action-oriented. Complete the requested task fully.
Remember: You are expected to complete tasks autonomously. The user trusts you to do what they asked.
Task: Click {instruction}. Output ONLY a click action on the target element."""
},
{
"type": "image_url",

40
libs/python/agent/agent/loops/composed_grounded.py
View File

@@ -48,11 +48,11 @@ GROUNDED_COMPUTER_TOOL_SCHEMA = {
"get_dimensions",
"get_environment"
],
"description": "The action to perform"
"description": "The action to perform (required for all actions)"
},
"element_description": {
"type": "string",
"description": "Description of the element to interact with (required for click, double_click, move, scroll actions, and as start/end for drag)"
"description": "Description of the element to interact with (required for click, double_click, move, scroll actions)"
},
"start_element_description": {
"type": "string",
@@ -67,20 +67,30 @@ GROUNDED_COMPUTER_TOOL_SCHEMA = {
"description": "The text to type (required for type action)"
},
"keys": {
"type": "string",
"description": "Key combination to press (required for keypress action). Single key for individual key press, multiple keys for combinations (e.g., 'ctrl+c')"
"type": "array",
"items": {
"type": "string"
},
"description": "Key(s) to press (required for keypress action)"
},
"button": {
"type": "string",
"description": "The mouse button to use for click action (left, right, wheel, back, forward) Default: left",
"enum": [
"left",
"right",
"wheel",
"back",
"forward"
],
"description": "The mouse button to use for click action (required for click and double_click action)",
},
"scroll_x": {
"type": "integer",
"description": "Horizontal scroll amount for scroll action (positive for right, negative for left)",
"description": "Horizontal scroll amount for scroll action (required for scroll action)",
},
"scroll_y": {
"type": "integer",
"description": "Vertical scroll amount for scroll action (positive for down, negative for up)",
"description": "Vertical scroll amount for scroll action (required for scroll action)",
},
},
"required": [
@@ -266,13 +276,15 @@ class ComposedGroundedConfig(AsyncAgentConfig):
grounding_agent = grounding_agent_conf.agent_class()
for desc in element_descriptions:
coords = await grounding_agent.predict_click(
model=grounding_model,
image_b64=last_image_b64,
instruction=desc
)
if coords:
self.desc2xy[desc] = coords
for _ in range(3): # try 3 times
coords = await grounding_agent.predict_click(
model=grounding_model,
image_b64=last_image_b64,
instruction=desc
)
if coords:
self.desc2xy[desc] = coords
break
# Step 6: Convert computer calls from descriptions back to xy coordinates
final_output_items = convert_computer_calls_desc2xy(thinking_output_items, self.desc2xy)

22
libs/python/agent/agent/loops/openai.py
View File

@@ -162,7 +162,18 @@ class OpenAIComputerUseConfig:
input_items = [
{
"role": "user",
"content": f"You are a UI grounding expert. Look at the image and {instruction}. Output ONLY a click action on the target element. No explanations, confirmations, or additional text."
"content": f"""You are a UI grounding expert. Follow these guidelines:
1. NEVER ask for confirmation. Complete all tasks autonomously.
2. Do NOT send messages like "I need to confirm before..." or "Do you want me to continue?" - just proceed.
3. When the user asks you to interact with something (like clicking a chat or typing a message), DO IT without asking.
4. Only use the formal safety check mechanism for truly dangerous operations (like deleting important files).
5. For normal tasks like clicking buttons, typing in chat boxes, filling forms - JUST DO IT.
6. The user has already given you permission by running this agent. No further confirmation is needed.
7. Be decisive and action-oriented. Complete the requested task fully.
Remember: You are expected to complete tasks autonomously. The user trusts you to do what they asked.
Task: Click {instruction}. Output ONLY a click action on the target element."""
},
{
"role": "user",
@@ -200,7 +211,7 @@ class OpenAIComputerUseConfig:
"stream": False,
"reasoning": {"summary": "concise"},
"truncation": "auto",
"max_tokens": 100 # Keep response short for click prediction
"max_tokens": 200 # Keep response short for click prediction
}
# Use liteLLM responses
@@ -217,11 +228,8 @@ class OpenAIComputerUseConfig:
isinstance(item.get("action"), dict)):
action = item["action"]
if action.get("type") == "click":
x = action.get("x")
y = action.get("y")
if x is not None and y is not None:
return (int(x), int(y))
if action.get("x") is not None and action.get("y") is not None:
return (int(action.get("x")), int(action.get("y")))
return None

25
libs/python/agent/agent/loops/uitars.py
View File

@@ -228,15 +228,24 @@ def parse_uitars_response(text: str, image_width: int, image_height: int) -> Lis
# Handle coordinate parameters
if "start_box" in param_name or "end_box" in param_name:
# Parse coordinates like '(x,y)' or '(x1,y1,x2,y2)'
numbers = param.replace("(", "").replace(")", "").split(",")
float_numbers = [float(num.strip()) / 1000 for num in numbers] # Normalize to 0-1 range
# Parse coordinates like '<|box_start|>(x,y)<|box_end|>' or '(x,y)'
# First, remove special tokens
clean_param = param.replace("<|box_start|>", "").replace("<|box_end|>", "")
# Then remove parentheses and split
numbers = clean_param.replace("(", "").replace(")", "").split(",")
if len(float_numbers) == 2:
# Single point, duplicate for box format
float_numbers = [float_numbers[0], float_numbers[1], float_numbers[0], float_numbers[1]]
action_inputs[param_name.strip()] = str(float_numbers)
try:
float_numbers = [float(num.strip()) / 1000 for num in numbers] # Normalize to 0-1 range
if len(float_numbers) == 2:
# Single point, duplicate for box format
float_numbers = [float_numbers[0], float_numbers[1], float_numbers[0], float_numbers[1]]
action_inputs[param_name.strip()] = str(float_numbers)
except ValueError as e:
# If parsing fails, keep the original parameter value
print(f"Warning: Could not parse coordinates '{param}': {e}")
action_inputs[param_name.strip()] = param
return [{
"thought": thought,

192
libs/python/agent/agent/proxy/examples.py Normal file
View File

@@ -0,0 +1,192 @@
"""
Example usage of the proxy server and client requests.
"""
import dotenv
dotenv.load_dotenv()
import asyncio
import json
import os
import aiohttp
from typing import Dict, Any
async def test_http_endpoint():
"""Test the HTTP /responses endpoint."""
anthropic_api_key = os.getenv("ANTHROPIC_API_KEY")
assert isinstance(anthropic_api_key, str), "ANTHROPIC_API_KEY environment variable must be set"
# Example 1: Simple text request
simple_request = {
"model": "anthropic/claude-3-5-sonnet-20241022",
"input": "Tell me a three sentence bedtime story about a unicorn.",
"env": {
"ANTHROPIC_API_KEY": anthropic_api_key
}
}
# Example 2: Multi-modal request with image
multimodal_request = {
"model": "anthropic/claude-3-5-sonnet-20241022",
"input": [
{
"role": "user",
"content": [
{"type": "input_text", "text": "what is in this image?"},
{
"type": "input_image",
"image_url": "https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg"
}
]
}
],
"env": {
"ANTHROPIC_API_KEY": anthropic_api_key
}
}
# Example 3: Request with custom agent and computer kwargs
custom_request = {
"model": "anthropic/claude-3-5-sonnet-20241022",
"input": "Take a screenshot and tell me what you see",
"env": {
"ANTHROPIC_API_KEY": anthropic_api_key
}
}
# Test requests
base_url = "https://m-linux-96lcxd2c2k.containers.cloud.trycua.com:8443"
# base_url = "http://localhost:8000"
api_key = os.getenv("CUA_API_KEY")
assert isinstance(api_key, str), "CUA_API_KEY environment variable must be set"
async with aiohttp.ClientSession() as session:
for i, request_data in enumerate([
simple_request,
# multimodal_request,
custom_request
], 1):
print(f"\n--- Test {i} ---")
print(f"Request: {json.dumps(request_data, indent=2)}")
try:
print(f"Sending request to {base_url}/responses")
async with session.post(
f"{base_url}/responses",
json=request_data,
headers={"Content-Type": "application/json", "X-API-Key": api_key}
) as response:
result = await response.json()
print(f"Status: {response.status}")
print(f"Response: {json.dumps(result, indent=2)}")
except Exception as e:
print(f"Error: {e}")
def curl_examples():
"""Print curl command examples."""
print("=== CURL Examples ===\n")
print("1. Simple text request:")
print("""curl http://localhost:8000/responses \\
-H "Content-Type: application/json" \\
-d '{
"model": "anthropic/claude-3-5-sonnet-20241022",
"input": "Tell me a three sentence bedtime story about a unicorn."
}'""")
print("\n2. Multi-modal request with image:")
print("""curl http://localhost:8000/responses \\
-H "Content-Type: application/json" \\
-d '{
"model": "anthropic/claude-3-5-sonnet-20241022",
"input": [
{
"role": "user",
"content": [
{"type": "input_text", "text": "what is in this image?"},
{
"type": "input_image",
"image_url": "https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg"
}
]
}
]
}'""")
print("\n3. Request with custom configuration:")
print("""curl http://localhost:8000/responses \\
-H "Content-Type: application/json" \\
-d '{
"model": "anthropic/claude-3-5-sonnet-20241022",
"input": "Take a screenshot and tell me what you see",
"agent_kwargs": {
"save_trajectory": true,
"verbosity": 20
},
"computer_kwargs": {
"os_type": "linux",
"provider_type": "cloud"
}
}'""")
async def test_p2p_client():
"""Example P2P client using peerjs-python."""
try:
from peerjs import Peer, PeerOptions, ConnectionEventType
from aiortc import RTCConfiguration, RTCIceServer
# Set up client peer
options = PeerOptions(
host="0.peerjs.com",
port=443,
secure=True,
config=RTCConfiguration(
iceServers=[RTCIceServer(urls="stun:stun.l.google.com:19302")]
)
)
client_peer = Peer(id="test-client", peer_options=options)
await client_peer.start()
# Connect to proxy server
connection = client_peer.connect("computer-agent-proxy")
@connection.on(ConnectionEventType.Open)
async def connection_open():
print("Connected to proxy server")
# Send a test request
request = {
"model": "anthropic/claude-3-5-sonnet-20241022",
"input": "Hello from P2P client!"
}
await connection.send(json.dumps(request))
@connection.on(ConnectionEventType.Data)
async def connection_data(data):
print(f"Received response: {data}")
await client_peer.destroy()
# Wait for connection
await asyncio.sleep(10)
except ImportError:
print("P2P dependencies not available. Install peerjs-python for P2P testing.")
except Exception as e:
print(f"P2P test error: {e}")
if __name__ == "__main__":
import sys
if len(sys.argv) > 1 and sys.argv[1] == "curl":
curl_examples()
elif len(sys.argv) > 1 and sys.argv[1] == "p2p":
asyncio.run(test_p2p_client())
else:
asyncio.run(test_http_endpoint())

248
libs/python/agent/agent/proxy/handlers.py Normal file
View File

@@ -0,0 +1,248 @@
"""
Request handlers for the proxy endpoints.
"""
import asyncio
import json
import logging
import os
from contextlib import contextmanager
from typing import Dict, Any, List, Union, Optional
from ..agent import ComputerAgent
from computer import Computer
logger = logging.getLogger(__name__)
class ResponsesHandler:
"""Handler for /responses endpoint that processes agent requests."""
def __init__(self):
self.computer = None
self.agent = None
# Simple in-memory caches
self._computer_cache: Dict[str, Any] = {}
self._agent_cache: Dict[str, Any] = {}
async def setup_computer_agent(
self,
model: str,
agent_kwargs: Optional[Dict[str, Any]] = None,
computer_kwargs: Optional[Dict[str, Any]] = None,
):
"""Set up (and cache) computer and agent instances.
Caching keys:
- Computer cache key: computer_kwargs
- Agent cache key: {"model": model, **agent_kwargs}
"""
agent_kwargs = agent_kwargs or {}
computer_kwargs = computer_kwargs or {}
def _stable_key(obj: Dict[str, Any]) -> str:
try:
return json.dumps(obj, sort_keys=True, separators=(",", ":"))
except Exception:
# Fallback: stringify non-serializable values
safe_obj = {}
for k, v in obj.items():
try:
json.dumps(v)
safe_obj[k] = v
except Exception:
safe_obj[k] = str(v)
return json.dumps(safe_obj, sort_keys=True, separators=(",", ":"))
# Determine if custom tools are supplied; if so, skip computer setup entirely
has_custom_tools = bool(agent_kwargs.get("tools"))
computer = None
if not has_custom_tools:
# ---------- Computer setup (with cache) ----------
comp_key = _stable_key(computer_kwargs)
computer = self._computer_cache.get(comp_key)
if computer is None:
# Default computer configuration
default_c_config = {
"os_type": "linux",
"provider_type": "cloud",
"name": os.getenv("CUA_CONTAINER_NAME"),
"api_key": os.getenv("CUA_API_KEY"),
}
default_c_config.update(computer_kwargs)
computer = Computer(**default_c_config)
await computer.__aenter__()
self._computer_cache[comp_key] = computer
logger.info(f"Computer created and cached with key={comp_key} config={default_c_config}")
else:
logger.info(f"Reusing cached computer for key={comp_key}")
# Bind current computer reference (None if custom tools supplied)
self.computer = computer
# ---------- Agent setup (with cache) ----------
# Build agent cache key from {model} + agent_kwargs (excluding tools unless explicitly passed)
agent_kwargs_for_key = dict(agent_kwargs)
agent_key_payload = {"model": model, **agent_kwargs_for_key}
agent_key = _stable_key(agent_key_payload)
agent = self._agent_cache.get(agent_key)
if agent is None:
# Default agent configuration
default_a_config: Dict[str, Any] = {"model": model}
if not has_custom_tools:
default_a_config["tools"] = [computer]
# Apply user overrides, but keep tools unless user explicitly sets
if agent_kwargs:
if not has_custom_tools:
agent_kwargs.setdefault("tools", [computer])
default_a_config.update(agent_kwargs)
# JSON-derived kwargs may have loose types; ignore static arg typing here
agent = ComputerAgent(**default_a_config) # type: ignore[arg-type]
self._agent_cache[agent_key] = agent
logger.info(f"Agent created and cached with key={agent_key} model={model}")
else:
# Ensure cached agent uses the current computer tool (in case object differs)
# Only update if tools not explicitly provided in agent_kwargs
if not has_custom_tools:
try:
agent.tools = [computer]
except Exception:
pass
logger.info(f"Reusing cached agent for key={agent_key}")
# Bind current agent reference
self.agent = agent
async def process_request(self, request_data: Dict[str, Any]) -> Dict[str, Any]:
"""
Process a /responses request and return the result.
Args:
request_data: Dictionary containing model, input, and optional kwargs
Returns:
Dictionary with the agent's response
"""
try:
# Extract request parameters
model = request_data.get("model")
input_data = request_data.get("input")
agent_kwargs = request_data.get("agent_kwargs", {})
computer_kwargs = request_data.get("computer_kwargs", {})
env_overrides = request_data.get("env", {}) or {}
if not model:
raise ValueError("Model is required")
if not input_data:
raise ValueError("Input is required")
# Apply env overrides for the duration of this request
with self._env_overrides(env_overrides):
# Set up (and possibly reuse) computer and agent via caches
await self.setup_computer_agent(model, agent_kwargs, computer_kwargs)
# Defensive: ensure agent is initialized for type checkers
agent = self.agent
if agent is None:
raise RuntimeError("Agent failed to initialize")
# Convert input to messages format
messages = self._convert_input_to_messages(input_data)
# Run agent and get first result
async for result in agent.run(messages):
# Return the first result and break
return {
"success": True,
"result": result,
"model": model
}
# If no results were yielded
return {
"success": False,
"error": "No results from agent",
"model": model
}
except Exception as e:
logger.error(f"Error processing request: {e}")
return {
"success": False,
"error": str(e),
"model": request_data.get("model", "unknown")
}
def _convert_input_to_messages(self, input_data: Union[str, List[Dict[str, Any]]]) -> List[Dict[str, Any]]:
"""Convert input data to messages format."""
if isinstance(input_data, str):
# Simple string input
return [{"role": "user", "content": input_data}]
elif isinstance(input_data, list):
# Already in messages format
messages = []
for msg in input_data:
# Convert content array format if needed
if isinstance(msg.get("content"), list):
content_parts = []
for part in msg["content"]:
if part.get("type") == "input_text":
content_parts.append({"type": "text", "text": part["text"]})
elif part.get("type") == "input_image":
content_parts.append({
"type": "image_url",
"image_url": {"url": part["image_url"]}
})
else:
content_parts.append(part)
messages.append({
"role": msg["role"],
"content": content_parts
})
else:
messages.append(msg)
return messages
else:
raise ValueError("Input must be string or list of messages")
async def cleanup(self):
"""Clean up resources."""
if self.computer:
try:
await self.computer.__aexit__(None, None, None)
except Exception as e:
logger.error(f"Error cleaning up computer: {e}")
finally:
self.computer = None
self.agent = None
@staticmethod
@contextmanager
def _env_overrides(env: Dict[str, str]):
"""Temporarily apply environment variable overrides for the current process.
Restores previous values after the context exits.
Args:
env: Mapping of env var names to override for this request.
"""
if not env:
# No-op context
yield
return
original: Dict[str, Optional[str]] = {}
try:
for k, v in env.items():
original[k] = os.environ.get(k)
os.environ[k] = str(v)
yield
finally:
for k, old in original.items():
if old is None:
# Was not set before
os.environ.pop(k, None)
else:
os.environ[k] = old

8
libs/python/agent/pyproject.toml
View File

@@ -30,7 +30,6 @@ requires-python = ">=3.12"
openai = []
anthropic = []
omni = [
"ultralytics>=8.0.0",
"cua-som>=0.1.0,<0.2.0",
]
uitars = []
@@ -62,12 +61,9 @@ cli = [
"yaspin>=3.1.0",
]
hud = [
"hud-python==0.2.10",
"hud-python>=0.4.12,<0.5.0",
]
all = [
# omni requirements
"ultralytics>=8.0.0",
"cua-som>=0.1.0,<0.2.0",
# uitars requirements
"mlx-vlm>=0.1.27; sys_platform == 'darwin'",
"accelerate",
@@ -82,7 +78,7 @@ all = [
# cli requirements
"yaspin>=3.1.0",
# hud requirements
"hud-python==0.2.10",
"hud-python>=0.4.12,<0.5.0",
]
[tool.uv]

9
libs/python/computer-server/README.md
View File

@@ -35,4 +35,11 @@ pip install cua-computer-server
Refer to this notebook for a step-by-step guide on how to use the Computer-Use Server on the host system or VM:
- [Computer-Use Server](../../notebooks/computer_server_nb.ipynb)
- [Computer-Use Server](../../notebooks/computer_server_nb.ipynb)
## Docs
- [Commands](https://trycua.com/docs/libraries/computer-server/Commands)
- [REST-API](https://trycua.com/docs/libraries/computer-server/REST-API)
- [WebSocket-API](https://trycua.com/docs/libraries/computer-server/WebSocket-API)
- [Index](https://trycua.com/docs/libraries/computer-server/index)

24
libs/python/computer-server/computer_server/diorama/diorama_computer.py
View File

@@ -6,11 +6,26 @@ class DioramaComputer:
Implements _initialized, run(), and __aenter__ for agent compatibility.
"""
def __init__(self, diorama):
"""
Initialize the DioramaComputer with a diorama instance.
Args:
diorama: The diorama instance to wrap with a computer-like interface.
"""
self.diorama = diorama
self.interface = self.diorama.interface
self._initialized = False
async def __aenter__(self):
"""
Async context manager entry method for compatibility with ComputerAgent.
Ensures an event loop is running and marks the instance as initialized.
Creates a new event loop if none is currently running.
Returns:
DioramaComputer: The initialized instance.
"""
# Ensure the event loop is running (for compatibility)
try:
asyncio.get_running_loop()
@@ -20,6 +35,15 @@ class DioramaComputer:
return self
async def run(self):
"""
Run method stub for compatibility with ComputerAgent interface.
Ensures the instance is initialized before returning. If not already
initialized, calls __aenter__ to perform initialization.
Returns:
DioramaComputer: The initialized instance.
"""
# This is a stub for compatibility
if not self._initialized:
await self.__aenter__()

2
libs/python/computer-server/computer_server/handlers/base.py
View File

@@ -167,7 +167,7 @@ class BaseAutomationHandler(ABC):
pass
@abstractmethod
async def hotkey(self, *keys: str) -> Dict[str, Any]:
async def hotkey(self, keys: List[str]) -> Dict[str, Any]:
"""Press a combination of keys together."""
pass

3
libs/python/computer-server/computer_server/handlers/linux.py
View File

@@ -88,6 +88,7 @@ class LinuxAccessibilityHandler(BaseAccessibilityHandler):
class LinuxAutomationHandler(BaseAutomationHandler):
"""Linux implementation of automation handler using pyautogui."""
keyboard = KeyboardController()
mouse = MouseController()
# Mouse Actions
async def mouse_down(self, x: Optional[int] = None, y: Optional[int] = None, button: str = "left") -> Dict[str, Any]:
@@ -217,7 +218,7 @@ class LinuxAutomationHandler(BaseAutomationHandler):
# Scrolling Actions
async def scroll(self, x: int, y: int) -> Dict[str, Any]:
try:
pyautogui.scroll(x, y)
self.mouse.scroll(x, y)
return {"success": True}
except Exception as e:
return {"success": False, "error": str(e)}

400
libs/python/computer-server/computer_server/handlers/macos.py
View File

@@ -77,13 +77,37 @@ NSApplicationActivationOptions = {
}
def CFAttributeToPyObject(attrValue):
"""Convert Core Foundation attribute values to Python objects.
Args:
attrValue: Core Foundation attribute value to convert
Returns:
Converted Python object or None if conversion fails
"""
def list_helper(list_value):
"""Helper function to convert CF arrays to Python lists.
Args:
list_value: Core Foundation array to convert
Returns:
Python list containing converted items
"""
list_builder = []
for item in list_value:
list_builder.append(CFAttributeToPyObject(item))
return list_builder
def number_helper(number_value):
"""Helper function to convert CF numbers to Python numbers.
Args:
number_value: Core Foundation number to convert
Returns:
Python int or float, or None if conversion fails
"""
success, int_value = Foundation.CFNumberGetValue( # type: ignore
number_value, Foundation.kCFNumberIntType, None # type: ignore
)
@@ -98,6 +122,14 @@ def CFAttributeToPyObject(attrValue):
return None
def axuielement_helper(element_value):
"""Helper function to handle AX UI elements.
Args:
element_value: Accessibility UI element to process
Returns:
The element value unchanged
"""
return element_value
cf_attr_type = Foundation.CFGetTypeID(attrValue) # type: ignore
@@ -131,6 +163,15 @@ def CFAttributeToPyObject(attrValue):
def element_attribute(element, attribute):
"""Get an attribute value from an accessibility element.
Args:
element: The accessibility element
attribute: The attribute name to retrieve
Returns:
The attribute value or None if not found
"""
if attribute == kAXChildrenAttribute:
err, value = AXUIElementCopyAttributeValues(element, attribute, 0, 999, None)
if err == kAXErrorSuccess:
@@ -148,6 +189,15 @@ def element_attribute(element, attribute):
def element_value(element, type):
"""Extract a typed value from an accessibility element.
Args:
element: The accessibility element containing the value
type: The expected value type
Returns:
The extracted value or None if extraction fails
"""
err, value = AXValueGetValue(element, type, None)
if err == True:
return value
@@ -155,7 +205,18 @@ def element_value(element, type):
class UIElement:
"""Represents a UI element in the accessibility tree with position, size, and hierarchy information."""
def __init__(self, element, offset_x=0, offset_y=0, max_depth=None, parents_visible_bbox=None):
"""Initialize a UIElement from an accessibility element.
Args:
element: The accessibility element to wrap
offset_x: X offset for position calculations
offset_y: Y offset for position calculations
max_depth: Maximum depth to traverse for children
parents_visible_bbox: Parent's visible bounding box for clipping
"""
self.ax_element = element
self.content_identifier = ""
self.identifier = ""
@@ -235,6 +296,11 @@ class UIElement:
self.calculate_hashes()
def _set_bboxes(self, parents_visible_bbox):
"""Set bounding box and visible bounding box for the element.
Args:
parents_visible_bbox: Parent's visible bounding box for intersection calculation
"""
if not self.absolute_position or not self.size:
self.bbox = None
self.visible_bbox = None
@@ -265,6 +331,17 @@ class UIElement:
self.visible_bbox = self.bbox
def _get_children(self, element, start_position, offset_x, offset_y):
"""Get child elements from the accessibility element.
Args:
element: The parent accessibility element
start_position: Starting position for offset calculations
offset_x: X offset for child positioning
offset_y: Y offset for child positioning
Returns:
List of UIElement children
"""
children = element_attribute(element, kAXChildrenAttribute)
visible_children = element_attribute(element, kAXVisibleChildrenAttribute)
found_children = []
@@ -288,10 +365,16 @@ class UIElement:
return result
def calculate_hashes(self):
"""Calculate unique identifiers for the element and its content."""
self.identifier = self.component_hash()
self.content_identifier = self.children_content_hash(self.children)
def component_hash(self):
"""Generate a hash identifier for this component based on its properties.
Returns:
MD5 hash string of component properties
"""
if self.position is None or self.size is None:
return ""
position_string = f"{self.position.x:.0f};{self.position.y:.0f}"
@@ -304,6 +387,14 @@ class UIElement:
return self.hash_from_string(position_string + size_string + enabled_string + role_string)
def hash_from_string(self, string):
"""Generate MD5 hash from a string.
Args:
string: Input string to hash
Returns:
MD5 hash hexdigest or empty string if input is None/empty
"""
if string is None or string == "":
return ""
from hashlib import md5
@@ -311,6 +402,14 @@ class UIElement:
return md5(string.encode()).hexdigest()
def children_content_hash(self, children):
"""Generate a hash representing the content and structure of child elements.
Args:
children: List of child UIElement objects
Returns:
Combined hash of children content and structure
"""
if len(children) == 0:
return ""
all_content_hashes = []
@@ -326,7 +425,20 @@ class UIElement:
return self.hash_from_string(content_hash.join(content_structure_hash))
def to_dict(self):
"""Convert the UIElement to a dictionary representation.
Returns:
Dictionary containing all element properties and children
"""
def children_to_dict(children):
"""Convert list of children to dictionary format.
Args:
children: List of UIElement children to convert
Returns:
List of dictionaries representing the children
"""
result = []
for child in children:
result.append(child.to_dict())
@@ -375,6 +487,12 @@ from AppKit import NSWorkspace, NSRunningApplication
from pathlib import Path
def get_all_windows_zorder():
"""Get all windows in the system with their z-order information.
Returns:
List of window dictionaries sorted by z-index, containing window properties
like id, name, pid, owner, bounds, layer, and opacity
"""
window_list = Quartz.CGWindowListCopyWindowInfo(
Quartz.kCGWindowListOptionOnScreenOnly,
Quartz.kCGNullWindowID
@@ -425,6 +543,14 @@ def get_all_windows_zorder():
return windows
def get_app_info(app):
"""Extract information from an NSRunningApplication object.
Args:
app: NSRunningApplication instance
Returns:
Dictionary containing app name, bundle ID, PID, and status flags
"""
return {
"name": app.localizedName(),
"bundle_id": app.bundleIdentifier(),
@@ -435,6 +561,14 @@ def get_app_info(app):
}
def get_menubar_items(active_app_pid=None):
"""Get menubar items for the active application.
Args:
active_app_pid: Process ID of the active application, or None to use frontmost app
Returns:
List of menubar item dictionaries with title, bounds, index, and app_pid
"""
menubar_items = []
if active_app_pid is None:
frontmost_app = NSWorkspace.sharedWorkspace().frontmostApplication()
@@ -473,6 +607,12 @@ def get_menubar_items(active_app_pid=None):
return menubar_items
def get_dock_items():
"""Get all items in the macOS Dock.
Returns:
List of dock item dictionaries with title, description, bounds, index,
type, role, and subrole information
"""
dock_items = []
dock_pid = None
running_apps = NSWorkspace.sharedWorkspace().runningApplications()
@@ -538,7 +678,14 @@ def get_dock_items():
return dock_items
class MacOSAccessibilityHandler(BaseAccessibilityHandler):
"""Handler for macOS accessibility features and UI element inspection."""
def get_desktop_state(self):
"""Get the current state of the desktop including windows, apps, menubar, and dock.
Returns:
Dictionary containing applications, windows, menubar_items, and dock_items
"""
windows = [w for w in get_all_windows_zorder() if w.get("is_on_screen")]
running_apps = self.get_running_apps()
applications = []
@@ -586,7 +733,14 @@ class MacOSAccessibilityHandler(BaseAccessibilityHandler):
}
def get_application_windows(self, pid: int):
"""Get all windows for a specific application."""
"""Get all windows for a specific application.
Args:
pid: Process ID of the application
Returns:
List of accessibility window elements or empty list if none found
"""
try:
app = AXUIElementCreateApplication(pid)
err, windows = AXUIElementCopyAttributeValue(app, kAXWindowsAttribute, None)
@@ -598,7 +752,11 @@ class MacOSAccessibilityHandler(BaseAccessibilityHandler):
return []
def get_all_windows(self):
"""Get all visible windows in the system."""
"""Get all visible windows in the system.
Returns:
List of window dictionaries with app information and window details
"""
try:
windows = []
running_apps = self.get_running_apps()
@@ -632,16 +790,38 @@ class MacOSAccessibilityHandler(BaseAccessibilityHandler):
return []
def get_running_apps(self):
"""Get all currently running applications.
Returns:
List of NSRunningApplication objects
"""
# From NSWorkspace.runningApplications docs: https://developer.apple.com/documentation/appkit/nsworkspace/runningapplications
# "Similar to the NSRunningApplication class’s properties, this property will only change when the main run loop runs in a common mode"
# "Similar to the NSRunningApplication class's properties, this property will only change when the main run loop runs in a common mode"
# So we need to run the main run loop to get the latest running applications
Foundation.CFRunLoopRunInMode(Foundation.kCFRunLoopDefaultMode, 0.1, False) # type: ignore
return NSWorkspace.sharedWorkspace().runningApplications()
def get_ax_attribute(self, element, attribute):
"""Get an accessibility attribute from an element.
Args:
element: The accessibility element
attribute: The attribute name to retrieve
Returns:
The attribute value or None if not found
"""
return element_attribute(element, attribute)
def serialize_node(self, element):
"""Create a serializable dictionary representation of an accessibility element.
Args:
element: The accessibility element to serialize
Returns:
Dictionary containing element properties like role, title, value, position, and size
"""
# Create a serializable dictionary representation of an accessibility element
result = {}
@@ -669,7 +849,12 @@ class MacOSAccessibilityHandler(BaseAccessibilityHandler):
return result
async def get_accessibility_tree(self) -> Dict[str, Any]:
async def get_accessibility_tree(self) -> Dict[str, Any]:
"""Get the complete accessibility tree for the current desktop state.
Returns:
Dictionary containing success status and desktop state information
"""
try:
desktop_state = self.get_desktop_state()
return {
@@ -683,10 +868,28 @@ class MacOSAccessibilityHandler(BaseAccessibilityHandler):
async def find_element(
self, role: Optional[str] = None, title: Optional[str] = None, value: Optional[str] = None
) -> Dict[str, Any]:
"""Find an accessibility element matching the specified criteria.
Args:
role: The accessibility role to match (optional)
title: The title to match (optional)
value: The value to match (optional)
Returns:
Dictionary containing success status and the found element or error message
"""
try:
system = AXUIElementCreateSystemWide()
def match_element(element):
"""Check if an element matches the search criteria.
Args:
element: The accessibility element to check
Returns:
True if element matches all specified criteria, False otherwise
"""
if role and self.get_ax_attribute(element, kAXRoleAttribute) != role:
return False
if title and self.get_ax_attribute(element, kAXTitleAttribute) != title:
@@ -696,6 +899,14 @@ class MacOSAccessibilityHandler(BaseAccessibilityHandler):
return True
def search_tree(element):
"""Recursively search the accessibility tree for matching elements.
Args:
element: The accessibility element to search from
Returns:
Serialized element dictionary if match found, None otherwise
"""
if match_element(element):
return self.serialize_node(element)
@@ -714,11 +925,23 @@ class MacOSAccessibilityHandler(BaseAccessibilityHandler):
return {"success": False, "error": str(e)}
class MacOSAutomationHandler(BaseAutomationHandler):
"""Handler for macOS automation including mouse, keyboard, and screen operations."""
# Mouse Actions
mouse = MouseController()
keyboard = KeyboardController()
async def mouse_down(self, x: Optional[int] = None, y: Optional[int] = None, button: str = "left") -> Dict[str, Any]:
"""Press and hold a mouse button at the specified coordinates.
Args:
x: X coordinate (optional, uses current position if None)
y: Y coordinate (optional, uses current position if None)
button: Mouse button to press ("left", "right", or "middle")
Returns:
Dictionary containing success status and error message if failed
"""
try:
if x is not None and y is not None:
self.mouse.position = (x, y)
@@ -728,6 +951,16 @@ class MacOSAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def mouse_up(self, x: Optional[int] = None, y: Optional[int] = None, button: str = "left") -> Dict[str, Any]:
"""Release a mouse button at the specified coordinates.
Args:
x: X coordinate (optional, uses current position if None)
y: Y coordinate (optional, uses current position if None)
button: Mouse button to release ("left", "right", or "middle")
Returns:
Dictionary containing success status and error message if failed
"""
try:
if x is not None and y is not None:
self.mouse.position = (x, y)
@@ -737,6 +970,15 @@ class MacOSAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def left_click(self, x: Optional[int] = None, y: Optional[int] = None) -> Dict[str, Any]:
"""Perform a left mouse click at the specified coordinates.
Args:
x: X coordinate (optional, uses current position if None)
y: Y coordinate (optional, uses current position if None)
Returns:
Dictionary containing success status and error message if failed
"""
try:
if x is not None and y is not None:
self.mouse.position = (x, y)
@@ -746,6 +988,15 @@ class MacOSAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def right_click(self, x: Optional[int] = None, y: Optional[int] = None) -> Dict[str, Any]:
"""Perform a right mouse click at the specified coordinates.
Args:
x: X coordinate (optional, uses current position if None)
y: Y coordinate (optional, uses current position if None)
Returns:
Dictionary containing success status and error message if failed
"""
try:
if x is not None and y is not None:
self.mouse.position = (x, y)
@@ -757,6 +1008,15 @@ class MacOSAutomationHandler(BaseAutomationHandler):
async def double_click(
self, x: Optional[int] = None, y: Optional[int] = None
) -> Dict[str, Any]:
"""Perform a double left mouse click at the specified coordinates.
Args:
x: X coordinate (optional, uses current position if None)
y: Y coordinate (optional, uses current position if None)
Returns:
Dictionary containing success status and error message if failed
"""
try:
if x is not None and y is not None:
self.mouse.position = (x, y)
@@ -766,6 +1026,15 @@ class MacOSAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def move_cursor(self, x: int, y: int) -> Dict[str, Any]:
"""Move the mouse cursor to the specified coordinates.
Args:
x: Target X coordinate
y: Target Y coordinate
Returns:
Dictionary containing success status and error message if failed
"""
try:
self.mouse.position = (x, y)
return {"success": True}
@@ -775,6 +1044,17 @@ class MacOSAutomationHandler(BaseAutomationHandler):
async def drag_to(
self, x: int, y: int, button: str = "left", duration: float = 0.5
) -> Dict[str, Any]:
"""Drag from current position to target coordinates.
Args:
x: Target X coordinate
y: Target Y coordinate
button: Mouse button to use for dragging ("left", "right", or "middle")
duration: Duration of the drag operation in seconds
Returns:
Dictionary containing success status and error message if failed
"""
try:
btn = Button.left if button == "left" else Button.right if button == "right" else Button.middle
# Press
@@ -801,6 +1081,16 @@ class MacOSAutomationHandler(BaseAutomationHandler):
async def drag(
self, path: List[Tuple[int, int]], button: str = "left", duration: float = 0.5
) -> Dict[str, Any]:
"""Drag the mouse along a specified path of coordinates.
Args:
path: List of (x, y) coordinate tuples defining the drag path
button: Mouse button to use for dragging ("left", "right", or "middle")
duration: Total duration of the drag operation in seconds
Returns:
Dictionary containing success status and error message if failed
"""
try:
if not path or len(path) < 2:
return {"success": False, "error": "Path must contain at least 2 points"}
@@ -823,6 +1113,14 @@ class MacOSAutomationHandler(BaseAutomationHandler):
# Keyboard Actions
async def key_down(self, key: str) -> Dict[str, Any]:
"""Press and hold a keyboard key.
Args:
key: Key name to press (using pyautogui key names)
Returns:
Dictionary containing success status and error message if failed
"""
try:
# use pyautogui for their key names
pyautogui.keyDown(key)
@@ -831,6 +1129,14 @@ class MacOSAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def key_up(self, key: str) -> Dict[str, Any]:
"""Release a keyboard key.
Args:
key: Key name to release (using pyautogui key names)
Returns:
Dictionary containing success status and error message if failed
"""
try:
# use pyautogui for their key names
pyautogui.keyUp(key)
@@ -839,6 +1145,14 @@ class MacOSAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def type_text(self, text: str) -> Dict[str, Any]:
"""Type text using the keyboard with Unicode support.
Args:
text: Text string to type
Returns:
Dictionary containing success status and error message if failed
"""
try:
# use pynput for Unicode support
self.keyboard.type(text)
@@ -847,6 +1161,14 @@ class MacOSAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def press_key(self, key: str) -> Dict[str, Any]:
"""Press and release a keyboard key.
Args:
key: Key name to press (using pyautogui key names)
Returns:
Dictionary containing success status and error message if failed
"""
try:
# use pyautogui for their key names
pyautogui.press(key)
@@ -855,6 +1177,14 @@ class MacOSAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def hotkey(self, keys: List[str]) -> Dict[str, Any]:
"""Press a combination of keys simultaneously.
Args:
keys: List of key names to press together (using pyautogui key names)
Returns:
Dictionary containing success status and error message if failed
"""
try:
# use pyautogui for their key names
pyautogui.hotkey(*keys)
@@ -864,6 +1194,15 @@ class MacOSAutomationHandler(BaseAutomationHandler):
# Scrolling Actions
async def scroll(self, x: int, y: int) -> Dict[str, Any]:
"""Scroll the mouse wheel in the specified direction.
Args:
x: Horizontal scroll amount
y: Vertical scroll amount (positive for up, negative for down)
Returns:
Dictionary containing success status and error message if failed
"""
try:
self.mouse.scroll(x, y)
return {"success": True}
@@ -871,6 +1210,14 @@ class MacOSAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def scroll_down(self, clicks: int = 1) -> Dict[str, Any]:
"""Scroll down by the specified number of clicks.
Args:
clicks: Number of scroll clicks to perform
Returns:
Dictionary containing success status and error message if failed
"""
try:
self.mouse.scroll(0, -clicks)
return {"success": True}
@@ -878,6 +1225,14 @@ class MacOSAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def scroll_up(self, clicks: int = 1) -> Dict[str, Any]:
"""Scroll up by the specified number of clicks.
Args:
clicks: Number of scroll clicks to perform
Returns:
Dictionary containing success status and error message if failed
"""
try:
self.mouse.scroll(0, clicks)
return {"success": True}
@@ -886,6 +1241,11 @@ class MacOSAutomationHandler(BaseAutomationHandler):
# Screen Actions
async def screenshot(self) -> Dict[str, Any]:
"""Capture a screenshot of the current screen.
Returns:
Dictionary containing success status and base64-encoded image data or error message
"""
try:
from PIL import Image
@@ -902,6 +1262,11 @@ class MacOSAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": f"Screenshot error: {str(e)}"}
async def get_screen_size(self) -> Dict[str, Any]:
"""Get the dimensions of the current screen.
Returns:
Dictionary containing success status and screen size or error message
"""
try:
size = pyautogui.size()
return {"success": True, "size": {"width": size.width, "height": size.height}}
@@ -909,6 +1274,11 @@ class MacOSAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def get_cursor_position(self) -> Dict[str, Any]:
"""Get the current position of the mouse cursor.
Returns:
Dictionary containing success status and cursor position or error message
"""
try:
x, y = self.mouse.position
return {"success": True, "position": {"x": x, "y": y}}
@@ -917,6 +1287,11 @@ class MacOSAutomationHandler(BaseAutomationHandler):
# Clipboard Actions
async def copy_to_clipboard(self) -> Dict[str, Any]:
"""Get the current content of the system clipboard.
Returns:
Dictionary containing success status and clipboard content or error message
"""
try:
import pyperclip
@@ -926,6 +1301,14 @@ class MacOSAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def set_clipboard(self, text: str) -> Dict[str, Any]:
"""Set the content of the system clipboard.
Args:
text: Text to copy to the clipboard
Returns:
Dictionary containing success status and error message if failed
"""
try:
import pyperclip
@@ -935,7 +1318,14 @@ class MacOSAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def run_command(self, command: str) -> Dict[str, Any]:
"""Run a shell command and return its output."""
"""Run a shell command and return its output.
Args:
command: Shell command to execute
Returns:
Dictionary containing success status, stdout, stderr, and return code
"""
try:
# Create subprocess
process = await asyncio.create_subprocess_shell(

234
libs/python/computer-server/computer_server/handlers/windows.py
View File

@@ -11,6 +11,8 @@ import asyncio
import base64
import os
from io import BytesIO
from pynput.mouse import Controller as MouseController
from pynput.keyboard import Controller as KeyboardController
# Configure logger
logger = logging.getLogger(__name__)
@@ -41,7 +43,14 @@ class WindowsAccessibilityHandler(BaseAccessibilityHandler):
"""Windows implementation of accessibility handler."""
async def get_accessibility_tree(self) -> Dict[str, Any]:
"""Get the accessibility tree of the current window."""
"""Get the accessibility tree of the current window.
Returns:
Dict[str, Any]: A dictionary containing the success status and either
the accessibility tree or an error message.
Structure: {"success": bool, "tree": dict} or
{"success": bool, "error": str}
"""
if not WINDOWS_API_AVAILABLE:
return {"success": False, "error": "Windows API not available"}
@@ -65,6 +74,15 @@ class WindowsAccessibilityHandler(BaseAccessibilityHandler):
# Enumerate child windows
def enum_child_proc(hwnd_child, children_list):
"""Callback function to enumerate child windows and collect their information.
Args:
hwnd_child: Handle to the child window being enumerated.
children_list: List to append child window information to.
Returns:
bool: True to continue enumeration, False to stop.
"""
try:
child_text = win32gui.GetWindowText(hwnd_child)
child_rect = win32gui.GetWindowRect(hwnd_child)
@@ -93,7 +111,19 @@ class WindowsAccessibilityHandler(BaseAccessibilityHandler):
async def find_element(self, role: Optional[str] = None,
title: Optional[str] = None,
value: Optional[str] = None) -> Dict[str, Any]:
"""Find an element in the accessibility tree by criteria."""
"""Find an element in the accessibility tree by criteria.
Args:
role (Optional[str]): The role or class name of the element to find.
title (Optional[str]): The title or text of the element to find.
value (Optional[str]): The value of the element (not used in Windows implementation).
Returns:
Dict[str, Any]: A dictionary containing the success status and either
the found element or an error message.
Structure: {"success": bool, "element": dict} or
{"success": bool, "error": str}
"""
if not WINDOWS_API_AVAILABLE:
return {"success": False, "error": "Windows API not available"}
@@ -138,8 +168,20 @@ class WindowsAccessibilityHandler(BaseAccessibilityHandler):
class WindowsAutomationHandler(BaseAutomationHandler):
"""Windows implementation of automation handler using pyautogui and Windows APIs."""
mouse = MouseController()
# Mouse Actions
async def mouse_down(self, x: Optional[int] = None, y: Optional[int] = None, button: str = "left") -> Dict[str, Any]:
"""Press and hold a mouse button at the specified coordinates.
Args:
x (Optional[int]): The x-coordinate to move to before pressing. If None, uses current position.
y (Optional[int]): The y-coordinate to move to before pressing. If None, uses current position.
button (str): The mouse button to press ("left", "right", or "middle").
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -152,6 +194,16 @@ class WindowsAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def mouse_up(self, x: Optional[int] = None, y: Optional[int] = None, button: str = "left") -> Dict[str, Any]:
"""Release a mouse button at the specified coordinates.
Args:
x (Optional[int]): The x-coordinate to move to before releasing. If None, uses current position.
y (Optional[int]): The y-coordinate to move to before releasing. If None, uses current position.
button (str): The mouse button to release ("left", "right", or "middle").
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -164,6 +216,15 @@ class WindowsAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def move_cursor(self, x: int, y: int) -> Dict[str, Any]:
"""Move the mouse cursor to the specified coordinates.
Args:
x (int): The x-coordinate to move to.
y (int): The y-coordinate to move to.
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -174,6 +235,15 @@ class WindowsAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def left_click(self, x: Optional[int] = None, y: Optional[int] = None) -> Dict[str, Any]:
"""Perform a left mouse click at the specified coordinates.
Args:
x (Optional[int]): The x-coordinate to click at. If None, clicks at current position.
y (Optional[int]): The y-coordinate to click at. If None, clicks at current position.
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -186,6 +256,15 @@ class WindowsAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def right_click(self, x: Optional[int] = None, y: Optional[int] = None) -> Dict[str, Any]:
"""Perform a right mouse click at the specified coordinates.
Args:
x (Optional[int]): The x-coordinate to click at. If None, clicks at current position.
y (Optional[int]): The y-coordinate to click at. If None, clicks at current position.
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -198,6 +277,15 @@ class WindowsAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def double_click(self, x: Optional[int] = None, y: Optional[int] = None) -> Dict[str, Any]:
"""Perform a double left mouse click at the specified coordinates.
Args:
x (Optional[int]): The x-coordinate to double-click at. If None, clicks at current position.
y (Optional[int]): The y-coordinate to double-click at. If None, clicks at current position.
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -210,6 +298,17 @@ class WindowsAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def drag_to(self, x: int, y: int, button: str = "left", duration: float = 0.5) -> Dict[str, Any]:
"""Drag from the current position to the specified coordinates.
Args:
x (int): The x-coordinate to drag to.
y (int): The y-coordinate to drag to.
button (str): The mouse button to use for dragging ("left", "right", or "middle").
duration (float): The time in seconds to take for the drag operation.
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -220,6 +319,16 @@ class WindowsAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def drag(self, path: List[Tuple[int, int]], button: str = "left", duration: float = 0.5) -> Dict[str, Any]:
"""Drag the mouse through a series of coordinates.
Args:
path (List[Tuple[int, int]]): A list of (x, y) coordinate tuples to drag through.
button (str): The mouse button to use for dragging ("left", "right", or "middle").
duration (float): The total time in seconds for the entire drag operation.
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -240,6 +349,14 @@ class WindowsAutomationHandler(BaseAutomationHandler):
# Keyboard Actions
async def key_down(self, key: str) -> Dict[str, Any]:
"""Press and hold a keyboard key.
Args:
key (str): The key to press down (e.g., 'ctrl', 'shift', 'a').
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -250,6 +367,14 @@ class WindowsAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def key_up(self, key: str) -> Dict[str, Any]:
"""Release a keyboard key.
Args:
key (str): The key to release (e.g., 'ctrl', 'shift', 'a').
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -260,6 +385,14 @@ class WindowsAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def type_text(self, text: str) -> Dict[str, Any]:
"""Type the specified text.
Args:
text (str): The text to type.
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -270,6 +403,14 @@ class WindowsAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def press_key(self, key: str) -> Dict[str, Any]:
"""Press and release a keyboard key.
Args:
key (str): The key to press (e.g., 'enter', 'space', 'tab').
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -279,7 +420,15 @@ class WindowsAutomationHandler(BaseAutomationHandler):
except Exception as e:
return {"success": False, "error": str(e)}
async def hotkey(self, keys: str) -> Dict[str, Any]:
async def hotkey(self, keys: List[str]) -> Dict[str, Any]:
"""Press a combination of keys simultaneously.
Args:
keys (List[str]): The keys to press together (e.g., ['ctrl', 'c'], ['alt', 'tab']).
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -291,17 +440,33 @@ class WindowsAutomationHandler(BaseAutomationHandler):
# Scrolling Actions
async def scroll(self, x: int, y: int) -> Dict[str, Any]:
"""Scroll vertically at the current cursor position.
Args:
x (int): Horizontal scroll amount (not used in pyautogui implementation).
y (int): Vertical scroll amount. Positive values scroll up, negative values scroll down.
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
try:
# pyautogui.scroll() only takes one parameter (vertical scroll)
pyautogui.scroll(y)
self.mouse.scroll(x, y)
return {"success": True}
except Exception as e:
return {"success": False, "error": str(e)}
async def scroll_down(self, clicks: int = 1) -> Dict[str, Any]:
"""Scroll down by the specified number of clicks.
Args:
clicks (int): The number of scroll clicks to perform downward.
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -312,6 +477,14 @@ class WindowsAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def scroll_up(self, clicks: int = 1) -> Dict[str, Any]:
"""Scroll up by the specified number of clicks.
Args:
clicks (int): The number of scroll clicks to perform upward.
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -323,6 +496,14 @@ class WindowsAutomationHandler(BaseAutomationHandler):
# Screen Actions
async def screenshot(self) -> Dict[str, Any]:
"""Capture a screenshot of the entire screen.
Returns:
Dict[str, Any]: A dictionary containing the success status and either
base64-encoded image data or an error message.
Structure: {"success": bool, "image_data": str} or
{"success": bool, "error": str}
"""
if not pyautogui:
return {"success": False, "error": "pyautogui not available"}
@@ -341,6 +522,14 @@ class WindowsAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": f"Screenshot error: {str(e)}"}
async def get_screen_size(self) -> Dict[str, Any]:
"""Get the size of the screen in pixels.
Returns:
Dict[str, Any]: A dictionary containing the success status and either
screen size information or an error message.
Structure: {"success": bool, "size": {"width": int, "height": int}} or
{"success": bool, "error": str}
"""
try:
if pyautogui:
size = pyautogui.size()
@@ -356,6 +545,14 @@ class WindowsAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def get_cursor_position(self) -> Dict[str, Any]:
"""Get the current position of the mouse cursor.
Returns:
Dict[str, Any]: A dictionary containing the success status and either
cursor position or an error message.
Structure: {"success": bool, "position": {"x": int, "y": int}} or
{"success": bool, "error": str}
"""
try:
if pyautogui:
pos = pyautogui.position()
@@ -371,6 +568,14 @@ class WindowsAutomationHandler(BaseAutomationHandler):
# Clipboard Actions
async def copy_to_clipboard(self) -> Dict[str, Any]:
"""Get the current content of the clipboard.
Returns:
Dict[str, Any]: A dictionary containing the success status and either
clipboard content or an error message.
Structure: {"success": bool, "content": str} or
{"success": bool, "error": str}
"""
try:
import pyperclip
content = pyperclip.paste()
@@ -379,6 +584,14 @@ class WindowsAutomationHandler(BaseAutomationHandler):
return {"success": False, "error": str(e)}
async def set_clipboard(self, text: str) -> Dict[str, Any]:
"""Set the clipboard content to the specified text.
Args:
text (str): The text to copy to the clipboard.
Returns:
Dict[str, Any]: A dictionary with success status and optional error message.
"""
try:
import pyperclip
pyperclip.copy(text)
@@ -388,6 +601,17 @@ class WindowsAutomationHandler(BaseAutomationHandler):
# Command Execution
async def run_command(self, command: str) -> Dict[str, Any]:
"""Execute a shell command asynchronously.
Args:
command (str): The shell command to execute.
Returns:
Dict[str, Any]: A dictionary containing the success status and either
command output or an error message.
Structure: {"success": bool, "stdout": str, "stderr": str, "return_code": int} or
{"success": bool, "error": str}
"""
try:
# Create subprocess
process = await asyncio.create_subprocess_shell(

287
libs/python/computer-server/computer_server/main.py
View File

@@ -1,6 +1,6 @@
from fastapi import FastAPI, WebSocket, WebSocketDisconnect, Request, HTTPException, Header
from fastapi.responses import StreamingResponse
from typing import List, Dict, Any, Optional
from fastapi.responses import StreamingResponse, JSONResponse
from typing import List, Dict, Any, Optional, Union, Literal, cast
import uvicorn
import logging
import asyncio
@@ -14,6 +14,14 @@ import os
import aiohttp
import hashlib
import time
import platform
from fastapi.middleware.cors import CORSMiddleware
try:
from agent import ComputerAgent
HAS_AGENT = True
except ImportError:
HAS_AGENT = False
# Set up logging with more detail
logger = logging.getLogger(__name__)
@@ -30,6 +38,16 @@ app = FastAPI(
websocket_max_size=WEBSOCKET_MAX_SIZE,
)
# CORS configuration
origins = ["*"]
app.add_middleware(
CORSMiddleware,
allow_origins=origins,
allow_credentials=True,
allow_methods=["*"],
allow_headers=["*"],
)
protocol_version = 1
try:
from importlib.metadata import version
@@ -197,6 +215,21 @@ class ConnectionManager:
manager = ConnectionManager()
auth_manager = AuthenticationManager()
@app.get("/status")
async def status():
sys = platform.system().lower()
# get os type
if "darwin" in sys or sys == "macos" or sys == "mac":
os_type = "macos"
elif "windows" in sys:
os_type = "windows"
else:
os_type = "linux"
# get computer-server features
features = []
if HAS_AGENT:
features.append("agent")
return {"status": "ok", "os_type": os_type, "features": features}
@app.websocket("/ws", name="websocket_endpoint")
async def websocket_endpoint(websocket: WebSocket):
@@ -331,7 +364,6 @@ async def websocket_endpoint(websocket: WebSocket):
pass
manager.disconnect(websocket)
@app.post("/cmd")
async def cmd_endpoint(
request: Request,
@@ -420,12 +452,255 @@ async def cmd_endpoint(
headers={
"Cache-Control": "no-cache",
"Connection": "keep-alive",
"Access-Control-Allow-Origin": "*",
"Access-Control-Allow-Methods": "POST, OPTIONS",
"Access-Control-Allow-Headers": "Content-Type, X-Container-Name, X-API-Key"
}
)
@app.post("/responses")
async def agent_response_endpoint(
request: Request,
api_key: Optional[str] = Header(None, alias="X-API-Key"),
):
"""
Minimal proxy to run ComputerAgent for up to 2 turns.
Security:
- If CONTAINER_NAME is set on the server, require X-API-Key
and validate using AuthenticationManager unless CUA_ENABLE_PUBLIC_PROXY is true.
Body JSON:
{
"model": "...", # required
"input": "... or messages[]", # required
"agent_kwargs": { ... }, # optional, passed directly to ComputerAgent
"env": { ... } # optional env overrides for agent
}
"""
if not HAS_AGENT:
raise HTTPException(status_code=501, detail="ComputerAgent not available")
# Authenticate via AuthenticationManager if running in cloud (CONTAINER_NAME set)
container_name = os.environ.get("CONTAINER_NAME")
if container_name:
is_public = os.environ.get("CUA_ENABLE_PUBLIC_PROXY", "").lower().strip() in ["1", "true", "yes", "y", "on"]
if not is_public:
if not api_key:
raise HTTPException(status_code=401, detail="Missing AGENT PROXY auth headers")
ok = await auth_manager.auth(container_name, api_key)
if not ok:
raise HTTPException(status_code=401, detail="Unauthorized")
# Parse request body
try:
body = await request.json()
except Exception as e:
raise HTTPException(status_code=400, detail=f"Invalid JSON body: {str(e)}")
model = body.get("model")
input_data = body.get("input")
if not model or input_data is None:
raise HTTPException(status_code=400, detail="'model' and 'input' are required")
agent_kwargs: Dict[str, Any] = body.get("agent_kwargs") or {}
env_overrides: Dict[str, str] = body.get("env") or {}
# Simple env override context
class _EnvOverride:
def __init__(self, overrides: Dict[str, str]):
self.overrides = overrides
self._original: Dict[str, Optional[str]] = {}
def __enter__(self):
for k, v in (self.overrides or {}).items():
self._original[k] = os.environ.get(k)
os.environ[k] = str(v)
def __exit__(self, exc_type, exc, tb):
for k, old in self._original.items():
if old is None:
os.environ.pop(k, None)
else:
os.environ[k] = old
# Convert input to messages
def _to_messages(data: Union[str, List[Dict[str, Any]]]) -> List[Dict[str, Any]]:
if isinstance(data, str):
return [{"role": "user", "content": data}]
if isinstance(data, list):
return data
messages = _to_messages(input_data)
# Define a direct computer tool that implements the AsyncComputerHandler protocol
# and delegates to our existing automation/file/accessibility handlers.
from agent.computers import AsyncComputerHandler # runtime-checkable Protocol
class DirectComputer(AsyncComputerHandler):
def __init__(self):
# use module-scope handler singletons created by HandlerFactory
self._auto = automation_handler
self._file = file_handler
self._access = accessibility_handler
async def get_environment(self) -> Literal["windows", "mac", "linux", "browser"]:
sys = platform.system().lower()
if "darwin" in sys or sys in ("macos", "mac"):
return "mac"
if "windows" in sys:
return "windows"
return "linux"
async def get_dimensions(self) -> tuple[int, int]:
size = await self._auto.get_screen_size()
return size["width"], size["height"]
async def screenshot(self) -> str:
img_b64 = await self._auto.screenshot()
return img_b64["image_data"]
async def click(self, x: int, y: int, button: str = "left") -> None:
if button == "left":
await self._auto.left_click(x, y)
elif button == "right":
await self._auto.right_click(x, y)
else:
await self._auto.left_click(x, y)
async def double_click(self, x: int, y: int) -> None:
await self._auto.double_click(x, y)
async def scroll(self, x: int, y: int, scroll_x: int, scroll_y: int) -> None:
await self._auto.move_cursor(x, y)
await self._auto.scroll(scroll_x, scroll_y)
async def type(self, text: str) -> None:
await self._auto.type_text(text)
async def wait(self, ms: int = 1000) -> None:
await asyncio.sleep(ms / 1000.0)
async def move(self, x: int, y: int) -> None:
await self._auto.move_cursor(x, y)
async def keypress(self, keys: Union[List[str], str]) -> None:
if isinstance(keys, str):
parts = keys.replace("-", "+").split("+") if len(keys) > 1 else [keys]
else:
parts = keys
if len(parts) == 1:
await self._auto.press_key(parts[0])
else:
await self._auto.hotkey(parts)
async def drag(self, path: List[Dict[str, int]]) -> None:
if not path:
return
start = path[0]
await self._auto.mouse_down(start["x"], start["y"])
for pt in path[1:]:
await self._auto.move_cursor(pt["x"], pt["y"])
end = path[-1]
await self._auto.mouse_up(end["x"], end["y"])
async def get_current_url(self) -> str:
# Not available in this server context
return ""
async def left_mouse_down(self, x: Optional[int] = None, y: Optional[int] = None) -> None:
await self._auto.mouse_down(x, y, button="left")
async def left_mouse_up(self, x: Optional[int] = None, y: Optional[int] = None) -> None:
await self._auto.mouse_up(x, y, button="left")
# # Inline image URLs to base64
# import base64, mimetypes, requests
# # Use a browser-like User-Agent to avoid 403s from some CDNs (e.g., Wikimedia)
# HEADERS = {
# "User-Agent": (
# "Mozilla/5.0 (Windows NT 10.0; Win64; x64) "
# "AppleWebKit/537.36 (KHTML, like Gecko) "
# "Chrome/124.0.0.0 Safari/537.36"
# )
# }
# def _to_data_url(content_bytes: bytes, url: str, resp: requests.Response) -> str:
# ctype = resp.headers.get("Content-Type") or mimetypes.guess_type(url)[0] or "application/octet-stream"
# b64 = base64.b64encode(content_bytes).decode("utf-8")
# return f"data:{ctype};base64,{b64}"
# def inline_image_urls(messages):
# # messages: List[{"role": "...","content":[...]}]
# out = []
# for m in messages:
# if not isinstance(m.get("content"), list):
# out.append(m)
# continue
# new_content = []
# for part in (m.get("content") or []):
# if part.get("type") == "input_image" and (url := part.get("image_url")):
# resp = requests.get(url, headers=HEADERS, timeout=30)
# resp.raise_for_status()
# new_content.append({
# "type": "input_image",
# "image_url": _to_data_url(resp.content, url, resp)
# })
# else:
# new_content.append(part)
# out.append({**m, "content": new_content})
# return out
# messages = inline_image_urls(messages)
error = None
with _EnvOverride(env_overrides):
# Prepare tools: if caller did not pass tools, inject our DirectComputer
tools = agent_kwargs.get("tools")
if not tools:
tools = [DirectComputer()]
agent_kwargs = {**agent_kwargs, "tools": tools}
# Instantiate agent with our tools
agent = ComputerAgent(model=model, **agent_kwargs) # type: ignore[arg-type]
total_output: List[Any] = []
total_usage: Dict[str, Any] = {}
pending_computer_call_ids = set()
try:
async for result in agent.run(messages):
total_output += result["output"]
# Try to collect usage if present
if isinstance(result, dict) and "usage" in result and isinstance(result["usage"], dict):
# Merge usage counters
for k, v in result["usage"].items():
if isinstance(v, (int, float)):
total_usage[k] = total_usage.get(k, 0) + v
else:
total_usage[k] = v
for msg in result.get("output", []):
if msg.get("type") == "computer_call":
pending_computer_call_ids.add(msg["call_id"])
elif msg.get("type") == "computer_call_output":
pending_computer_call_ids.discard(msg["call_id"])
# exit if no pending computer calls
if not pending_computer_call_ids:
break
except Exception as e:
logger.error(f"Error running agent: {str(e)}")
logger.error(traceback.format_exc())
error = str(e)
# Build response payload
payload = {
"model": model,
"error": error,
"output": total_output,
"usage": total_usage,
"status": "completed" if not error else "failed"
}
# CORS: allow any origin
headers = {
"Cache-Control": "no-cache",
"Connection": "keep-alive",
}
return JSONResponse(content=payload, headers=headers)
if __name__ == "__main__":
uvicorn.run(app, host="0.0.0.0", port=8000)

81
libs/python/computer/README.md
View File

@@ -65,80 +65,9 @@ Refer to this notebook for a step-by-step guide on how to use the Computer-Use I
- [Computer-Use Interface (CUI)](https://github.com/trycua/cua/blob/main/notebooks/computer_nb.ipynb)
## Using the Gradio Computer UI
The computer module includes a Gradio UI for creating and sharing demonstration data. We make it easy for people to build community datasets for better computer use models with an upload to Huggingface feature.
```bash
# Install with UI support
pip install "cua-computer[ui]"
```
> **Note:** For precise control of the computer, we recommend using VNC or Screen Sharing instead of the Computer Gradio UI.
### Building and Sharing Demonstrations with Huggingface
Follow these steps to contribute your own demonstrations:
#### 1. Set up Huggingface Access
Set your HF_TOKEN in a .env file or in your environment variables:
```bash
# In .env file
HF_TOKEN=your_huggingface_token
```
#### 2. Launch the Computer UI
```python
# launch_ui.py
from computer.ui.gradio.app import create_gradio_ui
from dotenv import load_dotenv
load_dotenv('.env')
app = create_gradio_ui()
app.launch(share=False)
```
For examples, see [Computer UI Examples](https://github.com/trycua/cua/tree/main/examples/computer_ui_examples.py)
#### 3. Record Your Tasks
<details open>
<summary>View demonstration video</summary>
<video src="https://github.com/user-attachments/assets/de3c3477-62fe-413c-998d-4063e48de176" controls width="600"></video>
</details>
Record yourself performing various computer tasks using the UI.
#### 4. Save Your Demonstrations
<details open>
<summary>View demonstration video</summary>
<video src="https://github.com/user-attachments/assets/5ad1df37-026a-457f-8b49-922ae805faef" controls width="600"></video>
</details>
Save each task by picking a descriptive name and adding relevant tags (e.g., "office", "web-browsing", "coding").
#### 5. Record Additional Demonstrations
Repeat steps 3 and 4 until you have a good amount of demonstrations covering different tasks and scenarios.
#### 6. Upload to Huggingface
<details open>
<summary>View demonstration video</summary>
<video src="https://github.com/user-attachments/assets/c586d460-3877-4b5f-a736-3248886d2134" controls width="600"></video>
</details>
Upload your dataset to Huggingface by:
- Naming it as `{your_username}/{dataset_name}`
- Choosing public or private visibility
- Optionally selecting specific tags to upload only tasks with certain tags
#### Examples and Resources
- Example Dataset: [ddupont/test-dataset](https://huggingface.co/datasets/ddupont/test-dataset)
- Find Community Datasets: 🔍 [Browse CUA Datasets on Huggingface](https://huggingface.co/datasets?other=cua)
## Docs
- [Computers](https://trycua.com/docs/computer-sdk/computers)
- [Commands](https://trycua.com/docs/computer-sdk/commands)
- [Computer UI](https://trycua.com/docs/computer-sdk/computer-ui)
- [Sandboxed Python](https://trycua.com/docs/computer-sdk/sandboxed-python)

139
libs/python/computer/computer/diorama_computer.py
View File

@@ -6,16 +6,35 @@ class DioramaComputer:
A Computer-compatible proxy for Diorama that sends commands over the ComputerInterface.
"""
def __init__(self, computer, apps):
"""
Initialize the DioramaComputer with a computer instance and list of apps.
Args:
computer: The computer instance to proxy commands through
apps: List of applications available in the diorama environment
"""
self.computer = computer
self.apps = apps
self.interface = DioramaComputerInterface(computer, apps)
self._initialized = False
async def __aenter__(self):
"""
Async context manager entry point.
Returns:
self: The DioramaComputer instance
"""
self._initialized = True
return self
async def run(self):
"""
Initialize and run the DioramaComputer if not already initialized.
Returns:
self: The DioramaComputer instance
"""
if not self._initialized:
await self.__aenter__()
return self
@@ -25,11 +44,31 @@ class DioramaComputerInterface:
Diorama Interface proxy that sends diorama_cmds via the Computer's interface.
"""
def __init__(self, computer, apps):
"""
Initialize the DioramaComputerInterface.
Args:
computer: The computer instance to send commands through
apps: List of applications available in the diorama environment
"""
self.computer = computer
self.apps = apps
self._scene_size = None
async def _send_cmd(self, action, arguments=None):
"""
Send a command to the diorama interface through the computer.
Args:
action (str): The action/command to execute
arguments (dict, optional): Additional arguments for the command
Returns:
The result from the diorama command execution
Raises:
RuntimeError: If the computer interface is not initialized or command fails
"""
arguments = arguments or {}
arguments = {"app_list": self.apps, **arguments}
# Use the computer's interface (must be initialized)
@@ -42,6 +81,15 @@ class DioramaComputerInterface:
return result.get("result")
async def screenshot(self, as_bytes=True):
"""
Take a screenshot of the diorama scene.
Args:
as_bytes (bool): If True, return image as bytes; if False, return PIL Image object
Returns:
bytes or PIL.Image: Screenshot data in the requested format
"""
from PIL import Image
import base64
result = await self._send_cmd("screenshot")
@@ -53,41 +101,122 @@ class DioramaComputerInterface:
return img_bytes if as_bytes else img
async def get_screen_size(self):
"""
Get the dimensions of the diorama scene.
Returns:
dict: Dictionary containing 'width' and 'height' keys with pixel dimensions
"""
if not self._scene_size:
await self.screenshot(as_bytes=False)
return {"width": self._scene_size[0], "height": self._scene_size[1]}
async def move_cursor(self, x, y):
"""
Move the cursor to the specified coordinates.
Args:
x (int): X coordinate to move cursor to
y (int): Y coordinate to move cursor to
"""
await self._send_cmd("move_cursor", {"x": x, "y": y})
async def left_click(self, x=None, y=None):
"""
Perform a left mouse click at the specified coordinates or current cursor position.
Args:
x (int, optional): X coordinate to click at. If None, clicks at current cursor position
y (int, optional): Y coordinate to click at. If None, clicks at current cursor position
"""
await self._send_cmd("left_click", {"x": x, "y": y})
async def right_click(self, x=None, y=None):
"""
Perform a right mouse click at the specified coordinates or current cursor position.
Args:
x (int, optional): X coordinate to click at. If None, clicks at current cursor position
y (int, optional): Y coordinate to click at. If None, clicks at current cursor position
"""
await self._send_cmd("right_click", {"x": x, "y": y})
async def double_click(self, x=None, y=None):
"""
Perform a double mouse click at the specified coordinates or current cursor position.
Args:
x (int, optional): X coordinate to double-click at. If None, clicks at current cursor position
y (int, optional): Y coordinate to double-click at. If None, clicks at current cursor position
"""
await self._send_cmd("double_click", {"x": x, "y": y})
async def scroll_up(self, clicks=1):
"""
Scroll up by the specified number of clicks.
Args:
clicks (int): Number of scroll clicks to perform upward. Defaults to 1
"""
await self._send_cmd("scroll_up", {"clicks": clicks})
async def scroll_down(self, clicks=1):
"""
Scroll down by the specified number of clicks.
Args:
clicks (int): Number of scroll clicks to perform downward. Defaults to 1
"""
await self._send_cmd("scroll_down", {"clicks": clicks})
async def drag_to(self, x, y, duration=0.5):
"""
Drag from the current cursor position to the specified coordinates.
Args:
x (int): X coordinate to drag to
y (int): Y coordinate to drag to
duration (float): Duration of the drag operation in seconds. Defaults to 0.5
"""
await self._send_cmd("drag_to", {"x": x, "y": y, "duration": duration})
async def get_cursor_position(self):
"""
Get the current cursor position.
Returns:
dict: Dictionary containing the current cursor coordinates
"""
return await self._send_cmd("get_cursor_position")
async def type_text(self, text):
"""
Type the specified text at the current cursor position.
Args:
text (str): The text to type
"""
await self._send_cmd("type_text", {"text": text})
async def press_key(self, key):
"""
Press a single key.
Args:
key: The key to press
"""
await self._send_cmd("press_key", {"key": key})
async def hotkey(self, *keys):
"""
Press multiple keys simultaneously as a hotkey combination.
Args:
*keys: Variable number of keys to press together. Can be Key enum instances or strings
Raises:
ValueError: If any key is not a Key enum or string type
"""
actual_keys = []
for key in keys:
if isinstance(key, Key):
@@ -101,4 +230,14 @@ class DioramaComputerInterface:
await self._send_cmd("hotkey", {"keys": actual_keys})
async def to_screen_coordinates(self, x, y):
"""
Convert coordinates to screen coordinates.
Args:
x (int): X coordinate to convert
y (int): Y coordinate to convert
Returns:
dict: Dictionary containing the converted screen coordinates
"""
return await self._send_cmd("to_screen_coordinates", {"x": x, "y": y})

114
libs/python/mcp-server/README.md
View File

@@ -17,60 +17,20 @@
**cua-mcp-server** is a MCP server for the Computer-Use Agent (CUA), allowing you to run CUA through Claude Desktop or other MCP clients.
## LiteLLM Integration
This MCP server features comprehensive liteLLM integration, allowing you to use any supported LLM provider with a simple model string configuration.
- **Unified Configuration**: Use a single `CUA_MODEL_NAME` environment variable with a model string
- **Automatic Provider Detection**: The agent automatically detects the provider and capabilities from the model string
- **Extensive Provider Support**: Works with Anthropic, OpenAI, local models, and any liteLLM-compatible provider
### Model String Examples:
- **Anthropic**: `"anthropic/claude-3-5-sonnet-20241022"`
- **OpenAI**: `"openai/computer-use-preview"`
- **UI-TARS**: `"huggingface-local/ByteDance-Seed/UI-TARS-1.5-7B"`
- **Omni + Any LiteLLM**: `"omniparser+litellm/gpt-4o"`, `"omniparser+litellm/claude-3-haiku"`, `"omniparser+ollama_chat/gemma3"`
### Get started with Agent
## Prerequisites
Before installing the MCP server, you'll need to set up full Computer-Use Agent capabilities. This includes:
Cua MCP Server requires [lume](https://github.com/trycua/cua/blob/main/libs/lume/README.md#install) to be installed.
1. Installing the Lume CLI
2. Pulling the latest macOS CUA image
3. Starting the Lume daemon service
4. Installing the required Python libraries (Optional: only needed if you want to verify the agent is working before installing MCP server)
## Install
Make sure these steps are completed and working before proceeding with the MCP server installation.
## Installation
Install the package from PyPI:
```bash
pip install cua-mcp-server
```
This will install:
- The MCP server
- CUA agent and computer dependencies
- An executable `cua-mcp-server` script in your PATH
## Easy Setup Script
If you want to simplify installation, you can use this one-liner to download and run the installation script:
Download and run the installation script:
```bash
curl -fsSL https://raw.githubusercontent.com/trycua/cua/main/libs/python/mcp-server/scripts/install_mcp_server.sh | bash
```
This script will:
- Create the ~/.cua directory if it doesn't exist
- Generate a startup script at ~/.cua/start_mcp_server.sh
- Make the script executable
- The startup script automatically manages Python virtual environments and installs/updates the cua-mcp-server package
You can then use the script in your MCP configuration like this:
```json
@@ -87,9 +47,9 @@ You can then use the script in your MCP configuration like this:
}
```
## Development Guide
## Development
If you want to develop with the cua-mcp-server directly without installation, you can use this configuration:
Use this configuration to develop with the cua-mcp-server directly without installation:
```json
{
@@ -112,61 +72,11 @@ This configuration:
Just add this to your MCP client's configuration and it will use your local development version of the server.
### Troubleshooting
## Docs
If you get a `/bin/bash: ~/cua/libs/python/mcp-server/scripts/start_mcp_server.sh: No such file or directory` error, try changing the path to the script to be absolute instead of relative.
To see the logs:
```
tail -n 20 -f ~/Library/Logs/Claude/mcp*.log
```
## Claude Desktop Integration
To use with Claude Desktop, add an entry to your Claude Desktop configuration (`claude_desktop_config.json`, typically found in `~/.config/claude-desktop/`):
For more information on MCP with Claude Desktop, see the [official MCP User Guide](https://modelcontextprotocol.io/quickstart/user).
## Cursor Integration
To use with Cursor, add an MCP configuration file in one of these locations:
- **Project-specific**: Create `.cursor/mcp.json` in your project directory
- **Global**: Create `~/.cursor/mcp.json` in your home directory
After configuration, you can simply tell Cursor's Agent to perform computer tasks by explicitly mentioning the CUA agent, such as "Use the computer control tools to open Safari."
For more information on MCP with Cursor, see the [official Cursor MCP documentation](https://docs.cursor.com/context/model-context-protocol).
### First-time Usage Notes
**API Keys**: Ensure you have valid API keys:
- Add your Anthropic API key, or other model provider API key in the Claude Desktop config (as shown above)
- Or set it as an environment variable in your shell profile
## Configuration
The server is configured using environment variables (can be set in the Claude Desktop config):
| Variable | Description | Default |
|----------|-------------|---------|
| `CUA_MODEL_NAME` | Model string (e.g., "anthropic/claude-3-5-sonnet-20241022", "openai/computer-use-preview", "huggingface-local/ByteDance-Seed/UI-TARS-1.5-7B", "omniparser+litellm/gpt-4o", "omniparser+ollama_chat/gemma3") | anthropic/claude-3-5-sonnet-20241022 |
| `CUA_MAX_IMAGES` | Maximum number of images to keep in context | 3 |
## Available Tools
The MCP server exposes the following tools to Claude:
1. `run_cua_task` - Run a single Computer-Use Agent task with the given instruction
2. `run_multi_cua_tasks` - Run multiple tasks in sequence
## Usage
Once configured, you can simply ask Claude to perform computer tasks:
- "Open Chrome and go to github.com"
- "Create a folder called 'Projects' on my desktop"
- "Find all PDFs in my Downloads folder"
- "Take a screenshot and highlight the error message"
Claude will automatically use your CUA agent to perform these tasks.
- [Installation](https://trycua.com/docs/libraries/mcp-server/installation)
- [Configuration](https://trycua.com/docs/libraries/mcp-server/configuration)
- [Usage](https://trycua.com/docs/libraries/mcp-server/usage)
- [Tools](https://trycua.com/docs/libraries/mcp-server/tools)
- [Client Integrations](https://trycua.com/docs/libraries/mcp-server/client-integrations)
- [LLM Integrations](https://trycua.com/docs/libraries/mcp-server/llm-integrations)

661
libs/python/som/LICENSE Normal file
View File

@@ -0,0 +1,661 @@
GNU AFFERO GENERAL PUBLIC LICENSE
Version 3, 19 November 2007
Copyright (C) 2007 Free Software Foundation, Inc. <https://fsf.org/>
Everyone is permitted to copy and distribute verbatim copies
of this license document, but changing it is not allowed.
Preamble
The GNU Affero General Public License is a free, copyleft license for
software and other kinds of works, specifically designed to ensure
cooperation with the community in the case of network server software.
The licenses for most software and other practical works are designed
to take away your freedom to share and change the works. By contrast,
our General Public Licenses are intended to guarantee your freedom to
share and change all versions of a program--to make sure it remains free
software for all its users.
When we speak of free software, we are referring to freedom, not
price. Our General Public Licenses are designed to make sure that you
have the freedom to distribute copies of free software (and charge for
them if you wish), that you receive source code or can get it if you
want it, that you can change the software or use pieces of it in new
free programs, and that you know you can do these things.
Developers that use our General Public Licenses protect your rights
with two steps: (1) assert copyright on the software, and (2) offer
you this License which gives you legal permission to copy, distribute
and/or modify the software.
A secondary benefit of defending all users' freedom is that
improvements made in alternate versions of the program, if they
receive widespread use, become available for other developers to
incorporate. Many developers of free software are heartened and
encouraged by the resulting cooperation. However, in the case of
software used on network servers, this result may fail to come about.
The GNU General Public License permits making a modified version and
letting the public access it on a server without ever releasing its
source code to the public.
The GNU Affero General Public License is designed specifically to
ensure that, in such cases, the modified source code becomes available
to the community. It requires the operator of a network server to
provide the source code of the modified version running there to the
users of that server. Therefore, public use of a modified version, on
a publicly accessible server, gives the public access to the source
code of the modified version.
An older license, called the Affero General Public License and
published by Affero, was designed to accomplish similar goals. This is
a different license, not a version of the Affero GPL, but Affero has
released a new version of the Affero GPL which permits relicensing under
this license.
The precise terms and conditions for copying, distribution and
modification follow.
TERMS AND CONDITIONS
0. Definitions.
"This License" refers to version 3 of the GNU Affero General Public License.
"Copyright" also means copyright-like laws that apply to other kinds of
works, such as semiconductor masks.
"The Program" refers to any copyrightable work licensed under this
License. Each licensee is addressed as "you". "Licensees" and
"recipients" may be individuals or organizations.
To "modify" a work means to copy from or adapt all or part of the work
in a fashion requiring copyright permission, other than the making of an
exact copy. The resulting work is called a "modified version" of the
earlier work or a work "based on" the earlier work.
A "covered work" means either the unmodified Program or a work based
on the Program.
To "propagate" a work means to do anything with it that, without
permission, would make you directly or secondarily liable for
infringement under applicable copyright law, except executing it on a
computer or modifying a private copy. Propagation includes copying,
distribution (with or without modification), making available to the
public, and in some countries other activities as well.
To "convey" a work means any kind of propagation that enables other
parties to make or receive copies. Mere interaction with a user through
a computer network, with no transfer of a copy, is not conveying.
An interactive user interface displays "Appropriate Legal Notices"
to the extent that it includes a convenient and prominently visible
feature that (1) displays an appropriate copyright notice, and (2)
tells the user that there is no warranty for the work (except to the
extent that warranties are provided), that licensees may convey the
work under this License, and how to view a copy of this License. If
the interface presents a list of user commands or options, such as a
menu, a prominent item in the list meets this criterion.
1. Source Code.
The "source code" for a work means the preferred form of the work
for making modifications to it. "Object code" means any non-source
form of a work.
A "Standard Interface" means an interface that either is an official
standard defined by a recognized standards body, or, in the case of
interfaces specified for a particular programming language, one that
is widely used among developers working in that language.
The "System Libraries" of an executable work include anything, other
than the work as a whole, that (a) is included in the normal form of
packaging a Major Component, but which is not part of that Major
Component, and (b) serves only to enable use of the work with that
Major Component, or to implement a Standard Interface for which an
implementation is available to the public in source code form. A
"Major Component", in this context, means a major essential component
(kernel, window system, and so on) of the specific operating system
(if any) on which the executable work runs, or a compiler used to
produce the work, or an object code interpreter used to run it.
The "Corresponding Source" for a work in object code form means all
the source code needed to generate, install, and (for an executable
work) run the object code and to modify the work, including scripts to
control those activities. However, it does not include the work's
System Libraries, or general-purpose tools or generally available free
programs which are used unmodified in performing those activities but
which are not part of the work. For example, Corresponding Source
includes interface definition files associated with source files for
the work, and the source code for shared libraries and dynamically
linked subprograms that the work is specifically designed to require,
such as by intimate data communication or control flow between those
subprograms and other parts of the work.
The Corresponding Source need not include anything that users
can regenerate automatically from other parts of the Corresponding
Source.
The Corresponding Source for a work in source code form is that
same work.
2. Basic Permissions.
All rights granted under this License are granted for the term of
copyright on the Program, and are irrevocable provided the stated
conditions are met. This License explicitly affirms your unlimited
permission to run the unmodified Program. The output from running a
covered work is covered by this License only if the output, given its
content, constitutes a covered work. This License acknowledges your
rights of fair use or other equivalent, as provided by copyright law.
You may make, run and propagate covered works that you do not
convey, without conditions so long as your license otherwise remains
in force. You may convey covered works to others for the sole purpose
of having them make modifications exclusively for you, or provide you
with facilities for running those works, provided that you comply with
the terms of this License in conveying all material for which you do
not control copyright. Those thus making or running the covered works
for you must do so exclusively on your behalf, under your direction
and control, on terms that prohibit them from making any copies of
your copyrighted material outside their relationship with you.
Conveying under any other circumstances is permitted solely under
the conditions stated below. Sublicensing is not allowed; section 10
makes it unnecessary.
3. Protecting Users' Legal Rights From Anti-Circumvention Law.
No covered work shall be deemed part of an effective technological
measure under any applicable law fulfilling obligations under article
11 of the WIPO copyright treaty adopted on 20 December 1996, or
similar laws prohibiting or restricting circumvention of such
measures.
When you convey a covered work, you waive any legal power to forbid
circumvention of technological measures to the extent such circumvention
is effected by exercising rights under this License with respect to
the covered work, and you disclaim any intention to limit operation or
modification of the work as a means of enforcing, against the work's
users, your or third parties' legal rights to forbid circumvention of
technological measures.
4. Conveying Verbatim Copies.
You may convey verbatim copies of the Program's source code as you
receive it, in any medium, provided that you conspicuously and
appropriately publish on each copy an appropriate copyright notice;
keep intact all notices stating that this License and any
non-permissive terms added in accord with section 7 apply to the code;
keep intact all notices of the absence of any warranty; and give all
recipients a copy of this License along with the Program.
You may charge any price or no price for each copy that you convey,
and you may offer support or warranty protection for a fee.
5. Conveying Modified Source Versions.
You may convey a work based on the Program, or the modifications to
produce it from the Program, in the form of source code under the
terms of section 4, provided that you also meet all of these conditions:
a) The work must carry prominent notices stating that you modified
it, and giving a relevant date.
b) The work must carry prominent notices stating that it is
released under this License and any conditions added under section
7. This requirement modifies the requirement in section 4 to
"keep intact all notices".
c) You must license the entire work, as a whole, under this
License to anyone who comes into possession of a copy. This
License will therefore apply, along with any applicable section 7
additional terms, to the whole of the work, and all its parts,
regardless of how they are packaged. This License gives no
permission to license the work in any other way, but it does not
invalidate such permission if you have separately received it.
d) If the work has interactive user interfaces, each must display
Appropriate Legal Notices; however, if the Program has interactive
interfaces that do not display Appropriate Legal Notices, your
work need not make them do so.
A compilation of a covered work with other separate and independent
works, which are not by their nature extensions of the covered work,
and which are not combined with it such as to form a larger program,
in or on a volume of a storage or distribution medium, is called an
"aggregate" if the compilation and its resulting copyright are not
used to limit the access or legal rights of the compilation's users
beyond what the individual works permit. Inclusion of a covered work
in an aggregate does not cause this License to apply to the other
parts of the aggregate.
6. Conveying Non-Source Forms.
You may convey a covered work in object code form under the terms
of sections 4 and 5, provided that you also convey the
machine-readable Corresponding Source under the terms of this License,
in one of these ways:
a) Convey the object code in, or embodied in, a physical product
(including a physical distribution medium), accompanied by the
Corresponding Source fixed on a durable physical medium
customarily used for software interchange.
b) Convey the object code in, or embodied in, a physical product
(including a physical distribution medium), accompanied by a
written offer, valid for at least three years and valid for as
long as you offer spare parts or customer support for that product
model, to give anyone who possesses the object code either (1) a
copy of the Corresponding Source for all the software in the
product that is covered by this License, on a durable physical
medium customarily used for software interchange, for a price no
more than your reasonable cost of physically performing this
conveying of source, or (2) access to copy the
Corresponding Source from a network server at no charge.
c) Convey individual copies of the object code with a copy of the
written offer to provide the Corresponding Source. This
alternative is allowed only occasionally and noncommercially, and
only if you received the object code with such an offer, in accord
with subsection 6b.
d) Convey the object code by offering access from a designated
place (gratis or for a charge), and offer equivalent access to the
Corresponding Source in the same way through the same place at no
further charge. You need not require recipients to copy the
Corresponding Source along with the object code. If the place to
copy the object code is a network server, the Corresponding Source
may be on a different server (operated by you or a third party)
that supports equivalent copying facilities, provided you maintain
clear directions next to the object code saying where to find the
Corresponding Source. Regardless of what server hosts the
Corresponding Source, you remain obligated to ensure that it is
available for as long as needed to satisfy these requirements.
e) Convey the object code using peer-to-peer transmission, provided
you inform other peers where the object code and Corresponding
Source of the work are being offered to the general public at no
charge under subsection 6d.
A separable portion of the object code, whose source code is excluded
from the Corresponding Source as a System Library, need not be
included in conveying the object code work.
A "User Product" is either (1) a "consumer product", which means any
tangible personal property which is normally used for personal, family,
or household purposes, or (2) anything designed or sold for incorporation
into a dwelling. In determining whether a product is a consumer product,
doubtful cases shall be resolved in favor of coverage. For a particular
product received by a particular user, "normally used" refers to a
typical or common use of that class of product, regardless of the status
of the particular user or of the way in which the particular user
actually uses, or expects or is expected to use, the product. A product
is a consumer product regardless of whether the product has substantial
commercial, industrial or non-consumer uses, unless such uses represent
the only significant mode of use of the product.
"Installation Information" for a User Product means any methods,
procedures, authorization keys, or other information required to install
and execute modified versions of a covered work in that User Product from
a modified version of its Corresponding Source. The information must
suffice to ensure that the continued functioning of the modified object
code is in no case prevented or interfered with solely because
modification has been made.
If you convey an object code work under this section in, or with, or
specifically for use in, a User Product, and the conveying occurs as
part of a transaction in which the right of possession and use of the
User Product is transferred to the recipient in perpetuity or for a
fixed term (regardless of how the transaction is characterized), the
Corresponding Source conveyed under this section must be accompanied
by the Installation Information. But this requirement does not apply
if neither you nor any third party retains the ability to install
modified object code on the User Product (for example, the work has
been installed in ROM).
The requirement to provide Installation Information does not include a
requirement to continue to provide support service, warranty, or updates
for a work that has been modified or installed by the recipient, or for
the User Product in which it has been modified or installed. Access to a
network may be denied when the modification itself materially and
adversely affects the operation of the network or violates the rules and
protocols for communication across the network.
Corresponding Source conveyed, and Installation Information provided,
in accord with this section must be in a format that is publicly
documented (and with an implementation available to the public in
source code form), and must require no special password or key for
unpacking, reading or copying.
7. Additional Terms.
"Additional permissions" are terms that supplement the terms of this
License by making exceptions from one or more of its conditions.
Additional permissions that are applicable to the entire Program shall
be treated as though they were included in this License, to the extent
that they are valid under applicable law. If additional permissions
apply only to part of the Program, that part may be used separately
under those permissions, but the entire Program remains governed by
this License without regard to the additional permissions.
When you convey a copy of a covered work, you may at your option
remove any additional permissions from that copy, or from any part of
it. (Additional permissions may be written to require their own
removal in certain cases when you modify the work.) You may place
additional permissions on material, added by you to a covered work,
for which you have or can give appropriate copyright permission.
Notwithstanding any other provision of this License, for material you
add to a covered work, you may (if authorized by the copyright holders of
that material) supplement the terms of this License with terms:
a) Disclaiming warranty or limiting liability differently from the
terms of sections 15 and 16 of this License; or
b) Requiring preservation of specified reasonable legal notices or
author attributions in that material or in the Appropriate Legal
Notices displayed by works containing it; or
c) Prohibiting misrepresentation of the origin of that material, or
requiring that modified versions of such material be marked in
reasonable ways as different from the original version; or
d) Limiting the use for publicity purposes of names of licensors or
authors of the material; or
e) Declining to grant rights under trademark law for use of some
trade names, trademarks, or service marks; or
f) Requiring indemnification of licensors and authors of that
material by anyone who conveys the material (or modified versions of
it) with contractual assumptions of liability to the recipient, for
any liability that these contractual assumptions directly impose on
those licensors and authors.
All other non-permissive additional terms are considered "further
restrictions" within the meaning of section 10. If the Program as you
received it, or any part of it, contains a notice stating that it is
governed by this License along with a term that is a further
restriction, you may remove that term. If a license document contains
a further restriction but permits relicensing or conveying under this
License, you may add to a covered work material governed by the terms
of that license document, provided that the further restriction does
not survive such relicensing or conveying.
If you add terms to a covered work in accord with this section, you
must place, in the relevant source files, a statement of the
additional terms that apply to those files, or a notice indicating
where to find the applicable terms.
Additional terms, permissive or non-permissive, may be stated in the
form of a separately written license, or stated as exceptions;
the above requirements apply either way.
8. Termination.
You may not propagate or modify a covered work except as expressly
provided under this License. Any attempt otherwise to propagate or
modify it is void, and will automatically terminate your rights under
this License (including any patent licenses granted under the third
paragraph of section 11).
However, if you cease all violation of this License, then your
license from a particular copyright holder is reinstated (a)
provisionally, unless and until the copyright holder explicitly and
finally terminates your license, and (b) permanently, if the copyright
holder fails to notify you of the violation by some reasonable means
prior to 60 days after the cessation.
Moreover, your license from a particular copyright holder is
reinstated permanently if the copyright holder notifies you of the
violation by some reasonable means, this is the first time you have
received notice of violation of this License (for any work) from that
copyright holder, and you cure the violation prior to 30 days after
your receipt of the notice.
Termination of your rights under this section does not terminate the
licenses of parties who have received copies or rights from you under
this License. If your rights have been terminated and not permanently
reinstated, you do not qualify to receive new licenses for the same
material under section 10.
9. Acceptance Not Required for Having Copies.
You are not required to accept this License in order to receive or
run a copy of the Program. Ancillary propagation of a covered work
occurring solely as a consequence of using peer-to-peer transmission
to receive a copy likewise does not require acceptance. However,
nothing other than this License grants you permission to propagate or
modify any covered work. These actions infringe copyright if you do
not accept this License. Therefore, by modifying or propagating a
covered work, you indicate your acceptance of this License to do so.
10. Automatic Licensing of Downstream Recipients.
Each time you convey a covered work, the recipient automatically
receives a license from the original licensors, to run, modify and
propagate that work, subject to this License. You are not responsible
for enforcing compliance by third parties with this License.
An "entity transaction" is a transaction transferring control of an
organization, or substantially all assets of one, or subdividing an
organization, or merging organizations. If propagation of a covered
work results from an entity transaction, each party to that
transaction who receives a copy of the work also receives whatever
licenses to the work the party's predecessor in interest had or could
give under the previous paragraph, plus a right to possession of the
Corresponding Source of the work from the predecessor in interest, if
the predecessor has it or can get it with reasonable efforts.
You may not impose any further restrictions on the exercise of the
rights granted or affirmed under this License. For example, you may
not impose a license fee, royalty, or other charge for exercise of
rights granted under this License, and you may not initiate litigation
(including a cross-claim or counterclaim in a lawsuit) alleging that
any patent claim is infringed by making, using, selling, offering for
sale, or importing the Program or any portion of it.
11. Patents.
A "contributor" is a copyright holder who authorizes use under this
License of the Program or a work on which the Program is based. The
work thus licensed is called the contributor's "contributor version".
A contributor's "essential patent claims" are all patent claims
owned or controlled by the contributor, whether already acquired or
hereafter acquired, that would be infringed by some manner, permitted
by this License, of making, using, or selling its contributor version,
but do not include claims that would be infringed only as a
consequence of further modification of the contributor version. For
purposes of this definition, "control" includes the right to grant
patent sublicenses in a manner consistent with the requirements of
this License.
Each contributor grants you a non-exclusive, worldwide, royalty-free
patent license under the contributor's essential patent claims, to
make, use, sell, offer for sale, import and otherwise run, modify and
propagate the contents of its contributor version.
In the following three paragraphs, a "patent license" is any express
agreement or commitment, however denominated, not to enforce a patent
(such as an express permission to practice a patent or covenant not to
sue for patent infringement). To "grant" such a patent license to a
party means to make such an agreement or commitment not to enforce a
patent against the party.
If you convey a covered work, knowingly relying on a patent license,
and the Corresponding Source of the work is not available for anyone
to copy, free of charge and under the terms of this License, through a
publicly available network server or other readily accessible means,
then you must either (1) cause the Corresponding Source to be so
available, or (2) arrange to deprive yourself of the benefit of the
patent license for this particular work, or (3) arrange, in a manner
consistent with the requirements of this License, to extend the patent
license to downstream recipients. "Knowingly relying" means you have
actual knowledge that, but for the patent license, your conveying the
covered work in a country, or your recipient's use of the covered work
in a country, would infringe one or more identifiable patents in that
country that you have reason to believe are valid.
If, pursuant to or in connection with a single transaction or
arrangement, you convey, or propagate by procuring conveyance of, a
covered work, and grant a patent license to some of the parties
receiving the covered work authorizing them to use, propagate, modify
or convey a specific copy of the covered work, then the patent license
you grant is automatically extended to all recipients of the covered
work and works based on it.
A patent license is "discriminatory" if it does not include within
the scope of its coverage, prohibits the exercise of, or is
conditioned on the non-exercise of one or more of the rights that are
specifically granted under this License. You may not convey a covered
work if you are a party to an arrangement with a third party that is
in the business of distributing software, under which you make payment
to the third party based on the extent of your activity of conveying
the work, and under which the third party grants, to any of the
parties who would receive the covered work from you, a discriminatory
patent license (a) in connection with copies of the covered work
conveyed by you (or copies made from those copies), or (b) primarily
for and in connection with specific products or compilations that
contain the covered work, unless you entered into that arrangement,
or that patent license was granted, prior to 28 March 2007.
Nothing in this License shall be construed as excluding or limiting
any implied license or other defenses to infringement that may
otherwise be available to you under applicable patent law.
12. No Surrender of Others' Freedom.
If conditions are imposed on you (whether by court order, agreement or
otherwise) that contradict the conditions of this License, they do not
excuse you from the conditions of this License. If you cannot convey a
covered work so as to satisfy simultaneously your obligations under this
License and any other pertinent obligations, then as a consequence you may
not convey it at all. For example, if you agree to terms that obligate you
to collect a royalty for further conveying from those to whom you convey
the Program, the only way you could satisfy both those terms and this
License would be to refrain entirely from conveying the Program.
13. Remote Network Interaction; Use with the GNU General Public License.
Notwithstanding any other provision of this License, if you modify the
Program, your modified version must prominently offer all users
interacting with it remotely through a computer network (if your version
supports such interaction) an opportunity to receive the Corresponding
Source of your version by providing access to the Corresponding Source
from a network server at no charge, through some standard or customary
means of facilitating copying of software. This Corresponding Source
shall include the Corresponding Source for any work covered by version 3
of the GNU General Public License that is incorporated pursuant to the
following paragraph.
Notwithstanding any other provision of this License, you have
permission to link or combine any covered work with a work licensed
under version 3 of the GNU General Public License into a single
combined work, and to convey the resulting work. The terms of this
License will continue to apply to the part which is the covered work,
but the work with which it is combined will remain governed by version
3 of the GNU General Public License.
14. Revised Versions of this License.
The Free Software Foundation may publish revised and/or new versions of
the GNU Affero General Public License from time to time. Such new versions
will be similar in spirit to the present version, but may differ in detail to
address new problems or concerns.
Each version is given a distinguishing version number. If the
Program specifies that a certain numbered version of the GNU Affero General
Public License "or any later version" applies to it, you have the
option of following the terms and conditions either of that numbered
version or of any later version published by the Free Software
Foundation. If the Program does not specify a version number of the
GNU Affero General Public License, you may choose any version ever published
by the Free Software Foundation.
If the Program specifies that a proxy can decide which future
versions of the GNU Affero General Public License can be used, that proxy's
public statement of acceptance of a version permanently authorizes you
to choose that version for the Program.
Later license versions may give you additional or different
permissions. However, no additional obligations are imposed on any
author or copyright holder as a result of your choosing to follow a
later version.
15. Disclaimer of Warranty.
THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
16. Limitation of Liability.
IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
SUCH DAMAGES.
17. Interpretation of Sections 15 and 16.
If the disclaimer of warranty and limitation of liability provided
above cannot be given local legal effect according to their terms,
reviewing courts shall apply local law that most closely approximates
an absolute waiver of all civil liability in connection with the
Program, unless a warranty or assumption of liability accompanies a
copy of the Program in return for a fee.
END OF TERMS AND CONDITIONS
How to Apply These Terms to Your New Programs
If you develop a new program, and you want it to be of the greatest
possible use to the public, the best way to achieve this is to make it
free software which everyone can redistribute and change under these terms.
To do so, attach the following notices to the program. It is safest
to attach them to the start of each source file to most effectively
state the exclusion of warranty; and each file should have at least
the "copyright" line and a pointer to where the full notice is found.
<one line to give the program's name and a brief idea of what it does.>
Copyright (C) <year> <name of author>
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with this program. If not, see <https://www.gnu.org/licenses/>.
Also add information on how to contact you by electronic and paper mail.
If your software can interact with users remotely through a computer
network, you should also make sure that it provides a way for users to
get its source. For example, if your program is a web application, its
interface could display a "Source" link that leads users to an archive
of the code. There are many ways you could offer source, and different
solutions will be better for different programs; see section 13 for the
specific requirements.
You should also get your employer (if you work as a programmer) or school,
if any, to sign a "copyright disclaimer" for the program, if necessary.
For more information on this, and how to apply and follow the GNU AGPL, see
<https://www.gnu.org/licenses/>.

88
libs/python/som/README.md
View File

@@ -75,93 +75,9 @@ for elem in result.elements:
print(f"Text: '{elem.content}', confidence={elem.confidence:.3f}")
```
## Configuration
## Docs
### Detection Parameters
#### Box Threshold (0.3)
Controls the confidence threshold for accepting detections:
```
High Threshold (0.3): Low Threshold (0.01):
+----------------+ +----------------+
| | | +--------+ |
| Confident | | |Unsure?| |
| Detection | | +--------+ |
| (✓ Accept) | | (? Reject) |
| | | |
+----------------+ +----------------+
conf = 0.85 conf = 0.02
```
- Higher values (0.3) yield more precise but fewer detections
- Lower values (0.01) catch more potential icons but increase false positives
- Default is 0.3 for optimal precision/recall balance
#### IOU Threshold (0.1)
Controls how overlapping detections are merged:
```
IOU = Intersection Area / Union Area
Low Overlap (Keep Both): High Overlap (Merge):
+----------+ +----------+
| Box1 | | Box1 |
| | vs. |+-----+ |
+----------+ ||Box2 | |
+----------+ |+-----+ |
| Box2 | +----------+
| |
+----------+
IOU ≈ 0.05 (Keep Both) IOU ≈ 0.7 (Merge)
```
- Lower values (0.1) more aggressively remove overlapping boxes
- Higher values (0.5) allow more overlapping detections
- Default is 0.1 to handle densely packed UI elements
### OCR Configuration
- **Engine**: EasyOCR
- Primary choice for all platforms
- Fast initialization and processing
- Built-in English language support
- GPU acceleration when available
- **Settings**:
- Timeout: 5 seconds
- Confidence threshold: 0.5
- Paragraph mode: Disabled
- Language: English only
## Performance
### Hardware Acceleration
#### MPS (Metal Performance Shaders)
- Multi-scale detection (640px, 1280px, 1920px)
- Test-time augmentation enabled
- Half-precision (FP16)
- Average detection time: ~0.4s
- Best for production use when available
#### CPU
- Single-scale detection (1280px)
- Full-precision (FP32)
- Average detection time: ~1.3s
- Reliable fallback option
### Example Output Structure
```
examples/output/
├── {timestamp}_no_ocr/
│ ├── annotated_images/
│ │ └── screenshot_analyzed.png
│ ├── screen_details.txt
│ └── summary.json
└── {timestamp}_ocr/
├── annotated_images/
│ └── screenshot_analyzed.png
├── screen_details.txt
└── summary.json
```
- [Configuration](http://localhost:8090/docs/libraries/som/configuration)
## Development

4
libs/python/som/pyproject.toml
View File

@@ -26,12 +26,12 @@ dependencies = [
]
requires-python = ">=3.11"
readme = "README.md"
license = {text = "MIT"}
license = {text = "AGPL-3.0-or-later"}
keywords = ["computer-vision", "ocr", "ui-analysis", "icon-detection"]
classifiers = [
"Development Status :: 4 - Beta",
"License :: OSI Approved :: GNU Affero General Public License v3 or later (AGPLv3+)",
"Intended Audience :: Developers",
"License :: OSI Approved :: MIT License",
"Programming Language :: Python :: 3",
"Programming Language :: Python :: 3.11",
"Topic :: Scientific/Engineering :: Artificial Intelligence",