Commit Graph

578 Commits

Author SHA1 Message Date
github-actions[bot]
dd9f13349b Bump cua-agent to v0.7.3 2025-12-31 13:43:28 +00:00
Harsh Verma
1557c2a9ea Merge pull request #684 from trycua/fix/fara-impl
[Agent] Fix FARA 7B loop implementation
2025-12-31 19:10:49 +05:30
synacktra.work@gmail.com
a3aca69e27 fix: return value if key exists regardless of its truthiness 2025-12-31 18:57:43 +05:30
synacktra.work@gmail.com
5001fa7821 chore: lint & format 2025-12-29 17:46:38 +05:30
synacktra.work@gmail.com
31568dc334 fix(agent): detect terminate in function_call_output to stop agent loop
After the terminate tool is executed, the agent loop adds function_call_output
to new_items, making it the last item (without role field), so the loop continues.

Solution: Check at the start of predict_step if the last message is a
function_call_output with "terminated": True, and immediately return a
final assistant message without calling the model.
2025-12-29 15:06:37 +05:30
synacktra.work@gmail.com
c4297ae441 fix(agent): always add final assistant message when terminate is detected 2025-12-29 15:02:43 +05:30
synacktra.work@gmail.com
c5a85e866c fix(agent): correct fara loop's terminate action detection for computer_call vs function_call 2025-12-29 15:00:10 +05:30
synacktra.work@gmail.com
3bbb8a7e71 fix(agent): use computer's interface for direct interaction 2025-12-29 12:59:50 +05:30
synacktra.work@gmail.com
e94b0e850e refactor(agent): fix fara implementation
- Use original Microsoft FARA prompt templates from qwen_helpers
- Fix tool call parsing to handle nested JSON braces
- Extract and preserve assistant thoughts before tool calls
- Add termination detection to exit agent loop properly
2025-12-29 12:57:52 +05:30
Adam
4fd6c9908c formatting fix 2025-12-28 21:39:09 -05:00
Adam
f36e335c29 formatting fix 2025-12-28 21:31:38 -05:00
Adam
545c2180f8 pyautogui cleanup 2025-12-28 21:25:49 -05:00
Adam
76c0ca4302 Merge branch 'main' into feat/pynput-server 2025-12-28 21:06:47 -05:00
Adam
5631c69af9 Revert "refractor docs into 6 sections" 2025-12-24 14:19:24 -05:00
Adam
13f3fb800d refractor docs into 6 sections
Agent - /agent
Computer - /computer
Lume - /lume
Lumier - /lumier
Set-of-Mark - /som
MCP - /mcp
2025-12-24 14:06:43 -05:00
github-actions[bot]
b0390ec648 Bump cua-agent to v0.7.2 2025-12-23 09:30:00 +00:00
github-actions[bot]
5bdf17b92d Bump cua-computer to v0.5.1 2025-12-23 09:30:00 +00:00
synacktra.work@gmail.com
e0ed712ea3 Merge branch 'main' of github.com:trycua/cua into feat/android-docker-support-&-docs 2025-12-23 01:36:43 +05:30
synacktra.work@gmail.com
13ffe4ace2 fix: run formatters to fix lint workflow failure 2025-12-23 01:20:52 +05:30
synacktra.work@gmail.com
7114e6bef2 chore: move android to qemu-docker 2025-12-22 18:21:41 +05:30
synacktra.work@gmail.com
c0ad2cb233 rename "cua-droid" to "cua-android-docker" 2025-12-22 17:23:44 +05:30
github-actions[bot]
6e697042bb Bump cua-computer-server to v0.3.0 2025-12-20 21:56:29 +00:00
synacktra.work@gmail.com
023e797936 feat(docker): add Android emulator image support
- Detect android images (checks for "droid" in image name)
- Require /dev/kvm device for Android emulator
- Configure Android-specific ports (VNC: 6080, API: 8000)
- Set home directory to /home/androidusr
- Add default env vars: EMULATOR_DEVICE and WEB_VNC
2025-12-21 03:18:32 +05:30
synacktra.work@gmail.com
20fe640127 feat: add android to OS type literal 2025-12-21 03:16:04 +05:30
synacktra.work@gmail.com
948265f62d Merge branch 'main' of github.com:trycua/cua into feat/android-docker 2025-12-21 01:36:51 +05:30
github-actions[bot]
0bbbf545bd Bump cua-agent to v0.7.1 2025-12-20 19:26:11 +00:00
synacktra.work@gmail.com
eef03294e4 fix(agent): update cua-computer dependency to >=0.5.0,<0.6.0 2025-12-21 00:53:26 +05:30
github-actions[bot]
37ce59de07 Bump cua-agent to v0.7.0 2025-12-20 19:02:57 +00:00
github-actions[bot]
793aaacc94 Bump cua-computer to v0.5.0 2025-12-20 19:02:56 +00:00
Francesco Bonacci
c96d2d97dd Merge pull request #670 from trycua/feat/qemu-docker-support-&-docs
Add QEMU Docker support to Python SDK
2025-12-20 10:53:32 -08:00
Sarina Li
175685e1bf allow for v1 route prefix 2025-12-19 14:33:10 -05:00
Sarina Li
05b1e5a984 add cua adapter support for gemini 2025-12-19 12:34:16 -05:00
synacktra.work@gmail.com
0c77bf627f chore(qemu): rename "cua-windows" -> "cua-qemu-windows" 2025-12-19 18:12:14 +05:30
synacktra.work@gmail.com
2aeafa69b2 chore(qemu): rename "cua-linux" -> "cua-qemu-linx" 2025-12-19 18:07:18 +05:30
synacktra.work@gmail.com
eb04b5b023 feat(docker): add QEMU image support
- Detect QEMU images (cua-linux, cua-windows) and configure accordingly
- Validate golden image storage path for QEMU images
- Add /dev/kvm device support with warning if not provided
- Add NET_ADMIN capability for QEMU networking
- Map correct ports (8006 VNC, 5000 API) for QEMU vs Kasm/XFCE
- Support custom environment variables via run_opts (RAM_SIZE, CPU_CORES, DISK_SIZE)
- Skip Docker memory/CPU limits for QEMU (uses env vars instead)
2025-12-19 14:23:14 +05:30
synacktra.work@gmail.com
7b7c75293f feat(computer): add timeout and run_opts parameters
- Add configurable timeout parameter (default 100s) for interface connection
- Add run_opts parameter for provider-specific options (devices, env vars)
- Include memory/CPU from config in run_opts
- Use configurable timeout instead of hardcoded 30s in wait_for_ready calls"
2025-12-19 14:21:00 +05:30
github-actions[bot]
eb89fd937f Bump cua-agent to v0.6.2 2025-12-18 22:20:49 +00:00
ddupont
60d88817a2 Merge pull request #668 from sarinali/feat/gemini-3-flash-native-integration
Gemini 3 Pro Preview and Flash Integration Native
2025-12-18 17:20:18 -05:00
ddupont
5e94349bb0 Modify API key retrieval method
Updated API key retrieval to allow for kwargs input.
2025-12-18 17:19:54 -05:00
Sarina Li
8e044585ea fix minor docs stuff 2025-12-18 12:39:15 -05:00
github-actions[bot]
d8e0fdd3bd Bump cua-agent to v0.6.1 2025-12-18 03:58:40 +00:00
github-actions[bot]
3e38d17dc1 Bump cua-computer to v0.4.19 2025-12-18 03:58:40 +00:00
Dillon DuPont
23d3a8b3e0 Add api_host kwarg 2025-12-17 22:56:46 -05:00
Sarina Li
3b947c5c13 merge main 2025-12-17 21:41:54 -05:00
Sarina Li
d85fc10adb add custom tools instead and support gemini 3 pro and flash 2025-12-17 21:41:15 -05:00
Sarina Li
ff7f88e1a6 add back and forth tools in browser 2025-12-17 21:41:15 -05:00
Sarina Li
fda7c65ccd expose browser tool execution in cua 2025-12-17 21:41:15 -05:00
Sarina Li
3dff086749 add back and forth tools in browser 2025-12-17 19:00:59 -05:00
Sarina Li
019bdc5e5d expose browser tool execution in cua 2025-12-17 19:00:38 -05:00
Sarina Li
c8c2996bc7 gemini loop with serialization checking 2025-12-17 18:53:47 -05:00