f7147ab312
dev: generalize how OpenAIUtil handles usage
...
The current implementation would break for XAIService because it gets
cumulative usage reports as the stream goes on (whereas OpenAI and
DeepSeek only provide this at the end of the stream).
2025-02-06 17:34:57 -05:00
16e2729b1d
dev: migrate XAIService to normalized messages
2025-02-06 17:34:57 -05:00
ce0ab02f39
dev: add file output mode for convert
2025-02-06 17:34:57 -05:00
5e30958a14
tweak: enable convert module on selfhosted run
2025-02-06 17:34:57 -05:00
2b0e173975
dev: add convert module (and ConvertAPI)
2025-02-06 17:34:57 -05:00
12806a2395
only close TLS socket once
2025-02-06 17:13:04 -05:00
26666fe666
new API puter.net.generateWispV1URL()
2025-02-05 16:56:57 -05:00
c8bed6eaee
dev: add deepseek tool calls
2025-02-05 15:32:35 -05:00
d4eff82d3c
dev: add email attachments
2025-02-05 14:12:05 -05:00
da4a1a182b
dev: de-couple OpenAI SDK utilities from service
...
This change will be helpful for making behavior across OpenAI-compatible
APIs consistent.
2025-02-05 11:36:07 -05:00
50383aab59
dev: try this for now
2025-02-04 16:45:42 -05:00
e6fc773706
fix: unfixed fix from earlier
2025-02-04 16:36:06 -05:00
9dbba292c4
dev: coerce semantics of tool calls/use in messages
...
If these APIs were standardized this would have been so much less
painful, but alas, tool calls are implemented completely differently
between Claude and OpenAI. Good job guys.
2025-02-04 16:15:47 -05:00
484bb5c201
fix: parser error for empty json buffer
2025-02-04 15:42:03 -05:00
d2358d234b
fix: fix name and id for openai tool calls
2025-02-04 14:36:30 -05:00
29c0241119
fix: let messages with tool_calls have content=null
2025-02-04 14:21:14 -05:00
8f27742038
fix: repair stream end
2025-02-04 14:09:28 -05:00
e2797c38d0
fix: add type=text
2025-02-04 14:02:01 -05:00
747614c28b
dev: integrate stream normalization for OpenAI
2025-02-04 13:57:10 -05:00
437467e0ce
dev: integrate stream normalization for Claude
2025-02-04 13:10:14 -05:00
0ce8fdc9f6
Merge pull request #1037 from HeyPuter/Add-TRADEMARK.md
...
Create TRADEMARK.md
2025-02-03 14:23:24 -08:00
4e17908664
Update TRADEMARK.md
2025-02-03 14:22:01 -08:00
f54b7b7b3d
dev: PoC for stream normalization
2025-02-03 16:57:13 -05:00
f0434435c4
dev: option to normalize AI response (non-streaming)
2025-02-03 14:03:27 -05:00
57231d31e4
dev: add email module to run-selfhosted
2025-02-03 13:14:52 -05:00
cfa3fde436
dev: error handling and text encoding
2025-02-03 13:11:29 -05:00
55d052cfc2
fix: various issues with Mail module
2025-02-03 12:25:22 -05:00
e281dc92e3
dev: add mail module
2025-02-03 10:08:39 -05:00
dc90999fee
dev: enable gpt-o3-mini
2025-01-31 14:48:49 -05:00
d88dedb66c
dev: normalize Claude input
...
- add merging of same-role messages for all normalized inputs
- generalize Claude-specific system message extraction
- remove redundant behavior from Claude
2025-01-31 11:22:06 -05:00
40aa52225f
dev: make ai fallback configurable
2025-01-31 10:32:36 -05:00
e257dceb93
Merge pull request #1100 from Zac0511/patch-4
...
Update fr.js
2025-01-31 05:04:24 -08:00
6c2fa4d398
Update fr.js
...
Fixed some translations and deleted a copy of billing translations
2025-01-31 09:50:49 +01:00
2d12949a08
dev: improve user request error detection
2025-01-30 15:40:10 -05:00
60eef2fc67
fix: buffer incomplete JSON objects from AI stream
...
- simplify ndjson stream to get complete lines instead of chunks
- add buffering for incomplete lines
2025-01-30 15:21:05 -05:00
af3d732fc4
dev: async await for puter-exec
2025-01-30 13:37:15 -05:00
d0eca6dafb
Revert "dev: remove redundant message adaption from claude"
...
This reverts commit 0c42613ff0
2025-01-30 12:23:27 -05:00
760eb55b69
Revert "fix: mistake in 0c42613"
...
This reverts commit 8ffd0004b3
2025-01-30 12:23:06 -05:00
8ffd0004b3
fix: mistake in 0c42613
2025-01-30 12:19:07 -05:00
05ced4e625
Revert "dev: remove max_tokens from Claude"
...
This reverts commit 3bce60893e
2025-01-30 12:18:26 -05:00
3bce60893e
dev: remove max_tokens from Claude
2025-01-30 12:09:35 -05:00
0c42613ff0
dev: remove redundant message adaption from claude
2025-01-30 10:11:52 -05:00
babe9f8ede
dev: remove now-redundant token count logic from openai service
2025-01-29 16:58:30 -05:00
7151723894
dev: use message normalization and clean up openai service
2025-01-29 16:45:15 -05:00
0009ededfa
dev: implement generic extract_text for ai messages
2025-01-29 16:08:30 -05:00
e9279ffb36
dev: implement message normalization
2025-01-29 15:28:14 -05:00
e6ad54be19
dev: puterjs exec module (2)
2025-01-29 13:58:21 -05:00
cbe02a4489
dev: puterjs exec module
2025-01-29 13:57:50 -05:00
77d6bb2d8c
dev: always select highest version of the specified language
2025-01-29 10:31:58 -05:00
dcd47bc4cf
fix: fix microcents to USD conversion in AIChatService
...
This only affects a value sent to TrackSpendingService. Cost
calculations related to user's quotas / billing remain unaffected.
2025-01-28 16:45:48 -05:00
KernelDeimos