Commit History
-
Use provider=auto instead of hardcoding novita
Novita specifically may not have spend authorized on this account; auto lets huggingface_hub pick the first provider actually available/authorized for the user rather than failing on one specific provider.fec35e0 -
Fix broken tools surfaced by the 20% GAIA run's trace logs
- get_youtube_transcript called the old YouTubeTranscriptApi.get_transcript() classmethod, which no longer exists in the installed youtube-transcript-api version (it now requires YouTubeTranscriptApi().fetch()). This silently broke every YouTube-transcript question. - VisitWebpageTool sends no User-Agent header, so Wikipedia (and other sites) 403 it -- this hit at least 4 of the 20 questions in the last run. Added fetch_webpage, a drop-in replacement that sends a descriptive UA, and made it the agent's default over visit_webpage. - Fetching an arXiv PDF through visit_webpage crashed with 'maximum recursion depth exceeded' trying to markdownify binary PDF content. Added read_pdf_from_url to download + parse PDFs by URL with pypdf. - Added html5lib so pandas.read_html can parse HTML tables directly, and pointed the agent at it -- the trace shows the model repeatedly writing fragile regex against raw table markdown and failing/timing out on max_steps instead.a10a8ac -
Fix task_id/read_pdf bugs, upgrade model, tighten answer formatting for GAIA scoring
- agent.run() was called with the raw question, never full_prompt, so the agent never saw task_id and could not call fetch_task_file for any GAIA question with an attached file. - read_pdf was instantiated but commented out of the tools list, so the agent had no way to actually read PDFs despite instructions telling it to. - Swap Qwen2.5-Coder-32B-Instruct for Qwen3-235B-A22B-Instruct-2507 (via novita) with temperature=0.3 for stronger general reasoning alongside code generation. - Add explicit answer-formatting guidance matching GAIA's quasi-exact match grader (bare numbers, minimal strings, comma-separated lists).36cc9c4 -
Update app.py
Verified signatureb3c29a7 -
Update app.py
Verified signaturec2f26e7 -
Update requirements.txt
Verified signature09afd7f -
Update requirements.txt
Verified signature278268c -
Update app.py
Verified signature532c5d2 -
Update app.py
Verified signaturea7ec422 -
Update app.py
Verified signature69befd0 -
Update app.py
Verified signature918a8b2 -
Update app.py
Verified signaturedcedcc3 -
Update app.py
Verified signature364c958 -
Update app.py
Verified signature8721e6b -
Update app.py
Verified signature2f72188 -
Update app.py
Verified signature1b77204 -
Update app.py
Verified signaturef5be765 -
Update requirements.txt
Verified signaturea5ac7bb -
Update app.py
Verified signature4e53f08 -
Update app.py
Verified signature73668ae -
Update app.py
Verified signature5106442 -
Update app.py
Verified signature85904f7 -
Update app.py
Verified signature9121c5a -
Update app.py
Verified signaturec989430 -
Update app.py
Verified signature489fea7 -
Update app.py
Verified signature100e2a5 -
Update app.py
Verified signature7f13a52 -
Update app.py
Verified signature484e478 -
Update app.py
Verified signature38f2d88 -
Update app.py
Verified signatureadf86e2
-
Update requirements.txt
Verified signatureb23c295 -
Update app.py
Verified signaturee8fc12a -
Update app.py
Verified signature2c1e9ac -
Update app.py
Verified signature4b87506 -
Update app.py
Verified signature403daa3 -
Update app.py
Verified signaturecda85c9 -
Update app.py
Verified signaturef70d238 -
Update app.py
Verified signatureea16981
-
Update app.py
Verified signature127ed31 -
Update app.py
Verified signaturea3ca106 -
Update app.py
Verified signature2f79205 -
Update app.py
Verified signaturebe7ba79 -
Update app.py
Verified signaturee2f8cfb -
Update app.py
Verified signature408669a -
Update app.py
Verified signature2ce9a12 -
Update app.py
Verified signaturee809807 -
Update app.py
Verified signaturef0cfd31 -
Update app.py
Verified signature1e96b44 -
Update app.py
Verified signaturedb8845f -
Update app.py
Verified signature1d9ad59