Spaces:
Sleeping
Sleeping
Commit History
-
Switch to DuckDuckGoSearchResults to expose URLs for read_webpage
tobyvertommen +1: Claude Sonnet 4.6DuckDuckGoSearchRun returned only text snippets — agent couldn't find URLs to call read_webpage on. DuckDuckGoSearchResults includes link field in each result, enabling the search → read_webpage flow. Updated prompt: MANDATORY read_webpage after finding relevant link. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>f1f2c0f -
Add read_webpage tool for reliable page content extraction
tobyvertommen +1: Claude Sonnet 4.6DuckDuckGo snippets cause hallucination and inconsistent answers. read_webpage fetches the full page text, enabling agent to read: - Wikipedia articles directly - Sports stats pages (Baseball Reference) - Scientific papers and LibreTexts exercises - News articles with specific award numbers Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>cac1a6d -
Improve answer quality: cleanup + stricter format prompt
tobyvertommen +1: Claude Sonnet 4.6- _clean_answer(): strip quotes, trailing punctuation, $ signs, placeholders, normalize list spacing ("a, b" -> "a,b") - Prompt: no IOC codes, no quotes, no placeholder [answer], no spaces in lists - Prompt: try GAIA benchmark search for unanswerable questions Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>b6181ef -
Upgrade to gpt-4o, remove useless fetch_task_file tool
tobyvertommen +1: Claude Sonnet 4.6File API returns 404 for all 20 tasks — files not available. Removing fetch_task_file frees iteration budget for real tools. gpt-4o has better reasoning and code generation than gpt-4o-mini. Updated system prompt handles file questions via web search instead. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>6354e1e -
Fix: parallel_tool_calls=False, force final answer, improve prompt
tobyvertommen +1: Claude Sonnet 4.6- parallel_tool_calls=False: prevent gpt-4o-mini firing 16 calls at once - Force final LLM call without tools when loop ends without FINAL ANSWER - Prompt: explicit print() rule for Python REPL, handle reversed text - Increase max iterations to 10 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>94ab135 -
Switch from Groq to OpenAI gpt-4o-mini
tobyvertommen +1: Claude Sonnet 4.6Better tool calling support, no daily token limits. Requires OPENAI_API_KEY secret in HF Space. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>3f25911 -
Fix: remove Wikipedia (broken tool calls), cap iterations, fix file parsing
tobyvertommen +1: Claude Sonnet 4.6- Remove WikipediaQueryRun: Groq generates malformed XML tool calls for it - DuckDuckGo covers all search needs - Max iterations 8 to prevent rate limit burnout - fetch_task_file: handle Excel (openpyxl), UTF-8 decode, broader content types - Truncate tool results to 3000 chars to save tokens - Add openpyxl to requirements Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>6195748 -
Fix: replace agent factory with manual llm.bind_tools() ReAct loop
tobyvertommen +1: Claude Sonnet 4.6Removes dependency on langchain.agents factory functions (not available in installed version). Custom tool-calling loop works with any recent langchain-groq version. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>b5c651d -
Fix: use create_openai_tools_agent instead of create_tool_calling_agent
tobyvertommen +1: Claude Sonnet 4.6create_tool_calling_agent not available in installed langchain version. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>df5ced9 -
Fix: switch to create_tool_calling_agent + AgentExecutor
tobyvertommen +1: Claude Sonnet 4.6Replaces LangGraph create_react_agent which caused malformed tool calls on Groq. LangChain's tool-calling agent uses proper JSON tool format. Also adds handle_parsing_errors=True and enforces non-empty answers. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>45b0018 -
Fix: update Groq model to llama-3.3-70b-versatile
tobyvertommen +1: Claude Sonnet 4.6llama-3.1-70b-versatile decommissioned, replaced by 3.3-70b-versatile. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>2bd5204 -
Fix: pass system prompt via SystemMessage instead of state_modifier
tobyvertommen +1: Claude Sonnet 4.6state_modifier not supported in installed LangGraph version. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>5762615 -
Fix: replace duckduckgo-search with ddgs package
tobyvertommen +1: Claude Sonnet 4.6Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>b363f5d -
Add LangGraph + Groq ReAct agent for GAIA benchmark
tobyvertommen +1: Claude Sonnet 4.6Replace placeholder BasicAgent with LangGraph ReAct agent using Groq (llama-3.1-70b-versatile). Tools: DuckDuckGo, Wikipedia, Python REPL, GAIA file fetcher. System prompt enforces exact-match answer format. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>fbc830e
-
Update app.py
Verified signature81917a3 -
Update app.py
Verified signature0ee0419 -
Update app.py
Verified signaturee514fd7 -
Update app.py
Verified signatured59f015 -
Update requirements.txt
Verified signature1232637 -
Update app.py
Verified signature36ed51a -
Update app.py
Verified signature3db6293 -
Update app.py
Verified signaturec1fd3d2
-
Update app.py
Verified signatureb177367 -
Update app.py
Verified signature7d65c66 -
Update app.py
Verified signature9088b99 -
Update app.py
Verified signature3c4371f -
Update app.py
Verified signature9cde2c4 -
Update app.py
Verified signature3e5c3a6
-
Update app.py
Verified signature45d779e -
Update app.py
Verified signature1ca9f65 -
Update app.py
Verified signature4021bf3 -
Update app.py
Verified signaturef6f7832 -
Update app.py
Verified signatureb90251f -
Update app.py
Verified signature7e4a06b -
Update README.md
Verified signatured123508 -
Update app.py
Verified signature335f9f4 -
Update app.py
Verified signature31243f4 -
Update app.py
Verified signaturee80aab9 -
Update app.py
Verified signaturefa7d846 -
Update app.py
Verified signaturee101d07 -
Update README.md
Verified signature62ad9da -
Update app.py
Verified signaturee326bf0
-
Update app.py
Verified signaturedb2d57b -
Update app.py
Verified signatureeccf8e4 -
Update app.py
Verified signatureb5c6c62 -
Update requirements.txt
Verified signaturecb5f8c6 -
Update app.py
Verified signatureba7497e -
Update requirements.txt
Verified signature5aede90 -
Update requirements.txt
Verified signaturea38f55a -
Create app.py
Verified signature10e9b7d