Added openai

Leaderboard score 40% using single GPT-4.1

Switched to single agent powered by GPT-4.1, added step wait function to avoid hitting the OpenAI API rate limit.

Added function to save chemistry textbook contents as markdown, refactored tool helper functions.

Added chapter parser function.

Added custom selenium based web search functions to look for academic text books via LibreTexts.

Added summarization of old messages when context window gets too long, switched to DeepSeek-V3, scored 30%!

Cleaned up

Downgraded to smolagents 1.18.0 and switched back to single codeagent - scored 20% on GAIA benchmark.

Split tools up amoung wikipedia agent and websearch agent and implementend manager agent, all using qwen2.5-coder-32B.

Updated wikipedia search tools to get and parse wikipedia pages to HTML so that tables and other non-text elements are visible to agent. Allowed agent to import BeautifulSoup.

Added unittest for google_search() function.

Added langchain-community

Added custom google search tool using googlesearch-python.

Refactored and cleaned up.

Updated langchain import

Reverted

Added CodeAgent with web search and scraping tools, added GAIA prompt.

Updated environment

Update README.md

Pinned space to profile

Changed HF space title

Updated README metadata

Set Gradio app to start on attach.

Added Oauth

Add devcontainer configuration.

Updated environment

Typo

Update app.py

Update app.py

Update app.py

Update app.py

Update requirements.txt

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update README.md

Update app.py