Update README.md

README.md

@@ -1,14 +1,42 @@
 ---
-title: GAIA Agent Evaluator (Display Only) # Or your title
+title: GAIA Agent Evaluator
 emoji: 🚀
 colorFrom: blue
 colorTo: green
 sdk: gradio
-sdk_version: 5.25.2 # Use your Gradio version
+sdk_version: 5.25.2
 app_file: app.py
 pinned: false
-hf_oauth: true
-# Add this section to install system packages:
+hf_oauth: true # Enable Login button
 packages:
   - ffmpeg
-  - stockfish
+  - stockfish
+---
+
+# GAIA Agent Evaluation Runner
+
+This Space runs an AI agent designed to answer questions from the GAIA benchmark (Level 1 subset).
+
+**Dependencies:**
+
+This space requires Python packages listed in `requirements.txt`.
+
+It also requires the following system packages:
+*   `ffmpeg`: For processing audio files (used by Whisper).
+*   `stockfish`: The chess engine used for Question 4.
+
+Add this to your Dockerfile or specify system packages if using other methods. For standard Spaces, add `apt-get install -y ffmpeg stockfish` commands appropriately (e.g., some spaces allow a startup script or Docker commands).
+If using default Spaces runtime, you might need to handle installing these differently, potentially bundling Stockfish or checking if ffmpeg is pre-installed.
+
+**Setup:**
+
+1.  Add your `OPENAI_API_KEY` as a Secret in the Space settings.
+2.  (Optional) Add `TAVILY_API_KEY` as a Secret for Tavily search.
+3.  Ensure Stockfish is installed and accessible via the `stockfish` command or set the `STOCKFISH_PATH` secret.
+
+**Usage:**
+
+1.  Log in using the Hugging Face Login button.
+2.  Click "Run Evaluation & Submit All Answers".
+3.  Wait for the agent to process all questions (this can take several minutes).
+4.  View the results and score.