metadata

title: Template Final Assignment
emoji: 🕵🏻‍♂️
colorFrom: indigo
colorTo: indigo
sdk: gradio
sdk_version: 5.25.2
app_file: app.py
pinned: false
hf_oauth: true
hf_oauth_expiration_minutes: 480

GAIA Benchmark Agent - Modular Multi-Modal Architecture

This Space is built on the official agents-course/Final_Assignment_Template base. The architecture strictly preserves the original constants and UI, but replaces the agent logic with a fully modular, multi-modal, GAIA-compliant agent.

Key Features

ModularGAIAAgent: Handles multi-modal, multi-step reasoning, tool use, file handling, and strict GAIA output formatting.
Tool/LLM Registry: Easily extensible for new tools, models, and modalities.
File Handling: Supports text, CSV, Excel, JSON, images, audio, and code files, with automatic type detection and routing.
Adaptive Reasoning: Plans and chains tool/model calls as needed for each question.
GAIA-Compliant Output: Ensures answers are formatted to GAIA standards.
Trace Logging: Internal reasoning trace for each answer (for debugging and transparency).

Usage

Log in with your Hugging Face account.
Click 'Run Evaluation & Submit All Answers' to fetch questions, run the agent, and submit answers for scoring.
The UI and constants (such as DEFAULT_API_URL) are unchanged from the official template, ensuring full compatibility with the GAIA evaluation system.

Customization

To extend the agent, add new tools or models to the TOOL_REGISTRY and update the logic in ModularGAIAAgent.
The agent is designed for easy adaptation to new modalities and reasoning strategies.

Note: This implementation is intentionally modular and extensible, but the public interface and constants remain as required by the course template.

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference