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-Agent

A general-purpose LLM agent built to solve the final assignment of the Hugging Face Agents course, which consists of 20 Level 1 questions from the GAIA benchmark.

The agent uses a variety of tools—web search, Wikipedia extraction, file parsing, audio transcription, and more—to gather evidence and reason through answers. It is built with LangGraph and LangChain, and all interactions are tracked with LangFuse.

🛠️ Technical Details

Component	Technology
Agent Framework	LangGraph + LangChain
LLM	Meta LLaMA 4 Maverick 17B 128E Instruct via Groq
Web Search	SerperAPI (Google Search), requests, Playwright for dynamic content rendering
HTML Parsing	BeautifulSoup, markdownify to convert HTML to Markdown
Wikipedia	Wikimedia API
File Parsing	Unstructured (PDF, DOCX, PPTX), pandas (CSV, TSV, XLSX)
Audio Transcription	OpenAI Whisper (base)
Monitoring	LangFuse
Frontend UI	Gradio, based on the Final Assignment Template

🔐 Setup Instructions

Clone the repository:

git clone https://github.com/phucdev/GAIA-agent.git
 cd GAIA-agent

Install dependencies:
```
pip install -r requirements.txt
```
Create API keys and set-up environment variables: Use env.example to create a .env file in the root directory and replace the placeholders with your actual API keys:
```
GROQ_API_KEY=your_groq_api_key
SERPER_API_KEY=your_serper_api_key
... (other API keys as needed)
```
Optional: For rendering JS with Playwright, you may need some additional setup:
```
playwright install
sudo apt-get install libgtk-3-0
```
Run the app with:
```
python app.py
```