README.md

metadata

title: Template Final Assignment
emoji: 🕵🏻‍♂️
colorFrom: indigo
colorTo: indigo
sdk: gradio
sdk_version: 5.25.2
python_version: 3.11
app_file: app.py
pinned: false
hf_oauth: true
hf_oauth_expiration_minutes: 480

SmolaAgents Course Final Assignment

This application is a demonstration of creating a robust agent system with monitoring and tracing capabilities.

Features

• Multi-Agent System: Uses a manager agent to orchestrate a specialized web agent.
• Advanced Models: Leverages Qwen and DeepSeek models through HuggingFace's Inference API.
• Gradio Interface: Simple UI for agent interaction and evaluation.
• Comprehensive Tracing: Integrated observability through Langfuse and Phoenix.

Tracing and Monitoring

The application includes a modular tracing system that supports both:

Langfuse

• Cloud-based LLM observability platform
• Configure with environment variables:
  • LANGFUSE_PUBLIC_KEY
  • LANGFUSE_SECRET_KEY
  • LANGFUSE_HOST (defaults to EU region)

Arize Phoenix

• Local, open-source LLM observability tool
• Start the Phoenix UI with:

  python -m phoenix.server.main serve
  

• View traces at http://localhost:6006

You can configure which tracing system to use:

# Use Langfuse (default)
initialize_tracing(provider="langfuse")

# Use Phoenix
initialize_tracing(provider="phoenix")

Due to limitations in the OpenTelemetry integration, only one provider can be active at a time. Simply choose the provider that best meets your needs.

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