Agentic Research

Develop an AI agent using SmolAgent Framework that uses ReAct(Reason - Act) feedback system with deep research capabilities that can be easily used by users to tackle real world problems in a faster method.

Langfuse-Powered Smolagents with Hugging Face

Overview

This project integrates Langfuse with Smolagents to monitor and improve reasoning and decision-making capabilities when executing multi-step tasks. By utilizing OpenTelemetry tracing and AI models from Hugging Face, this implementation allows real-time performance tracking and optimization of autonomous agents.

Features

• Langfuse Integration: Observability with OpenTelemetry.
• Hugging Face AI Model: Uses Qwen/Qwen2.5-Coder-32B-Instruct.
• Multi-Agent System: Implements CodeAgent and ToolCallingAgent.
• Web Search & Data Retrieval: DuckDuckGo and Webpage visiting tools.
• Replay Functionality: Allows re-execution of previous runs.

Demo