Agent Framework Code Execution Demos

Collection of code execution agents demonstrating different AI agent frameworks with OpenTelemetry tracing.

Frameworks Included

Framework	Agent	Description
LangGraph	langgraph_code_agent.py	Graph-based code generation with retry logic
OpenAI	openai_agents_code.py	Function calling with code execution
AutoGen	autogen_code_agent.py	Multi-agent conversation for coding tasks
CrewAI	crewai_code_agent.py	Role-based crew for code development

Features

• Code Execution: All agents can write and execute Python code
• OpenTelemetry: Full OTEL tracing to Jaeger/OTLP collectors
• Error Recovery: Automatic retry and error fixing
• Multiple Tasks: Each agent demonstrates various coding challenges

Quick Start

1. Start OTEL Collector

cd otel-collector
docker compose up -d

2. Set Environment Variables

export OPENAI_API_KEY=your_key
export OTEL_EXPORTER_OTLP_ENDPOINT=localhost:4317

3. Install Dependencies

pip install langgraph langchain langchain-openai opentelemetry-api opentelemetry-sdk opentelemetry-exporter-otlp
pip install crewai crewai-tools
pip install autogen-agentchat autogen-ext[openai]

4. Run Agents

# LangGraph
python langgraph_code_agent.py

# OpenAI Function Calling
python openai_agents_code.py

# AutoGen Multi-Agent
python autogen_code_agent.py

5. View Traces

Open Jaeger UI at http://localhost:16686

Trace Services

• langgraph-code-agent
• openai-agents-code
• autogen-code-agent
• crewai-code-agent

Use with ShadowMap

Export traces for ShadowMap analysis:

curl "http://localhost:16686/api/traces?service=langgraph-code-agent&limit=10" > traces.json

License

MIT