- Python 3.12 or higher
- pip (Python package installer)
- ERC3 API key (get it at https://erc.timetoact-group.at/)
- LLM provider API key (OpenRouter or Cerebras)
- Node.js v18 or higher (optional - only for trace visualizer)
git clone https://github.com/IlyaRice/Enterprise-RAG-Challenge-3-AI-Agents.git
cd Enterprise-RAG-Challenge-3-AI-Agentspython -m venv .venv
source .venv/bin/activate # or .venv\Scripts\activate on Windowspip install -r requirements.txtCopy .env.example to .env:
cp .env.example .env # or copy .env.example .env on WindowsEdit .env and set the following required variables:
# Required
ERC3_API_KEY=your_erc3_api_key_here
OPENROUTER_API_KEY=your_openrouter_api_key_here
LANGFUSE_TRACING_ENABLED=0Optional variables:
CEREBRAS_API_KEY- Direct Cerebras API access (alternative to OpenRouter)USER_NAME- Prefix for session names in dashboardLOCAL_DEV- Set to1to prioritize .env over system environmentLANGFUSE_SECRET_KEY,LANGFUSE_PUBLIC_KEY,LANGFUSE_BASE_URL- LLM observability setup
OpenRouter (recommended): Works reliably for all tasks. Use this unless you have specific requirements.
Cerebras: Direct API access needed if you want to use reasoning_effort parameter (ignored when routing through OpenRouter).
Test with a single task:
python main.py erc3-dev --tasks 1 -vExpected output: Task execution with detailed logs (verbose mode), ending with score and completion message.
Open the visualizer - runs in your browser, no installation needed.
The visualizer has three tabs in the left sidebar:
- Tasks - View the execution tree and details for the current trace
- Demos - Pre-loaded demo traces
- Files - Browse traces from your local
traces/folder or upload individual files
Only needed if you want to modify the visualizer code or keep everything offline.
cd trace-viewer
npm installcd trace-viewer
npm run devOpen http://localhost:3000 in your browser.