NLIP Knowledge Agent

Long-running NLIP knowledge agent for answering questions about NLIP.

The service accepts ECMA-430-shaped NLIP messages:

{
  "messageType": "request",
  "format": "text",
  "subformat": "english",
  "content": "What is NLIP?"
}

It responds with:

{
  "messageType": "response",
  "format": "json",
  "subformat": "nlip-agent-response",
  "content": {
    "agent": "nlip-knowledge-agent",
    "answer": "...",
    "sources": []
  }
}

Run Locally

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
mkdir -p data
cp data/knowledge_base.sample.json data/knowledge_base.json
uvicorn app:app --host 0.0.0.0 --port 8080

Build The Knowledge Base

pip install -r requirements.txt
python build_knowledge_base.py

The builder loads the NLIP website, ECMA TC56 page, ECMA-430 through ECMA-434, ECMA TR-113, and raw Markdown files from nlip-project/security_guidelines.

Optional Environment Variables

• JETSTREAM_CHAT_URL: OpenAI-compatible Jetstream inference endpoint
• JETSTREAM_API_KEY: optional API key if required by the endpoint
• NLIP_MODEL_NAME: model name, for example llama-4-scout
• KNOWLEDGE_BASE_PATH: path to the retrieval knowledge base
• CONFIDENTIAL_AGENT_URL: Skupper-reachable AWS confidential agent URL

Test

curl -s -X POST http://localhost:8080/nlip \
  -H 'Content-Type: application/json' \
  -d '{"messageType":"request","format":"text","subformat":"english","content":"What is NLIP?"}'