Documentation
¶
Overview ¶
Command runbook-ingest builds the runbook-RAG corpus for the SRE agent's read-only find_runbook tool. It scans the runbook source directory of Markdown runbooks, embeds each through the configured OpenAI-compatible embeddings endpoint, and persists the vectors to the same storage backend the server reads at boot.
You usually do NOT need to run this: when tools.find_runbook.embedding_model is set, the server auto-ingests the runbook source directory at boot (incrementally — only new or edited runbooks are embedded). This command exists for pre-baking the corpus out-of-band, e.g. in CI or an air-gapped image build, so the server starts with the corpus already populated.
This is the WRITE path: it lives in its own command so the corpus can be built without a running server, and the analyze tool catalog stays read-only.
Usage:
runbook-ingest [-config config/config.yaml] [-org ORG]
Place your *.md runbooks in the data folder under runbooks/ (e.g. ./data/runbooks; /app/data/runbooks in the container image). The embedding model comes from tools.find_runbook.embedding_model and the API credential is the shared agent.ai.api_key. The embedding backend is selected by agent.ai.provider (openai / ollama / gemini), so pointing provider at a local Ollama keeps embeddings inside the operator's own network.
Source Files