RagLogic AI Technical Documentation

This site is the internal technical source of truth for the current RagLogic AI stack: runtime architecture, ingestion pipelines, and observability.

Note

The documentation is Markdown-first and built with Sphinx. Long-form legacy notes remain in docs/ as source material, but the pages below are the curated entry points.

Entry points

  • Getting Started: repo layout, local commands, and documentation workflow.

  • Architecture: system overview, architecture notes, request flow, ingestion pipeline, and diagram audit.

  • Runtime: query modes, retrieval, graph support, agentic orchestration, streaming semantics, and the GraphRAG commercial brief.

  • Operations: observability, production surfaces, dashboards, and chart verification.

  • API Reference: generated package reference built from Python docstrings.

What is covered here

  • How requests move from the frontend to the API gateway and the RAG runtime.

  • How the repository tree maps to the runtime and data architecture.

  • How indexing and enrichment jobs populate PostgreSQL, Qdrant, and Neo4j.

  • How the GraphRAG layer translates into a commercially defensible product story.

  • Which metrics exist today, which dashboards consume them, and where coverage is incomplete.

  • Which production URLs are exposed behind Traefik, and which ones are internal-only.

  • How the reusable Python packages are exposed through docstring-driven API documentation.

Documentation rules

  • Narrative pages explain behavior and expected system shape.

  • Chart pages document the source of truth, intended audience, and validation status.

  • Generated HTML is not committed; docs/_build/ stays local or CI-only.