System Overview

The online user path is:

  1. The React frontend issues /chat/* requests.

  2. Local proxying or Nginx rewrites those routes to /api/v1/*.

  3. The API gateway enforces authentication, rate limiting, and proxy behavior.

  4. The RAG service orchestrates retrieval, graph support, generation, citations, and persistence.

Simplified RagLogic AI architecture diagram

Simplified architecture diagram generated from docs/architecture_simple.puml.

Runtime building blocks

Layer

Main responsibilities

Representative code

Frontend

Request initiation, chat state, source and timing display

frontend/src/contexts/ChatContext.tsx

API gateway

Auth, proxying, rate limiting, edge metrics

services/api-gateway/main.py

RAG runtime

Query orchestration, retrieval, generation, SSE

services/rag-service/main.py

Data stores

PostgreSQL, Qdrant, Neo4j, Redis

infra/compose/*.yml

Async pipeline

Chunking, embedding, extraction workers

services/*worker/, packages/lalandre_*

Observability

Prometheus, Grafana, exporters

monitoring/

For the rationale behind this split, see Architecture Notes.

Notes

  • Purpose: give engineers a stable overview of the online and offline system boundaries.

  • Source of truth: docs/architecture_simple.puml and the maintained SVG export.

  • Audience: engineers onboarding to the current stack.

  • Notes: this diagram is intentionally simplified; the detailed version lives in the audit page.