Self-hosted stack
FastAPI, LiteLLM, vLLM, PostgreSQL + pgvector, Redis, Kafka, Langfuse, OCR, and Huawei Cloud services.
Technical architecture notes for the self-hosted government RAG platform
FastAPI, LiteLLM, vLLM, PostgreSQL + pgvector, Redis, Kafka, Langfuse, OCR, and Huawei Cloud services.
Detailed pages cover OCR, data ingestion, long-term memory, retrieval, model serving, APIs, security, operations, and evaluation.
Hybrid retrieval, source citations, faithfulness checks, and the Government Policy Guard protect every generated answer.