Static Git Graph RAG · Case Study 05

Astra Knowledge Graph Engine

I built a file-based Graph RAG layer inside the repo. The build writes graph and MiniSearch indexes. Runtime loads them into memory and serves cached results quickly.

PRODUCTIONBack to case studies

status

PRODUCTION

environment

Git-controlled static JSON + in-memory MiniSearch

ingress

Next.js API Routes

runtime graph

5 nodes / 4 edges

View system map

System map

Astra Knowledge Graph Engine

Env: Git-controlled static JSON + in-memory MiniSearchIngress: Next.js API Routes

Problem: I wanted fast site retrieval without a managed vector database. The system needed low token use, fast lookup, and a Git-friendly build path.

My engineering note

I built a file-based Graph RAG layer inside the repo. The build writes graph and MiniSearch indexes. Runtime loads them into memory and serves cached results quickly.

Live path

deserialization: Manifest Index -> MiniSearch trie

Path running

Architecture Decision

Why I chose this design.

Short decision notes tied to the code or config that mattered.

Decision

lib/context.ts

I chose build-time indexing because the site content already lives in Git. It keeps retrieval cheap, fast, and easy to deploy.

Code noteSee engineering notes

lib/context.tstypescript

typescript

import { getCompressedContext } from "@/lib/context";import { runManojFastAgent } from "@/lib/manoj-agent";
// Eager index deserialization eliminates runtime overhead// Warm memory cache resolves identical queries in ~0.1msconst cache = new Map<string, CompressedContext>();
export function getContext(query: string) {  const key = query.trim().toLowerCase();  if (cache.has(key)) return cache.get(key);
  const context = getCompressedContext(query);  cache.set(key, context);  return context;}

import { getCompressedContext } from "@/lib/context";import { runManojFastAgent } from "@/lib/manoj-agent";
// Eager index deserialization eliminates runtime overhead// Warm memory cache resolves identical queries in ~0.1msconst cache = new Map<string, CompressedContext>();
export function getContext(query: string) {  const key = query.trim().toLowerCase();  if (cache.has(key)) return cache.get(key);
  const context = getCompressedContext(query);  cache.set(key, context);  return context;}

Next case study

AI Home Lab API Gateway

Python FastAPI gateway - I built a local OpenAI-compatible gateway so my site, agents, and scripts could call one stable /v1 contract. Ollama, Po...

Need this level of architecture review?

Bring the hard system constraint: retrieval quality, agent failure modes, latency, evaluation, deployment topology, or technical market education.

Start Advisory Intake View GitHub