Post

go-second-brain — Knowledge Graph RAG Platform

Markdown KB → Neo4j + Qdrant → Matrix RAG bot and voice assistant.

Posted
By Andriy Oblivantsev
1 min read
go-second-brain — Knowledge Graph RAG Platform

Self-hosted second brain platform: a Markdown knowledge base flows into a Neo4j knowledge graph and Qdrant vector store, then serves answers through a Matrix RAG bot and optional voice assistant.

Repository: github.com/eSlider/go-second-brain

Architecture

  • Ingest: docsparse parses Markdown corpus → graph nodes/edges in Neo4j and embedded chunks in Qdrant
  • RAG: hybrid retrieval (vector search + graph-linked context) → Ollama generation
  • Serving: Matrix bot (!brain prefix), optional low-latency voice CLI (STT/TTS)
  • SDK: public Go packages (services/pkg/*) shipped Jun 2026

Docker Compose stack

ServiceRole
docsMkDocs knowledge base
neo4jKnowledge graph
qdrantVector store
kg-ingestorMarkdown → graph + embeddings
matrix-botMatrix RAG + voice bot

Outcomes

  • Hybrid GraphRAG retrieval combining vector similarity and graph traversal
  • Self-hosted AI stack — no external LLM API required when using Ollama
  • Production-oriented Go SDK for embedding the pipeline in other services

Tech stack

Go · Neo4j (Cypher) · Qdrant · Ollama · Matrix · Docker Compose

Projects, Software Architectures
Edit this post
Go Neo4j Qdrant Ollama RAG Matrix Docker LLM
This post is licensed under CC BY 4.0 by the author.