geek007git/hogyoku-oss

How geek007git/hogyoku-oss works

Hogyoku is a self-hosted, multimodal RAG workspace: users upload private documents, ask questions in managed threads, and receive cited answers whose claims are independently checked before presentation. The documented lifecycle and sampled routes show upload-to-background-ingestion and ask-to-retrieval-to-verification. The graph also preserves the documented browser workspace and offline Python guardrails. Several capabilities, including parsing/OCR internals and answer verification, are documented but not excerpted; their subsystem paths are therefore exact enclosing directories, and uncertain implementation links are omitted.

Main entry points: ingestion.ts Document ingestion · threads.ts Thread routes · app.ts Fastify API · ai.ts Model and embeddings

Components 26

Workspace and API

Browser workspace public/app.js
Fastify API src/app.ts
Accounts and sessions src/http/auth.ts
Document routes src/http/documents.ts
Thread routes src/http/threads.ts
Web server src/server.ts

Document ingestion

Ingestion queue src/lib/queue.ts
Ingestion worker src/worker.ts
Document ingestion src/services/ingestion.ts
Extraction and OCR src/services
Page-aware chunking src/lib/chunking.ts

Conversational RAG

Query rewriting src/services/conversation.ts
Hybrid retrieval src/services/retrieval.ts
Grounded verification src/services/answers.ts

Persistence and providers

Model and embeddings src/services/ai.ts
PostgreSQL and pgvector src/db/client.ts
Private object storage src/lib/storage.ts
Redis
Google Gemini

Offline guardrails

Guardrail utilities python/hogyoku_guardrails/common.py
Citation audit python/hogyoku_guardrails/citation_audit.py
Chunk audit python/hogyoku_guardrails/chunk_audit.py
RAG lint python/hogyoku_guardrails/rag_lint.py
Security scanning python/hogyoku_guardrails/security_scan.py
Security report python/hogyoku_guardrails/security_report.py

Other components

Workspace user

Connections 33

Workspace user → Browser workspace: uses README.md
Browser workspace → Fastify API: calls API
Web server → Fastify API: starts src/server.ts
Fastify API → Accounts and sessions: registers routes src/app.ts
Fastify API → Document routes: registers routes src/app.ts
Fastify API → Thread routes: registers routes src/app.ts
Document routes → Private object storage: uploads files src/http/documents.ts
Document routes → Ingestion queue: enqueues jobs src/http/documents.ts
Web server → Private object storage: ensures bucket src/server.ts
Ingestion worker → Ingestion queue: consumes jobs src/worker.ts
Ingestion worker → Document ingestion: invokes src/worker.ts
Document ingestion → PostgreSQL and pgvector: reads and writes src/services/ingestion.ts
Document ingestion → Private object storage: downloads object src/services/ingestion.ts
Document ingestion → Extraction and OCR: extracts pages src/services/ingestion.ts
Document ingestion → Page-aware chunking: chunks pages src/services/ingestion.ts
Document ingestion → Model and embeddings: embeds chunks src/services/ingestion.ts
Thread routes → Query rewriting: rewrites query src/http/threads.ts
Thread routes → Hybrid retrieval: retrieves evidence src/http/threads.ts
Thread routes → Grounded verification: requests answer src/http/threads.ts
Thread routes → PostgreSQL and pgvector: reads and saves turns src/http/threads.ts
Hybrid retrieval → Model and embeddings: embeds query src/services/retrieval.ts
Hybrid retrieval → PostgreSQL and pgvector: searches indexed chunks src/services/retrieval.ts
Grounded verification → Hybrid retrieval: uses evidence src/services/answers.ts
Grounded verification → Query rewriting: uses history src/services/answers.ts
Grounded verification → Model and embeddings: generates and checks src/services/answers.ts
Model and embeddings → Google Gemini: calls when configured src/services/ai.ts
Model and embeddings → Google Gemini: embeds remotely src/services/ai.ts
Citation audit → Guardrail utilities: uses helpers python/hogyoku_guardrails/citation_audit.py
Chunk audit → Guardrail utilities: uses helpers python/hogyoku_guardrails/chunk_audit.py
RAG lint → Guardrail utilities: uses helpers python/hogyoku_guardrails/rag_lint.py
Security scanning → Guardrail utilities: uses helpers python/hogyoku_guardrails/security_scan.py
Security report → Guardrail utilities: uses helpers python/hogyoku_guardrails/security_report.py
Ingestion queue → Redis: stores jobs

More TypeScript diagrams: superfluid-org/protocol-monorepo · inngest/inngest-js · drawrowfly/instagram-scraper · google/adk-web · redpangilinan/credenza

Drawn by GitDiagram on September 30, 2026 from the repository's file tree, README and source files. Read as Markdown ·