⚠️ IMPORTANT NOTICE REGARDING PLAGIARISM & AUTHORSHIP ⚠️

It has come to my attention that the GitHub user @tfantas (Thiago Antas) and his automated account @jarvis-aix are falsely claiming credit for my architecture. They have explicitly listed my original repositories (including RAG_enterprise_core, smart-ingest-kit, DAUT, etc.) as their own '🔬 Featured Work' on their public profile without authorization or proper attribution. Below is the documented proof.

https://github.com/tfantas seems to have 20+ years of expirience but no own ideas .... Im gonna make him famous...... If you enjoyed my repos and found them useful, Im sorry but im out of this game !!! No more opensource Sorry Im sure you will find my further developed Repos at https://github.com/jarvis-aix .... What a disgrace and disrespect !

This repository, the Multi-Lane Consensus Architecture, and the V4.0 Manifest are 100% my original work, built over two years. Please be highly cautious of actors in the AI space attempting to rebrand, clone, or take credit for this Enterprise RAG system

⚠️ ⚠️ ⚠️

Knowledge Base Self-Hosting Kit

A production-ready, self-hosted RAG memory layer that pairs with local LLMs, chains, and autonomous agents.

• ChromaDB + Docling powers ingestion of everything from PDFs to code repositories.
• FastAPI backend exposes CRUD/ingestion/search APIs plus a log-friendly health endpoint.
• Modern UI includes dashboards, ingestion tooling, and now a dedicated Agent Configuration tab for MCP clients.

---

🚀 Quick Start

1. git clone https://github.com/2dogsandanerd/Knowledge-Base-Self-Hosting-Kit.git
2. cd Knowledge-Base-Self-Hosting-Kit
3. cp .env.example .env and adjust DOCS_DIR, LLM provider settings, etc.
4. docker compose up -d
5. curl http://localhost:8080/health to check the backend.
6. Open http://localhost:8080 in your browser.

Services (all exposed via the root nginx gateway):

• Web UI + Agent Configuration tab: http://localhost:8080/
• OpenAPI docs: http://localhost:8080/docs
• Health check: http://localhost:8080/health
• API root: http://localhost:8080/api/v1/rag

---

⚙️ Features at a Glance

• Ingestion modes: single uploads, folder scans, hybrid chunking (vector + BM25).
• Agent-ready search: run semantic queries with k tuning, view citations, and see streaming logs.
• MCP integration: connect OpenClaw or other MCP-aware agents via the included MCP server.
• Agent Configuration tab: configure MCP connection details without touching .env or the CLI.
• Local LLM friendly: default Ollama setup shares host models, but OpenAI-compatible servers are supported via OPENAI_BASE_URL.

---

🧠 Agent Configuration (New Tab)

The UI now surfaces an Agent Configuration tab where you can:

• Review the MCP-friendly CLI snippet.
• Set KNOWLEDGE_BASE_API_URL, timeouts, and log levels for the MCP server directly.
• Test connectivity before handing the config to your agent.

Recommended OpenClaw command

openclaw mcp add --transport stdio knowledge-kit npx -y @knowledge-kit/mcp-server

Existing agents can point at the tab’s configuration values (copyable) or use the CLI above to bootstrap the connector.

---

📁 Configuration

Variable	Default	Description
PORT	8080	External port for the nginx gateway
DOCS_DIR	./data/docs	Host path mounted at /host_root for folder ingestion
LLM_PROVIDER	ollama	Model provider (ollama, openai, anthropic, gemini, openai_compatible)
LLM_MODEL	llama3:latest	Selected model for LLM calls
EMBEDDING_PROVIDER	ollama	Embedding provider (usually same as LLM)
EMBEDDING_MODEL	nomic-embed-text	Embedding model name
CHUNK_SIZE	512	Document chunk size
CHUNK_OVERLAP	128	Chunk overlap for context continuity
DEBUG	false	Toggle verbose logging
LOG_LEVEL	INFO	DEBUG/INFO/WARN/ERROR

For OpenAI-compatible servers, set LLM_PROVIDER=openai_compatible and point OPENAI_BASE_URL to your local endpoint.

---

🔍 Ingestion & Search API

• POST /api/v1/rag/collections: create a collection with embedding metadata.
• POST /api/v1/rag/documents/upload: upload PDF/MD/TXT files.
• POST /api/v1/rag/ingest-folder: kick off folder ingestion (uploads via mounted DOCS_DIR).
• GET /api/v1/rag/collections/:name/stats: inspect document counts.
• POST /api/v1/rag/query: semantic queries (vector + BM25 fusion) with citation data.
• POST /api/v1/rag/search: immediate keyword search without LLM generation.

Use the frontend or any HTTP client; JSON responses include answer, sources, and scoring metadata.

---

🧪 MCP Configuration API

• GET /api/v1/config: read current .env values via ConfigService.
• POST /api/v1/config: persist updates sent from the UI.
• POST /api/v1/config/test: validate downstream connections (LLM, Ollama, etc.) before deploying the change.

This powers the new Agent Configuration tab so you can tune the connector without restarting containers.

---

🛠️ Development

• Backend: cd backend && pip install -r requirements.txt && uvicorn src.main:app --reload.
• Frontend is static; editing frontend/index.html suffices.
• Build stack with Docker: docker compose build to refresh containers.
• Configuration persists in .env, but the UI uses /api/v1/config for live editing.

MCP Server

The companion connector lives in mcp-server/ and publishes @knowledge-kit/mcp-server. Run it with:

cd mcp-server
npm install
npm run build
npm start

---

📚 Documentation & Design Notes

• Architecture diagrams, MCP guides, and customization tips live under docs/.
• CHROMADB and Ollama containers are orchestrated via docker-compose.yml.
• Use smithery/ configs to run the MCP server inside Smithery.

---

❤️ Contributing

1. Open an issue describing your feature or bug.
2. Create a branch off main and target your change at the repo’s layout.
3. Keep docs, tests, and UI changes in sync when you refactor functionality.
4. If you update MCP wiring, ensure the Agent Configuration tab and mcp-server read the same env names.