🧠 RAG Memory Bot

A local-first RAG system with persistent memory, exact replay and interactive learning via web interface — no retraining required.

Learn new knowledge instantly by adding and verifying content directly in the UI.
Generate once, store forever, retrieve exactly — or refine and evolve your knowledge over time.

---

🚀 Key Idea

This system separates generation from memory:

• The LLM generates content
• The system stores and curates it
• Future answers are retrieved — not hallucinated

👉 Knowledge improves over time through human feedback.

---

🚀 Overview

RAG Memory Bot is a local FastAPI-based prototype for a retrieval-augmented long-term memory system with focus on three core modes:

• Generate: create new content
• Replay: reproduce stored content exactly
• RAG Answer / Variation: use stored knowledge for answers or variations

The system combines Markdown storage, SQLite metadata, full-text search (FTS5), embedding-based retrieval, and a lightweight web interface.

---

✨ Features

• Persistent storage of generated content as Markdown artifacts
• Metadata management via SQLite
• Hybrid search combining:
  • Full-text search (FTS5)
  • Semantic search (embeddings)
  • Status / feedback ranking
• Session-aware replay and variation
• Review workflow with statuses:
  • draft, saved, favorite, fixed, verified, archived
• Import of existing Markdown files
• Web interface for:
  • Chat
  • Artifact management
  • Review and verification
  • Data import

---

🧱 Architecture

User → API (FastAPI) → Orchestrator

Orchestrator coordinates:

• Retrieval (FTS + vector search)
• Storage (Markdown + SQLite)
• LLM (Ollama)

---

🖼️ Screenshots

(Add your screenshots here)

---

🎯 Use Cases

🧑‍🏫 Education

• Course-specific assistants
• FAQ systems
• Knowledge-based tutoring

🏢 IT / Support

• Internal knowledge systems
• Policy-aware assistants
• Local, privacy-friendly AI

✍️ Creative Systems

• Story generation with memory
• Replayable narratives
• Interactive storytelling

🧠 Personal Knowledge Base

• Persistent assistant memory
• Retrieval of past answers
• Knowledge refinement via feedback

---

⚙️ Installation

1. Clone repository

git clone https://github.com/YOUR_USERNAME/rag-memory-bot.git
cd rag-memory-bot

2. Setup environment

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

3. Install Ollama

https://ollama.com

Check installation:

ollama list

4. Initialize database

python3 scripts/init_db.py

5. Start server

uvicorn app.main:app --host 0.0.0.0 --port 8000

---

🌐 Web Interface

Open in browser:

http://localhost:8000

---

🧪 Example Workflow

1. Generate content
   → "Tell me a story about a hedgehog in a storm"

2. Verify content in the UI
   → Mark artifact as verified

3. Replay
   → "Tell me the same story again"
   → Exact same output

4. Ask a knowledge question
   → "What was the story about?"
   → Answer is generated via RAG using stored content

---

📂 Project Structure

app/
  api/
  core/
  storage/
  retrieval/
  llm/
  indexing/
  templates/
  static/

data/
  artifacts/
  db/

docs/
  Startanleitung.md

scripts/

---

🔐 License

This project is licensed under the MIT License.

Note: This project uses third-party models (e.g. via Ollama).
Their respective licenses apply separately.