Knowledge Reuse Platform

An AI assistant that answers questions using your Obsidian Vault as the knowledge base, powered by RAG (Retrieval-Augmented Generation) with hybrid search.

Features

• Hybrid Search: Combines vector similarity (ChromaDB) and BM25 keyword search for accurate retrieval
• AI-Generated Answers: GPT-4o-mini synthesizes answers grounded in your vault content, with source citations
• Incremental Indexing: Detects modified notes via file modification time and indexes only the changes
• Full Re-indexing: Reset and rebuild the entire vector index on demand
• Auto Indexing: Background scheduler keeps the index up to date automatically
• Search History: Saves past queries for quick re-access
• Note Tagger: Standalone UI for bulk-labeling notes with content_origin frontmatter

Tech Stack

Component	Technology
UI	Streamlit
Vector Store	ChromaDB
Embedding	OpenAI text-embedding-3-small
Generation	OpenAI gpt-4o-mini
Keyword Search	BM25 (rank-bm25)
Scheduler	APScheduler

Prerequisites

• Python 3.9+
• OpenAI API key
• Obsidian vault

Installation

git clone <repo-url>
cd knowledge-reuse-platform

python -m venv venv
source venv/bin/activate       # Windows: venv\Scripts\activate

pip install -r requirements.txt

Configuration

Create a .env file in the project root:

OPENAI_API_KEY=sk-...
VAULT_PATH=/path/to/your/vault
VAULT_NAME=MyVault

# Optional — defaults shown
TOP_K=10
VECTOR_WEIGHT=0.7
BM25_WEIGHT=0.3
AUTO_INDEX_ENABLED=true
AUTO_INDEX_INTERVAL_MINUTES=5
GENERATION_MODEL=gpt-4o-mini
MAX_CONTEXT_CHUNKS=5

Alternatively, configure everything from the Settings page in the app.

Usage

Main App

streamlit run app/main.py

Page	Description
Search	Enter a query; get an AI-generated answer with source citations and Obsidian deep links
Indexing	Run incremental or full re-indexing with live log output and index statistics
Settings	Manage API key, vault path, search parameters, and auto-indexing schedule

Note Tagger

A separate Streamlit tool for bulk-labeling notes with content_origin frontmatter (scraps / original).

streamlit run scripts/tagger.py

• Filter notes by filename or path (e.g. clippings/2025)
• Select individual notes or use Select All
• Apply scraps or original label in bulk via the sidebar
• Automatically excludes Excalidraw, Templater, and Kanban files

Project Structure

knowledge-reuse-platform/
├── app/
│   └── main.py                    # Streamlit UI (Search / Indexing / Settings)
├── pipeline/
│   ├── indexing.py                # Indexing pipeline
│   └── search.py                  # Search pipeline
├── components/                    # Core processing components
│   ├── loader.py
│   ├── parser.py
│   ├── normalizer.py
│   ├── chunker.py
│   ├── retriever.py
│   ├── response_generator.py
│   └── bm25_index.py
├── infra/
│   ├── embedding/                 # EmbeddingModel interface + OpenAI impl
│   └── vectorstore/               # VectorStore interface + ChromaDB impl
├── scheduler/
│   └── incremental_indexer.py
├── domain/
│   ├── models.py
│   └── id_utils.py
├── scripts/
│   └── tagger.py                  # Standalone note tagging tool
├── config.py
└── requirements.txt

License

MIT