Semantic File System

Local-first semantic search and natural-language file operations powered by Ollama, ChromaDB, and a Python backend, with both Terminal and Tauri desktop UI workflows.

What This Project Includes

• Python semantic runtime (SLPFS) for indexing, search, and NL command execution
• FastAPI backend service used by the desktop app
• Tauri + React frontend for file tree, chat/search, and preview UX
• Optional multimodal indexing/search pipeline (Semantixel)
• Config-driven local setup through a single config file

Architecture

1. User submits a search or command from Terminal or UI.
2. Backend routes requests to runtime services.
3. Files are indexed to vectors using sentence-transformers.
4. ChromaDB returns semantic matches.
5. Ollama-powered LLM handles intent/command parsing and response shaping.

Primary paths:

• slpfs/ core semantic file system runtime
• semantixel/ multimodal services
• backend_api/app/ FastAPI API layer and runtime management
• tauri_ui/ desktop app frontend (React + Tauri)

Prerequisites

Required on all OS:

• Python 3.10+ (3.11 recommended)
• Node.js 18+ and npm
• Ollama installed and available in PATH
• Git

For Tauri desktop development, install Rust toolchain:

• Rustup + stable Rust + Cargo

Install Rust toolchain:

rustup default stable

Install and start Ollama, then pull a model:

ollama serve

In another terminal:

ollama pull qwen2.5:3b

OS Setup and Run Commands

Windows (PowerShell)

git clone https://github.com/taslim121/Semantic_file_system.git
cd Semantic_file_system

python -m venv myenv
.\myenv\Scripts\Activate.ps1

pip install --upgrade pip
pip install -r requirements.txt

cd tauri_ui
npm install
cd ..

Start backend API:

.\myenv\Scripts\Activate.ps1
python -m uvicorn backend_api.app.main:app --host 127.0.0.1 --port 8000 --reload

Start Tauri UI (new terminal):

cd tauri_ui
npm run tauri dev

Optional Terminal UI:

.\myenv\Scripts\Activate.ps1
python terminal.py

macOS (zsh/bash)

git clone https://github.com/taslim121/Semantic_file_system.git
cd Semantic_file_system

python3 -m venv myenv
source myenv/bin/activate

pip install --upgrade pip
pip install -r requirements.txt

cd tauri_ui
npm install
cd ..

Start backend API:

source myenv/bin/activate
python -m uvicorn backend_api.app.main:app --host 127.0.0.1 --port 8000 --reload

Start Tauri UI (new terminal):

cd tauri_ui
npm run tauri dev

Optional Terminal UI:

source myenv/bin/activate
python terminal.py

Linux (Ubuntu/Debian)

System packages commonly needed for Tauri/WebKit:

sudo apt update
sudo apt install -y build-essential curl wget file libxdo-dev libssl-dev libayatana-appindicator3-dev librsvg2-dev

Then project setup:

git clone https://github.com/taslim121/Semantic_file_system.git
cd Semantic_file_system

python3 -m venv myenv
source myenv/bin/activate

pip install --upgrade pip
pip install -r requirements.txt

cd tauri_ui
npm install
cd ..

Start backend API:

source myenv/bin/activate
python -m uvicorn backend_api.app.main:app --host 127.0.0.1 --port 8000 --reload

Start Tauri UI (new terminal):

cd tauri_ui
npm run tauri dev

Optional Terminal UI:

source myenv/bin/activate
python terminal.py

Configuration

Edit config.yaml before first use.

Current important keys:

• directories.root_dir: folder to index and operate on
• directories.vector_db_dir: local Chroma storage
• ollama.model: LLM model (currently qwen2.5:3b)
• ollama.url: Ollama server URL (default local)
• embedding.model: embedding model name
• multimodal.enabled: toggle multimodal indexing/search
• multimodal.db_path: multimodal vector DB path

Example:

directories:
  root_dir: C:\Users\YourUser\Desktop\your_folder
  vector_db_dir: ./.lsfs_db

ollama:
  model: qwen2.5:3b
  url: http://localhost:11434

embedding:
  model: all-mpnet-base-v2

search:
  default_results: 5
  max_file_size_mb: 10

multimodal:
  enabled: true
  db_path: ./db_multimodal

Running Modes

1) API + Desktop UI (recommended)

• Start Ollama
• Start backend API
• Start Tauri app

This mode is best for daily use.

2) Terminal-Only Mode

python terminal.py

Useful for debugging backend behavior without UI.

Backend Commands

Run these from the repository root.

Install backend dependencies

Windows PowerShell:

.\myenv\Scripts\Activate.ps1
pip install -r requirements.txt

macOS/Linux:

source myenv/bin/activate
pip install -r requirements.txt

Start backend API in development

Windows PowerShell:

.\myenv\Scripts\Activate.ps1
python -m uvicorn backend_api.app.main:app --host 127.0.0.1 --port 8000 --reload

macOS/Linux:

source myenv/bin/activate
python -m uvicorn backend_api.app.main:app --host 127.0.0.1 --port 8000 --reload

Start backend API without reload

python backend_api/app/main.py

Run terminal client

python terminal.py

Frontend Commands

Run these from tauri_ui/.

Install frontend dependencies

npm install

Start React frontend only

This starts the Vite dev server on http://localhost:1420.

npm run dev

Start desktop app in development

This starts the Tauri desktop shell. The Rust side will also start the Python backend automatically when the app launches.

npm run tauri dev

Build frontend assets

npm run build

Preview the production frontend build

npm run preview

Build desktop app bundle

npm run tauri build

API Endpoints (MVP)

Base URL:

http://127.0.0.1:8000/api/v1

Key endpoints:

• GET /health
• GET /config
• PUT /config
• PUT /config/root
• POST /search
• POST /command
• GET /tree
• POST /file/read
• POST /index/start
• GET /index/status
• POST /index/cancel

Troubleshooting

Ollama not reachable

Connection refused / model unavailable

Fix:

ollama serve
ollama pull qwen2.5:3b

Backend import errors

Make sure virtual environment is active and dependencies are installed from both requirement files.

Tauri build issues

• Confirm Rust is installed: rustc --version
• Confirm Node/npm are installed: node -v and npm -v
• On Linux, install missing system packages for WebKit/GTK

Reset vector stores

Windows PowerShell:

Remove-Item -Recurse -Force .lsfs_db, db_multimodal

macOS/Linux:

rm -rf .lsfs_db db_multimodal

Recommended Ollama Models

• Balanced default: qwen2.5:3b
• Better quality (if resources allow): qwen2.5:7b
• Alternative general model: llama3.1:8b

Update the model in config.yaml, then restart backend.

Developer Notes

• Backend API entrypoint: backend_api/app/main.py
• UI scripts: tauri_ui/package.json
• Contract and planning docs:
  • backend-contract.md
  • tauri-implementation-plan.md

License

Use according to repository license and applicable third-party dependency licenses.