Fetch: Development Workspace

Status: MVP / Recovery Release v0.2.0 Phase: 3 (Mobile Interface MVP)

This repository contains Fetch, a privacy-sovereign executive assistant designed to draft email responses using local AI.

Fetch features "Archimedes"—a wise, slightly sassy academic dog persona who helps triage your inbox with intellectual rigor and emotional empathy.

⚠️ Note: This is the v0.2.0 recovery release. Core functionality is implemented. See CHANGELOG.md for details.

Project Structure

• src/: Python backend (FastAPI + SQLModel + Ollama).
• mobile/: Mobile client (Expo + Tamagui + TypeScript).
• docs/: Project documentation and execution plans.
• data/: Local database and storage.

Getting Started

Backend (Python)

1. Environment: Ensure Python 3.11+ and Ollama are installed.
2. Install: pip install -r requirements.txt
3. Run:

   uvicorn src.main:app --reload --host 0.0.0.0

Mobile Client (Expo)

1. Environment: Ensure Node.js is installed.
2. Install:

   cd mobile
   npm install

3. Run:

   npx expo start --clear

   Note: If you encounter OOM errors, see docs/mobile_connectivity_diagnostics.md.

Features (Current State)

• Draft Generation: Ingests emails and generates drafts using local LLMs.
• Mobile Review: "Tinder-style" card stack UI for reviewing, approving, and rejecting drafts.
• Local-Only: Designed to run entirely on the local network for privacy.

Roadmap

• Phase 1: Core Backend & RAG
• Phase 2: Mobile Interface Prototype
• Phase 3: Refinement & Polish (MVP)