A fully local AI email assistant powered by Llama 3.2 (1B) via Ollama. No paid APIs. No internet needed at runtime. Runs on 8 GB RAM.
How many of us have sent an email we immediately wished we could rewrite? Or received a long message and thought — I just need the key points from this. That is exactly the problem this app solves.
The AI Email Assistant rewrites, summarizes, extracts action items, replies, translates, and searches your past emails — all running privately on your own machine, with no subscriptions, no data leaving your device, and no cost.
| Feature | Description |
|---|---|
| ✍️ Rewrite | Turns a casual draft into a professional email |
| 📋 Summarize | Compresses long emails into 3-5 bullet points |
| ✅ Extract | Pulls out tasks, deadlines, names, and decisions |
| 💡 Improve Clarity | Coaching feedback plus an improved version |
| ↩️ Reply Email | Drafts a professional reply to any incoming email |
| 🌐 Translate Email | Auto-detects language and translates to/from English or 19 other languages |
| 🗂️ Search History | Finds past emails by meaning using vector search |
| Layer | Tool |
|---|---|
| AI Model | Llama 3.2:1b via Ollama (local, free) |
| UI | Streamlit |
| Vector Database | ChromaDB (embedded) |
| Embeddings | all-MiniLM-L6-v2 (SentenceTransformers) |
| Package Manager | uv |
| Language | Python 3.10+ |
# Windows (PowerShell)
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"Download Ollama from https://ollama.com then run:
ollama pull llama3.2:1bgit clone https://github.com/YOUR_USERNAME/EmailAssistant_AI.git
cd EmailAssistant_AIuv sync# Terminal 1 — start Ollama
ollama serve
# Terminal 2 — start the app
uv run streamlit run app.pyOpen your browser at http://localhost:8501
| File | Description |
|---|---|
| PLAN.md | Problem thinking and design decisions |
| ARCHITECTURE.md | System design and data flow |
| docs/proposal.md | Capstone project proposal |
| docs/final_documentation.md | Final documentation |