Local AI Prompt
A VS Code extension that replicates the GitHub Copilot Chat experience but runs fully offline using a local Ollama LLM — no cloud APIs, no data leaving your machine.
Features
💬 Chat
- Ask mode — chat with AI about your code; automatically uses the active file as context (or selected text if highlighted)
- Streaming responses — tokens appear as the model generates them
- Markdown rendering — formatted output with code blocks and Copy button
- Chat history — conversations are saved per workspace; use the 🕐 button to browse, switch, or delete past chats
- New Chat — start a fresh conversation without losing history
📎 Context Attach
- Auto-context — active file is automatically attached when no text is selected
- Selection context — highlight code in the editor → it auto-appears as context chip
- 📎 button — attach current file, current selection, a folder (recursive), or browse individual files
- Drag & drop — drag files from macOS Finder directly onto the chat input
- Drop zone — drag files/folders from VS Code Explorer onto the "Drop Files Here" panel
- Right-click — right-click any file/folder in Explorer → Add to AI Chat Context
📝 Plan mode
- Describe what you want to build → AI analyzes whether files are needed first (YES/NO classification)
- If files are needed: shows a plan with file names, descriptions, and reasons why each file is needed
- You review and confirm before anything is created
- If no files needed: AI answers as a normal chat message
🤖 Agent mode
- Same smart 2-step flow as Plan, but designed for building features
- Shows plan with reasons → ▶ Create all files or ✕ Cancel
- Files are only created after your explicit confirmation
🔧 Model management
- Model dropdown in header — shows installed models (green dot) vs popular available models
- One-click ollama pull for any model directly from the dropdown with live progress bar
- Connection indicator — green = Ollama reachable, red = offline
Keyboard Shortcuts
| Action |
macOS |
Windows/Linux |
| Open Chat panel |
Cmd+Shift+L |
Ctrl+Shift+L |
| Send message |
Enter |
Enter |
| New line in input |
Shift+Enter |
Shift+Enter |
Prerequisites
1. Install Ollama (official installer — NOT Homebrew)
# macOS / Linux
curl -fsSL https://ollama.com/install.sh | sh
# Windows / macOS app — download from https://ollama.com/download
⚠️ Do not use brew install ollama — the Homebrew package is missing the llama-server binary.
2. Pull a model
# Lightweight & fast (recommended for low-resource machines)
ollama pull gemma:2b
ollama pull gemma2:2b
ollama pull llama3.2:3b
# Coding-focused
ollama pull codellama:7b
ollama pull qwen2.5-coder:7b
ollama pull deepseek-coder:6.7b
# Balanced quality/speed
ollama pull mistral:7b
ollama pull llama3.1:8b
ollama pull qwen2.5:7b
3. Make sure Ollama is running
ollama serve
# If you see "address already in use" — Ollama is already running, that's fine.
Run Locally (Development)
cd local-ai-prompt
npm install
npm run compile
# Press F5 in VS Code → opens Extension Development Host
The chat panel appears in the Activity Bar on the left sidebar.
Configuration
Open Settings (Cmd+,) and search for localAIPrompt:
| Setting |
Default |
Description |
localAIPrompt.ollamaUrl |
http://localhost:11434 |
Ollama API base URL |
localAIPrompt.model |
gemma:2b |
Default model |
localAIPrompt.systemPrompt |
(coding assistant) |
System prompt sent to the model |
Project Structure
local-ai-prompt/
├── src/
│ ├── extension.ts # Activation, command registration
│ ├── ollamaClient.ts # Streaming Ollama API client (NDJSON)
│ ├── chatController.ts # Multi-conversation history (persisted per workspace)
│ ├── webviewProvider.ts # Webview lifecycle, message routing, context handling
│ └── fileDropProvider.ts # TreeView drop target for VS Code Explorer drag
├── media/
│ ├── webview.html # Full chat UI (HTML + CSS + vanilla JS)
│ └── icon.svg # Activity Bar icon
├── package.json
├── tsconfig.json
└── README.md
Architecture
VS Code Extension Host
┌──────────────────────────────────────────┐
│ extension.ts │ ← commands, Explorer right-click
│ chatController.ts │ ← multi-conversation history (workspaceState)
│ ollamaClient.ts ───────────────────────┼──► POST /api/chat (streaming NDJSON)
│ webviewProvider.ts │ ← context detection, file attach, agent logic
│ fileDropProvider.ts │ ← TreeView DnD from Explorer
│ │ postMessage / onMessage │
│ ▼ │
│ media/webview.html │ ← Copilot-like chat UI
│ ├─ History sidebar │
│ ├─ Context chips (file/selection) │
│ ├─ Model dropdown + pull progress │
│ ├─ Ask / Plan / Agent modes │
│ └─ Streaming markdown renderer │
└──────────────────────────────────────────┘
│
http://localhost:11434
│
Ollama (local LLM)
Package as VSIX
npm install -g @vscode/vsce
vsce package
# Produces: local-ai-prompt-0.1.0.vsix
Install: Extensions view → ... → Install from VSIX…
Troubleshooting
| Error |
Fix |
llama-server binary not found |
Reinstall Ollama via official installer (not Homebrew) |
address already in use |
Ollama already running — no action needed |
model 'xxx' not found |
Run ollama pull <model-name> first |
| Red connection dot |
Run ollama serve |
| Agent always creates files |
Use a larger model (7B+); small models may not classify intent accurately |
License
MIT
