LocalForge

Chat with any major AI API without leaving VS Code: local Ollama models, OpenAI, OpenRouter, Anthropic Claude, Google Gemini, or any OpenAI-compatible server (LM Studio, llama.cpp, vLLM, Groq, Together, DeepSeek…).

What's new in 1.1.0

• New name: LocalForge — new display name, sidebar title, and LocalForge: … command titles (command IDs and settings keys are unchanged, so nothing breaks).
• Multi-provider chat — one dropdown switches between Ollama, OpenAI, OpenRouter, Anthropic, Gemini, and custom OpenAI-compatible endpoints, with per-provider model lists.
• Secure API keys — keys are stored in VS Code Secret Storage via LocalForge: Set API Key, never written to settings files.
• Stop generation — cancel a streaming reply mid-flight from the stop button or the LocalForge: Stop Generation command.
• Hardened webview — strict Content-Security-Policy, zero CDN/network dependencies (fully offline-safe), XSS-safe Markdown rendering, toasts instead of alert() popups, and fixed copy/insert buttons that no longer break on quotes or backticks.
• Smarter UX — connection-status banner with actionable errors, preferred-model auto-select, token-speed stats under every reply, chat persistence per workspace, one-click transcript export, and Ollama-only actions (pull/delete) cleanly hidden for cloud providers.

Features

• Sidebar chat panel — always one click away in the activity bar, themed to match your VS Code theme.

• Live streaming replies with tokens/sec, total time, and prompt-token stats.

  

• Editor actions — select code, right-click, and choose:

  • AI: Explain Code
  • AI: Find Bugs & Optimize
  • AI: Generate Unit Tests
  • AI: Write Documentation

  The analysis streams into the sidebar; hover any code block to Copy or Insert it at your cursor.

• Model manager —

  • Ollama: lists installed models (size, parameter count, format), pulls new ones with a live progress bar, deletes with confirmation.
  • Cloud providers: lists available models with a one-click Use shortcut.

  

• Tunable per chat — provider, model, system prompt, temperature, context window, max output tokens, and base URL, all from the settings drawer.

• Private by default — Ollama runs 100% locally; cloud keys stay in your OS keychain via Secret Storage.

Requirements

• VS Code ^1.96.0.
• For local models: Ollama running (ollama pull llama3.2 to get started).
• For cloud providers: an API key (OpenAI, OpenRouter, Anthropic, or Google AI Studio).

Installation

• Marketplace: search LocalForge in the Extensions view (Ctrl+Shift+X), or install from the Marketplace page.
• From .vsix: code --install-extension ollama-vscode-chat-1.1.0.vsix, or Extensions → … → Install from VSIX.

Quick Start

Option A — Local, free, private (Ollama)

1. Install Ollama: ollama.com/download, or curl -fsSL https://ollama.com/install.sh | sh.
2. Pull a model: ollama pull llama3.2.
3. Click the LocalForge activity-bar icon → provider Ollama → pick the model → chat.

Option B — Cloud APIs

1. Run LocalForge: Set API Key from the Command Palette (Ctrl+Shift+P) and paste your key (stored securely; repeat per provider as needed — the key is matched to the selected provider's service).
2. Select the provider in the sidebar dropdown, then pick a model:

Option C — Custom OpenAI-compatible server

1. Choose provider Custom OpenAI API.
2. Set the base URL, e.g. LM Studio http://localhost:1234/v1 or llama.cpp http://localhost:8080/v1 (add a key only if your server requires one).
3. Pick a loaded model and chat. Works with vLLM, Groq, Together, DeepSeek, and friends.

Settings (File → Preferences → Settings, search ollama)

| Setting | Default | Description | |---|---|---| | ollama.provider | ollama | ollama · openai · openrouter · anthropic · gemini · openai-compatible | | ollama.apiUrl | http://localhost:11434 | Base URL (http://localhost:1234/v1 for LM Studio-style servers) | | ollama.apiKey | | Optional fallback key field — the `LocalForge: Set API Key` command (Secret Storage) is recommended | | `ollama.defaultModel` | | Preferred model, auto-selected whenever it appears in the list | | ollama.systemPrompt | You are a helpful, expert AI software developer… | System instructions sent with every chat | | ollama.temperature | 0.7 | Creativity/randomness (0–2) | | ollama.contextLength | 4096 | Ollama num_ctx window | | ollama.maxTokens | 2048 | Max output tokens for cloud APIs |

Commands

• LocalForge: Open Chat Panel / LocalForge: Focus Chat Sidebar
• LocalForge: Set API Key (secure) / LocalForge: Clear API Key
• LocalForge: Stop Generation
• AI: Explain Code · AI: Find Bugs & Optimize · AI: Generate Unit Tests · AI: Write Documentation (editor right-click with a selection)

Troubleshooting

Privacy

• Ollama chats never leave your machine.
• Cloud chats go directly from VS Code to the selected provider — this extension runs no proxy and collects no telemetry.
• API keys live in VS Code Secret Storage (OS keychain), not in settings or workspace files.

Changelog

See CHANGELOG.md. Highlights: 1.1.0 LocalForge rename + docs refresh · 1.0.1 rename release · 1.0.0 multi-provider rewrite + security hardening.

Publishing a new version (maintainers)

npm run compile && npm run lint && npm test
npx vsce package        # produces ollama-vscode-chat-<version>.vsix
npx vsce publish        # needs a Marketplace PAT with Manage scope

Get a token at aka.ms/vscodepat (organization Vaibhavrathod), then npx vsce login Vaibhavrathod && npx vsce publish.

License

MIT — see LICENSE.