Collama

AI-powered code completion and editing for VS Code using local LLM backends

Features • Installation • Configuration • Models • Contributing

---

Overview

Collama is a VS Code extension that uses local LLM backends to get code completions, refactoring suggestions, and documentation generation — all running privately on your machine with no external API calls. Supports Ollama and OpenAI-compatible API (verified only vLLM).

Status: This project is in heavy active development. For sure there will be a lot of strange output. If you have any ideas to improve the quality just let me know and contribute!

Features

Code Completion

• Inline, multiline, and multiblock (more a "fun" feature) suggestions
• Uses currently opened tabs as context

Code Edits

• Generate docstrings and documentation
• Extract functions and refactor code
• Simplify complex code
• Fix syntax errors
• Manual instructions

Chat Interface

• Chat with multiple sessions - organize conversations by topic
• Send selected code or files to chat as context
• Context is automatically attached to messages with file reference and line numbers
• Real-time context usage bar showing token consumption vs model's context window
• Automatic context trimming — when conversations exceed the context window, older messages are removed from the LLM context while remaining visible in the chat
• Visual indicators for messages no longer included in the LLM context

AI Agent with Tool Calling

• LLM can interact with your workspace through function calling
• File System Tools: Read files, list directories, search contents with regex, create files and folders, edit files with diff preview
• Git Tools: List commits or branches, view diffs (working tree or compare commits/branches)
• Code Analysis Tools: Get diagnostics (errors, warnings, hints) from the language server
• Security: Path traversal protection, workspace boundary enforcement, .gitignore integration
• Real-time Feedback: Tool calls streamed to chat as they execute
• The agent can autonomously explore your codebase to provide context-aware assistance
• Read-Only Mode: Toggle edit tools to run the agent in safe, read-only mode

Commit Messages

• AI-generated conventional commit messages from staged changes
• Analyzes git diff to create meaningful commit descriptions
• Accessible via command palette or Source Control view

Context Management

• Automatic detection of the model's context window size (Ollama and OpenAI)
• Prompts that exceed the context window are blocked with a clear notification
• Open editor tabs are used as additional context for completions, with smart pruning to fit the context window

Quick Start

Prerequisites

• VS Code 1.109.0 or higher
• Ollama running locally (or accessible on your network), or any OpenAI-compatible API
• A supported code model (see Models)

Installation

Use the marketplace to install the extension or build the vsix yourself. Furthermore you need an ollama instance in your local network.
See this link how to install ollama or this link for the docker image.

Alternatively, you can use vLLM (tested). Point the endpoint settings to your server and the backend is auto-detected. If authentication is required, see Bearer Tokens.

Configuration

Configure Collama via VS Code Settings (Preferences → Settings, search "collama"):

Setting	Type	Default	Description
collama.apiEndpointCompletion	string	http://127.0.0.1:11434	Endpoint for code auto-completion
collama.apiEndpointInstruct	string	http://127.0.0.1:11434	Endpoint for code edits/chat
collama.apiModelCompletion	string	qwen2.5-coder:3b	Model for code completions
collama.apiModelInstruct	string	gpt-oss:20b	Model for code edits (use instruct/base variant)
collama.apiTokenContextLenCompletion	number	4096	Context window size (tokens) for the completion model
collama.apiTokenContextLenInstruct	number	4096	Context window size (tokens) for the instruct/chat model
collama.apiTokenPredictCompletion	number	400	Max tokens to generate per completion request
collama.apiTokenPredictInstruct	number	4096	Max tokens to generate per instruct/chat request
collama.autoComplete	boolean	true	Enable auto-suggestions
collama.suggestMode	string	inline	Suggestion style: inline, multiline, or multiblock
collama.suggestDelay	number	1500	Delay (ms) before requesting completion
collama.agentic	boolean	true	Use agentic (tool use) mode for chat
collama.enableEditTools	boolean	true	Enable edit tools (read-only mode when disabled)
collama.tlsRejectUnauthorized	boolean	true	Verify TLS certificates for HTTPS endpoints

Manual Token Settings

You have to manually adjust the token limits for context and prediction. The values are expressed in tokens, not characters.

• collama.apiTokenContextLenCompletion / collama.apiTokenContextLenInstruct – maximum number of tokens that can be sent to the model as context.
• collama.apiTokenPredictCompletion / collama.apiTokenPredictInstruct – maximum number of tokens the model can generate in a single request.

Changing these values will automatically trigger a re‑calculation of the context window and may affect the real‑time context usage bar shown in the chat view.

Note

Check the max context window of your model online.

Bearer Tokens (API Key)

If your API endpoints require authentication (e.g. vLLM with --api-key, or a reverse proxy), you can securely store bearer tokens using VS Code's encrypted credential storage. Tokens are sent as Authorization: Bearer <token> headers with every request.

1. Open the Command Palette (Ctrl+Shift+P / Cmd+Shift+P)
2. Run one of these commands:
   • collama: Set Bearer Token (Completion) - for the completion endpoint
   • collama: Set Bearer Token (Instruct) - for the instruct/chat endpoint
3. Enter your token in the password input (characters will be masked)
4. The token is stored encrypted in your system's credential manager

Note: Tokens are stored securely and never appear in plain text in your settings or configuration files. To clear a token, run the same command and leave the input empty.

Models

Recommended Models

Collama is tested primarily with the Qwen Coder for Completion and *gpt-oss for Instruction.

For Code Completion (FIM - Fill In Middle)

• any qwen coder > 3b recommended
• qwen2.5-coder:7b (stable quality)

For Code Edits (Instruct/Base Models)

• any instruct model and thinking recommended
• gpt-oss:20b (stable quality) Do not use a FIM model for instructions. It will produce very poor quality answers.

Model Completion Compatibility Table

Model	Tested Sizes	FIM Support	Status	Notes
qwen2.5-coder	1.5B, 3B, 7B	✅	Stable	Recommended for most use cases
qwen3-coder	30B	✅	Stable	Excellent quality, higher resource usage
starcoder	—	⚠️	Untested	May work; contributions welcome
starcoder2	3B	✅	Stable	Improved over v1
codellama	7B, 13B	⚠️	Limited	Limited file context support; FIM is ok
codeqwen	—	⚠️	Untested	May work; contributions welcome

Note: Models are tested primarily with quantization level q4. Results may vary with other quantization levels.

Note: ChatML format is not supported - that means only true fim-models will work for autocomplete!

Usage

Code Completion

1. Trigger Completion: Use editor.action.inlineSuggest.trigger (default keybinding varies by OS)
   • Set custom keybinding: Alt + S or Ctrl + NumPad 1 (example)
2. Auto-Trigger: Completions trigger automatically after 1.5 seconds of inactivity (configurable via suggestDelay)
3. Accept: Press Tab to accept suggestion, Esc to dismiss

Code Edits

1. Select code in the editor
2. Right-click → collama (on selection) and choose:
   • Write Docstring - Generate documentation
   • Extract Functions - Refactor into separate functions
   • Simplify Code - Improve readability and efficiency
   • Fix Syntax - Correct syntax errors
   • Edit (manual) - Custom AI-assisted edits

Chat Interface

1. Open the Chat view in the sidebar (Collama icon)
2. Right-click on selected code in editor → Send to Chat
3. Type your message - the context is automatically attached with:
   • File name and path
   • Line number references
   • Selected code or full file content
4. Monitor token usage with the real-time context bar (shows usage vs. model's max context)
5. Create multiple chat sessions to organize conversations by topic

Commit Message Generation

1. Stage your changes in Git (using Source Control view or git add)
2. Open Command Palette (Ctrl+Shift+P / Cmd+Shift+P)
3. Run collama: Generate Commit Message
4. The AI analyzes your staged diff and generates a conventional commit message
5. Review and edit the message in the Source Control input box before committing

AI Agent Usage

Important

It is recommended to turn off agentic-mode for small local/home models (like gpt-oss:20b) and use them in chat-only mode instead.

Agentic was tested on local hosted vLLM (Openai SDK) with

• gpt-oss:120b
• glm-4.7-fp8
• minimax2.5

Available Tools:

• File System Tools

  • readFile - Read the contents of a file in the workspace (supports optional line ranges)
  • searchFiles - Search file contents for a regex pattern (supports optional glob pattern to filter files)
  • lsPath - List files and folders in a workspace directory (supports recursion depth and glob pattern filtering)
  • create - Create a new file or folder (with content: creates a file with preview; without content: creates a folder)
  • editFile - Edit a file by replacing an exact string match with new content (shows diff preview and asks for confirmation)
  • deleteFile - Delete a file from the workspace (asks for user confirmation)

• Git Tools

  • gitLog - Git log info (list commits or list branches with optional filters)
  • gitDiff - Get a git diff (working tree changes or compare commits/branches)

• Code Analysis Tools

  • getDiagnostics - Get diagnostics (errors, warnings, hints) from the language server

Read-Only Mode:

When enableEditTools is disabled (toggle via status bar), the following tools are unavailable:

• editFile, create, deleteFile

The agent can still use read-only tools to explore and analyze your codebase safely.

Contributing

Contributions are welcome! Here's how you can help:

1. Report Issues Open an issue
2. Submit PRs: Fork, create a feature branch, and submit a pull request