Built-in Extensions

Overview of the built-in extensions that provide core functionality without additional dependencies

To minimize bloat and ensure a lean functional core, llms .py:

Maintains a single file main.py core which enables:
- Its CLI Features
- Support for Open AI Compatible providers:
  - Moonshot AI, Z.ai, Alibaba/Qwen, Groq, xAI, Mistral, Codestral, NVIDIA, GitHub, DeepSeek, Chutes, HuggingFace, OpenRouter, Fireworks AI, Ollama, LM Studio
Requires only a single aiohttp dependency
Maintains an up-to-date model provider configuration (from models.dev)
Supports loading extensions for any additional functionality, integrations, and UI features

All built-in extensions are located in the llms/extensions folder. They're resolved for generally useful functionality that does not require any additional dependencies.

app extension

Provides core application logic, data persistence (SQLite), and migration from client-side storage

Database: ~/.llms/user/default/app/app.sqlite
- request: Logs all requests for auditing and analytics
- thread: Persists chat threads and messages
analytics: Provides a UI for analytics - using request table data in app.sqlite

gallery extension

A UI and backend for intercepting, storing, and viewing generated assets and uploaded files

Database: ~/.llms/user/default/gallery/gallery.sqlite
- media: Stores metadata for generated assets and uploaded files

providers extension

Provides support for popular LLM providers requiring custom handling beyond OpenAI-compatible requests:

anthropic: Integration with Anthropic's Claude models
cerebras: Integration with Cerebras AI models
chutes: Integration with Chutes.ai for image generation
google: Integration with Google's Gemini models
nvidia: Integration with NVIDIA's GenAI APIs
openai: Integration with OpenAI's chat and image generation models
openrouter: Integration with OpenRouter for image generation
z.ai: Integration with Z.ai for image generation

Only built-in extensions or extensions in your local ~/.llms/extensions folder are loaded by default. To disable built-in extensions or temporarily disable local extensions, add their folder names to your ~/.llms/llms.json:

{
  "disable_extensions": [
    "xmas",
    "duckduckgo"
  ]
}

Alternatively you can set the LLMS_DISABLE environment variable with a comma-separated list of extension names to disable.

export LLMS_DISABLE="xmas,duckduckgo"

Other Built-in Extensions

core_tools - Essential tools for file operations, memory persistence, math calculation & safe code execution
katex - Enables rendering of LaTeX math expressions in chat responses using KaTeX
pdf - PDF Studio for code-first PDF design using Typst, real-time preview & AI assistance
system_prompts - Configures a library of over 200+ curated system prompts accessible from the UI
tools - Provides Tools UI and Exposes registered tool definitions via an API endpoint

share_static: the default Folder destination for static project exports. Requires no publisher account; exports under ./p/<user>/<project>/ in llms.py.
share_llmspy: the ai.llmspy.org destination for public threads, projects, media, and Decision Studio recipes. Requires a connected publisher account to publish.

They run independently and use the core Share panel. Folder appears only with a selected project or a project chat; the Share icon is hidden when no destinations are available. Add either or both names to disable_extensions to disable them. See sharing configuration and ctx.setShareOptions for custom tabs.

Workspace and Decision Extensions

projects: Git-backed creation, ordering, and archiving, plus the read-only file explorer.
git: Diffs, staging, commits, stashes, and remote synchronization. Requires the Git executable and disables itself when it is unavailable.
jev: Decision Studio, with portable recipes, recorded examples, history, public sharing, and collection imports. Running decisions requires OpenRouter.
openai_auth: Per-user ChatGPT sign-in and eligible subscription chat.
gemini: Gemini RAG, including portable saved imports and website crawls.

These ship with llms.py and require no separate extension install or additional Python framework.

Built-in Extensions