Configuration Reference

Open Settings to configure the active Larkup workspace. Changes apply to Chat, indexing, and the generated server.

​

AI Models

Use Settings → AI Models to choose the provider and model for embeddings, chat, and vision. The embedding model turns content and queries into vectors; changing it requires indexing existing content again. The chat model answers from retrieved context and can use a different provider than embeddings. The vision model powers video, image, and OCR tools and can also differ from the chat model.

​

Storage and retrieval

Use Settings → Storage to choose Semantic, Lexical, or Hybrid retrieval and the storage engine. LanceDB runs from a local path, LanceDB Cloud, or S3 compatible storage; use a persistent remote mode before deploying to a serverless platform.

​

General

Use Settings → General to name the workspace and set its theme. Web search is configured separately under Settings → Search & Scraping, covered in Web Crawl at Scale.

​

Installed tool settings

Marketplace tools add their own settings under Settings → Installed Tools.

​

Agent tools

Use Settings → Agent Customization to control the prompt, tools, skills, MCP connections, and sandbox the assistant can use. See Agent Server API.

​

Server configuration

After configuring models, storage, and tools, open Settings → Larkup Server to launch or deploy the API. See Deploy Your Server.