[drvt69talati]

Summary

• Adds llamay, a single-binary local inference server, to AI Endpoints: ai_endpoints/llamay.mdx plus a sidebar entry. The page follows the Ollama and Lemonade pages.
• Covers the custom endpoint YAML, the API-key placeholder, model fetching, and Docker networking. llamay binds to loopback by default, so the Docker section has users bind it with -api-key and put the same key in apiKey, rather than exposing an unauthenticated port.
• Adds a monochrome LlamayIcon drawn from https://llamay.com/logo.svg, registered as Llamay in lib/icons.tsx.

Disclosure: I work on llamay.

Verification

All against llamay serve -m qwen2.5:0.5b:

• GET /v1/models with Authorization: Bearer llamay lists the models on disk, so fetch: true works for an endpoint named llamay.
• POST /v1/chat/completions works with the documented settings, streaming and non-streaming, and so does a short title prompt (titleModel: "current_model").
• With -api-key, a request with no key returns 401 and a request with Authorization: Bearer <key> returns 200, which matches how LibreChat sends apiKey.
• Every CLI command on the page (pull, serve -m, models, -addr, -api-key) is checked against llamay help.

I could not run pnpm run typecheck locally. llamay.tsx has the same shape as lemonade.tsx, and the registry edits mirror the Lemonade ones from #759.

🤖 Generated with Claude Code

[vercel]

@drvt69talati is attempting to deploy a commit to the LibreChat's projects Team on Vercel.

A member of the Team first needs to authorize it.