Skip to content

AnythingLLM: using the VIDAI Control Plane as the backend

AnythingLLM's Generic OpenAI LLM provider takes a base URL and API key. Configure it once and every workspace, embedding call, and agent invocation on your AnythingLLM deployment runs through the control plane.

TL;DR

  1. Settings → LLM Preference → Generic OpenAI.
  2. Fill in:
  3. Base URL: https://your-vidai-server.example.com/v1
  4. API Key: your VIDAI API key
  5. Chat Model Name: a model registered on the control plane (e.g. gpt-4o-mini)
  6. Token context window: whatever fits the model (e.g. 128000)
  7. Max Tokens: the response cap (e.g. 4096)
  8. Save. Test with a new chat.

Prerequisites

  • AnythingLLM installed (desktop, Docker, or cloud-hosted).
  • Control plane base URL and an API key from API Keys.
  • A model registered on the Models page.

Per-workspace overrides

AnythingLLM's per-workspace LLM override lets different workspaces use different models. Pick Generic OpenAI on each workspace and put a different registered model name in — each workspace's traffic flows through the control plane on the same key (or different keys if you split per team).

Agents

AnythingLLM's agent skills (search, scrape, etc.) call the configured LLM for their reasoning. Every skill invocation is a control-plane request that appears on Request Logs.

Attribution

One VIDAI API key per AnythingLLM instance is the default. For per-team splits, deploy separate AnythingLLM instances (or use the workspace-level LLM override with different keys).

Verify it works

Open a new chat in any workspace and send:

reply with exactly: ok

A row appears on Request Logs.

If something's off

Raise an issue at github.com/vidaiUK/vidai-quickstart/issues with the LLM Preference config (masking the key). We'll get it sorted.

Where to go next