LibreChat: using the VIDAI Control Plane as the backend¶
LibreChat's librechat.yaml supports custom endpoints. Add one
pointing at the control plane and every user on your LibreChat
deployment can pick your VIDAI models from the endpoint dropdown.
TL;DR¶
Add to librechat.yaml:
endpoints:
custom:
- name: "VIDAI Control Plane"
apiKey: "${VIDAI_API_KEY}"
baseURL: "https://your-vidai-server.example.com/v1"
models:
default: ["gpt-4o-mini", "claude-sonnet-4.5", "claude-haiku-4-5"]
fetch: false
titleConvo: true
titleModel: "gpt-4o-mini"
modelDisplayLabel: "VIDAI"
Set VIDAI_API_KEY in your .env and restart LibreChat. The
endpoint appears in the top-of-conversation picker.
Prerequisites¶
- LibreChat deployed (Docker Compose, Helm, or bare install).
- Control plane base URL and an API key from API Keys.
- Models registered on the Models page — the
names go in
models.default.
Attribution¶
Every LibreChat user's conversation authenticates with the
single VIDAI_API_KEY env var. Per-user attribution requires
either:
- One custom endpoint per team in
librechat.yaml, each with its own key — teams appear as separate rows on Chargeback. - User-provided keys: set
userProvide: trueon the endpoint so each LibreChat user pastes their own VIDAI key in Settings → API Keys — clean per-user attribution.
endpoints:
custom:
- name: "VIDAI Control Plane"
apiKey: "user_provided" # each user brings their own
baseURL: "https://your-vidai-server.example.com/v1"
models:
default: ["gpt-4o-mini", "claude-sonnet-4.5"]
Multiple endpoints¶
Register several custom endpoints for different model families or different teams:
endpoints:
custom:
- name: "VIDAI · General"
apiKey: "${VIDAI_GENERAL_KEY}"
baseURL: "https://your-vidai-server.example.com/v1"
models: { default: ["gpt-4o-mini"] }
- name: "VIDAI · Compliance"
apiKey: "${VIDAI_COMPLIANCE_KEY}"
baseURL: "https://your-vidai-server.example.com/v1"
models: { default: ["claude-sonnet-4.5"] }
Each shows as a separate option in the endpoint dropdown.
Verify it works¶
Open LibreChat, pick VIDAI Control Plane, choose a model, and send:
A row appears on Request Logs.
If something's off¶
Raise an issue at github.com/vidaiUK/vidai-quickstart/issues
with the librechat.yaml endpoint block (masking keys). We'll
get it sorted.
Where to go next¶
- Client integrations overview
- API Keys — one per team or user
- Chargeback
- Guardrails — apply policies across LibreChat traffic