← Langdock integration

EU-resident inference via Langdock

Langdock is an EU-hosted AI platform whose completion API is OpenAI-compatible and served from EU infrastructure, with GDPR and SOC 2 compliance and bring-your-own-key (BYOK) support. In Daslab it acts as an inference route: connect a workspace key, and models run with EU data residency — nothing else about your scene changes.

Connecting

  1. Create an API key in your Langdock workspace settings (see docs.langdock.com).
  2. Add a Langdock Account asset with that key.

The key is workspace-scoped: it can only reach the models and data enabled in that Langdock workspace.

Addressing models

Prefix any workspace-enabled model with langdock/:

langdock/gpt-4o

Daslab strips the prefix on the wire and calls Langdock's EU route (https://api.langdock.com/openai/eu/v1/chat/completions) with your key as a Bearer token. There is no fixed model catalogue on the Daslab side — which models resolve is decided by your Langdock workspace configuration, so new models your workspace enables work immediately.

Rate limits

Langdock enforces 500 requests per minute and 60,000 tokens per minute, per model. Long agent runs on a single model share that budget; spreading work across models raises the effective ceiling.

What this route does and doesn't cover

  • Covered: chat completions through the OpenAI-compatible EU endpoint — the path Daslab agents use for inference.
  • Not wired up (exists on Langdock's side): their Anthropic/Google/Mistral-native routes, the assistant/agent API, knowledge folders, and usage-export/audit-log APIs. Usage and audit trails remain visible in your Langdock workspace itself.
  • Langdock's own chat UI and assistants are separate products — this integration uses only the API surface.