VS Code Custom Model Clients with Verdictan
This integration applies only when the client exposes an API base URL or a complete model endpoint. A generic network proxy setting is not equivalent.
Make sure that the client has the capabilities
Before you configure the client, make sure that it can set these values:
- the endpoint or OpenAI-compatible base URL.
- the model ID.
- a bearer/API-key credential.
- the API family it sends, such as Chat Completions or Responses.
- the tool-calling or streaming capabilities necessary for the selected feature.
An extension can prevent changes to one or more of these values. In that condition, the Verdictan model API cannot route that traffic.
Gateway values
| Client field | Value |
|---|---|
| OpenAI-compatible base URL | http://127.0.0.1:41002/v1 |
| Full Chat Completions URL | http://127.0.0.1:41002/v1/chat/completions |
| Full Responses URL | http://127.0.0.1:41002/v1/responses |
| API key | A Verdictan client API token |
| Model | An ID returned by /v1/models |
Use only the URL shape that the client must use. If the client must use a full
route, /v1 can cause a misleading 404.
VS Code also has a built-in Custom Endpoint provider in Chat: Manage Language Models. See VS Code and GitHub Copilot for the setup.
Verify
Select the custom model. Run verdictan events tail --since 10m --follow. Send one
small request. Until gateway events show each call, do not claim that the
integration governs other call types. These types include inline suggestions,
utility models, embeddings, and extension telemetry.