Continue¶
Add the gateway to Continue's config.yaml and chat, edits, agent mode and autocomplete run on governed calls:
checked against your policies, charged to your team and recorded in Request logs. Give each role its own model, a
strong one for agent work and a cheap one for autocomplete, and see what each costs.
What it is¶
Continue is an open-source coding assistant for VS Code and JetBrains IDEs. Its models are listed in a config.yaml
file; each entry names a provider, a model, and optionally an apiBase. With provider: openai and the gateway's
address, any model your organisation approved can be used. Continue calls the gateway from your machine.
The two settings¶
| Setting | config.yaml key |
Value |
|---|---|---|
| Base address | apiBase |
https://<your-gateway>/v1 |
| Key | apiKey |
Your Vidai key (starts vk_), ideally as a secret: ${{ secrets.VIDAI_API_KEY }} |
For a local config.yaml, put VIDAI_API_KEY=vk_… in ~/.continue/.env and restart the editor: the extension doesn't read your shell's variables.
name: My config
version: 1.0.0
schema: v1
models:
- name: GPT-4.1 (Vidai)
provider: openai
model: gpt-4.1
apiBase: https://<your-gateway>/v1
apiKey: ${{ secrets.VIDAI_API_KEY }}
roles: [chat, edit, apply]
capabilities: [tool_use]
- name: Claude Sonnet 4.5 (Vidai)
provider: openai
model: claude-sonnet-4.5
apiBase: https://<your-gateway>/v1
apiKey: ${{ secrets.VIDAI_API_KEY }}
roles: [chat, edit, apply]
capabilities: [tool_use, image_input]
- name: Autocomplete (Vidai)
provider: openai
model: gpt-4o-mini
apiBase: https://<your-gateway>/v1
apiKey: ${{ secrets.VIDAI_API_KEY }}
roles: [autocomplete]
Continue sends the key as Authorization: Bearer. The gateway's address is under Connect your app on the Keys
page and in System › Settings › Gateway; the key comes from Keys › Issue a key (its secret is shown once).
A minimal call¶
- Save
config.yaml(Continue reloads it). - Pick GPT-4.1 (Vidai) in Continue's model picker.
- Send "Reply with exactly: ok" in the chat. A short answer means the address and key are right.
Streaming¶
Continue streams answers. With provider: openai, a stream from Claude or Gemini arrives as OpenAI chunks; Vidai
records tokens and cost either way.
Tool calls¶
Agent mode needs tool calls. Add capabilities: [tool_use] to a model you use as an agent: Continue doesn't know what a
gateway name can do, so it relies on you saying so. Tool calls travel in the OpenAI format and work on every provider
the gateway translates to.
Reading what the gateway did¶
Continue doesn't show response headers. Look the call up in Request logs: filter on Key, Model or Path. Autocomplete shows as many short calls; give it its own key if you want to see its cost separately.
Errors¶
Continue shows the gateway's error message. Common ones: 401 invalid_api_key, 403 model_not_allowed, 404
model_not_found (check model), 400 for a call to an endpoint that model's provider doesn't serve (see below), 429
(a limit on the key or team). The full list is in Headers and errors.
Model names¶
model is the name your organisation approved under Models in the Control room; name is only the label in
Continue.
What goes through the gateway¶
| Continue feature | Through the gateway |
|---|---|
| Chat, edit, apply, agent mode | Yes, to every provider (Chat Completions) |
| Autocomplete | Yes. Model names starting gpt or o use Chat Completions (every provider); other names use /v1/completions, which reaches OpenAI and Azure OpenAI models only |
Embeddings (embed role), reranking (rerank role) |
Yes, to a provider that serves them in the OpenAI format; not translated to another provider |
Models Continue reaches with /responses |
OpenAI models only (see Good to know) |
Check it works in a minute¶
Send "Reply with exactly: ok" with a gateway model selected, then open Request logs a few seconds later and filter on your key.
Good to know¶
- For model names that look like OpenAI's reasoning models (o-series,
gpt-5…), Continue uses OpenAI's Responses API by default. That works when the name is served by OpenAI; if your organisation maps such a name to another provider, adduseResponsesApi: falseto the model entry so Continue uses Chat Completions. - Autocomplete sends a call on almost every pause in typing. Give it a fast, cheap model, and a limit on its key.
- Continue's docs: config.yaml reference, OpenAI provider.