Configure an AI provider
DevOpsPad’s AI assistant works with any OpenAI-compatible Chat Completions API, hosted or local. Open Settings, enter the API base URL, the Model ID and, if your provider needs one, an API key, then click Save provider. There is no built-in model and no DevOpsPad account: requests go straight from your machine to the provider you configure.
The assistant only suggests queries. It never runs anything. See Insert an AI suggestion into the editor.
Set up the provider
- Click Settings in the left rail.
- In API base URL, enter the provider’s base URL, without
/chat/completions. The field starts withhttps://api.openai.com/v1. DevOpsPad appends/chat/completions. - In Model, enter the model ID exactly as your provider names it. DevOpsPad does not pick or list models.
- In API key, paste your key. It is optional for local providers.
- Click Save provider.
Then open the assistant with AI assistant in the header (or the AI button in the rail) and ask a question.
Rules for the base URL
- It must use HTTPS. Plain HTTP is accepted only for loopback addresses:
localhost,127.0.0.1and::1. See Use a local AI model. - It cannot contain a user name or password, query parameters or a
#fragment. - Redirects are not followed. Use the final URL.
If the URL is invalid, saving fails with “Enter a valid API base URL without credentials or query parameters” or “Use HTTPS, or HTTP for a local AI provider”.
How the API key is handled
The key is stored in the OS keychain, never in workspace.json. After saving, the field shows “Saved in keychain · leave blank to keep”. Leave it blank to keep the saved key, or type a new one to replace it.
If you change the base URL and do not enter a new key, DevOpsPad deletes the old key. This prevents one provider’s key from being sent to another provider’s URL.
Request details
- Endpoint:
POST <base URL>/chat/completionswithstream: false. - The key is sent as
Authorization: Bearer <key>when one is saved. - Timeout: 60 seconds. Responses larger than 2 MiB are rejected.
- Prompts are limited to 16,000 characters.
Troubleshooting
| Message | What to check |
|---|---|
| Configure an AI provider in Settings first | No valid provider is saved yet |
| Could not reach the AI provider | Network, VPN, base URL, local server running |
| AI provider returned HTTP 401 / 404 / … | API key, model ID, or a base URL missing /v1 |
| AI provider returned an empty response | The model produced no content; try again or another model |
| Cannot read AI key from OS keychain | Keychain access was denied or the item was removed |
From the assistant’s error message, Configure provider opens Settings directly.
Limits
No streaming responses, no provider presets, and each prompt is sent on its own without earlier chat messages. See What DevOpsPad sends to the AI provider.
Frequently asked questions
Which AI providers work with DevOpsPad?
Any provider that implements the OpenAI-compatible /chat/completions endpoint, including local servers. You enter the base URL and the exact model ID your provider uses.
Where is my AI API key stored?
In your operating system's keychain, under the devopspad service. It is never written to the workspace file and is not shown again after saving.