Getting Started
Configure Models
Connect DeepSeek API, other providers and custom OpenAI-compatible endpoints
Configure in the Web UI
Open Settings → Models:
- Enter your DeepSeek API key;
- After saving, model routing takes effect immediately — no server restart needed.
Custom endpoints
The models page lets you add a custom provider with any OpenAI-compatible Base URL and key, so other model services can be connected as well.
Environment variables (CLI / headless / automation)
Command-line and automation scenarios read model configuration from environment variables:
| Variable | Purpose |
|---|---|
DEEPSEEK_API_KEY | DeepSeek API key |
DEEPSEEK_BASE_URL | Optional: custom Base URL |
Headless mode (dsh --profile headless "task") and the repository's automated tests rely on these variables; without a key, related tasks skip or fail.
Key safety
- Never commit keys: keep
.envand API keys out of your repository. - Prefer environment variables or Web UI settings over hardcoding keys in config files.
- Hide keys before screen sharing or recording.
Common questions
- Model not responding: check the key validity and network access to the API endpoint (see troubleshooting).
- Switching models: change it under Settings → Models; existing session history is unaffected.