LLM providers¶
Keprix routes inference through pluggable providers. Prefer the GUI on a running instance; .env / CLI remain valid for automation.
GUI (recommended for Contabo sidecar / app.keprixai.com)¶
- Sign in as admin at
https://app.keprixai.com/auth/login(or your instance). - Open Admin → Settings (
/dashboard/settings). - Tab LLM Providers → DeepSeek → paste API key → Save.
- Optionally Set as default, then Test.
- Keys persist under
KEPRIX_HOME/.env(Docker:/home/keprix/.keprix/.env) and apply immediately to the Carina/Aiva product sidecar on the same backend.
Do not put LLM keys only in the host compose .env and expect GUI edits to update that file; the GUI SoT is KEPRIX_HOME/.env (and KEPRIX_ENV_FILE when set).
Cloud providers¶
| Provider | Environment variable |
|---|---|
| OpenAI | OPENAI_API_KEY |
| Anthropic | ANTHROPIC_API_KEY |
| Google Gemini | GEMINI_API_KEY or GOOGLE_API_KEY |
| Groq | GROQ_API_KEY |
| OpenRouter | OPENROUTER_API_KEY |
| DeepSeek | DEEPSEEK_API_KEY |
| Together | TOGETHER_API_KEY |
| Mistral | MISTRAL_API_KEY |
Local models¶
| Backend | URL variable |
|---|---|
| Ollama | OLLAMA_BASE_URL (default http://host.docker.internal:11434/v1) |
| LM Studio | LM_STUDIO_URL |
| Custom OpenAI-compatible | CUSTOM_LLM_BASE_URL + CUSTOM_LLM_API_KEY |
Default selection¶
KEPRIX_DEFAULT_PROVIDER=auto
CLI:
python3 -m keprix.keprix_cli.main model
Docker note¶
Containers reach host Ollama via host.docker.internal (configured in Compose extra_hosts).