Providers
Hosted Providers
Managed provider and gateway setup for Facio.
Hosted providers cover direct model APIs, gateways, and subscription routers. Most use an API key plus an optional default API base from the provider registry.
Provider reference
| Provider | Config key | Key name | Default API base | Notes |
|---|---|---|---|---|
| Anthropic | anthropic | ANTHROPIC_API_KEY | Native SDK default | Routes Claude models and supports prompt caching. |
| OpenAI | openai | OPENAI_API_KEY | SDK default | Routes GPT/OpenAI models and supports max-completion-token style models. |
| OpenRouter | openrouter | OPENROUTER_API_KEY | https://openrouter.ai/api/v1 | Gateway provider; routes many model families. |
| Azure OpenAI | azure_openai | Custom key | Required apiBase | Model is the deployment name. Requires key and Azure endpoint. |
| AWS Bedrock | bedrock | AWS_BEARER_TOKEN_BEDROCK or AWS auth | Native Bedrock runtime | Uses Bedrock Converse through AWS configuration. |
| Hugging Face | huggingface | HF_TOKEN | https://router.huggingface.co/v1 | OpenAI-compatible inference router. |
| BytePlus | byteplus | BYTEPLUS_API_KEY | https://ark.ap-southeast.bytepluses.com/api/v3 | OpenAI-compatible VolcEngine international endpoint. |
| BytePlus Coding Plan | byteplus_coding_plan | BYTEPLUS_CODING_PLAN_API_KEY | https://ark.ap-southeast.bytepluses.com/api/coding/v3 | Coding-plan gateway, strips provider prefix before sending. |
| OpenCode Go | opencode_go | OPENCODE_GO_API_KEY | https://opencode.ai/zen/go/v1 | Subscription gateway with a models endpoint at /models. |
| DeepSeek | deepseek | DEEPSEEK_API_KEY | https://api.deepseek.com | OpenAI-compatible. |
| Gemini | gemini | GEMINI_API_KEY | https://generativelanguage.googleapis.com/v1beta/openai/ | Google's OpenAI-compatible endpoint. |
| Zhipu AI | zhipu | ZAI_API_KEY | https://open.bigmodel.cn/api/paas/v4 | Also exposes ZHIPUAI_API_KEY to provider tooling. |
| DashScope | dashscope | DASHSCOPE_API_KEY | https://dashscope.aliyuncs.com/compatible-mode/v1 | Qwen-compatible routing with provider thinking options. |
| Moonshot | moonshot | MOONSHOT_API_KEY | https://api.moonshot.ai/v1 | Kimi models use provider-specific temperature behavior. |
| MiniMax | minimax | MINIMAX_API_KEY | https://api.minimax.io/v1 | OpenAI-compatible. |
| Mistral | mistral | MISTRAL_API_KEY | https://api.mistral.ai/v1 | OpenAI-compatible. |
| Groq | groq | GROQ_API_KEY | https://api.groq.com/openai/v1 | Often used for transcription; can also route LLM calls. |
Quickstart envs
The Docker quickstart maps these simple host env names into Facio provider config:
Host .env | Facio config target |
|---|---|
OPENROUTER_API_KEY | providers.openrouter.apiKey |
ANTHROPIC_API_KEY | providers.anthropic.apiKey |
OPENAI_API_KEY | providers.openai.apiKey |
FACIO_MODEL | agents.defaults.model |
Other providers can be configured through Placet settings, config.json, the management API, or nested FACIO_... config environment variables.
Provider selection examples
| Goal | Example setting |
|---|---|
| Use Anthropic by model name | model: anthropic/claude-opus-4-7, provider: auto |
| Force OpenRouter | provider: openrouter, model can be an OpenRouter-routable model ID. |
| Use Azure deployment | provider: azure_openai, model equals the Azure deployment name, apiBase is the Azure endpoint. |
| Use OpenCode Go | provider: opencode_go, apiKey set, model selected from OpenCode Go's model catalog. |
Subagent and media routes
Facio can use separate defaults for different work classes:
| Setting | Purpose |
|---|---|
subagent_model / subagent_provider | Background research or coding helpers. |
vision_model / vision_provider | Fallback when images need a vision-capable model. |
image_model / image_provider | Default model for image generation tools. |
video_model / video_provider | Default model for video generation tools. |
Keep these empty until the main route works; then add overrides for cost, capability, or latency reasons.