`deepinfra://`¶

DeepInfra — serverless hosted open-source models via an OpenAI-compatible API. Pay per token, broad catalogue covering Llama, Mixtral, Phi, Qwen, DeepSeek, etc.

Built on pkg/component/llmutil/ — same chat/embeddings shape as openai://. This page covers only the DeepInfra-specific defaults.

URI shape¶

deepinfra://chat?model=<id>&apiKey=<key>&temperature=<float>&maxTokens=<int>&baseURL=<url>
deepinfra://embeddings?model=<id>&apiKey=<key>

Parameters¶

Param	Type	Default	Producer/Consumer/Both	Description
path	enum	required	Producer	`chat` or `embeddings`.
`model`	string	`meta-llama/Meta-Llama-3.1-70B-Instruct`	Producer	DeepInfra model id — typically `<org>/<model>`. Browse https://deepinfra.com/models.
`apiKey`	string	(from constructor)	Producer	DeepInfra API token. Sent as `Authorization: Bearer <key>`.
`baseURL`	string	`https://api.deepinfra.com/v1/openai`	Producer	Override for proxies. Note the URL ends in `/openai` — DeepInfra exposes the OpenAI-compatible subset under that prefix.
`temperature`	float	`0.7`	Producer	Sampling temperature.
`maxTokens`	int	`4096`	Producer	Output cap.

Producer-only. Default timeout 120 s.

Headers¶

Namespaced under deepinfra.* (deepinfra.model, deepinfra.id, deepinfra.finishReason, deepinfra.usage.*). Input header systemPrompt works the same way.

Examples¶

- to: deepinfra://chat?model=meta-llama/Meta-Llama-3.1-70B-Instruct&apiKey=${DEEPINFRA_API_KEY}

DeepSeek for code:

- to: deepinfra://chat?model=deepseek-ai/DeepSeek-V3&apiKey=${DEEPINFRA_API_KEY}

Embeddings:

- to: deepinfra://embeddings?model=BAAI/bge-large-en-v1.5&apiKey=${DEEPINFRA_API_KEY}

Behavior notes¶

See openai://. The /v1/openai suffix on baseURL is the DeepInfra-specific quirk to keep in mind if you ever override the base URL.

Source¶

pkg/component/deepinfra/deepinfra.go.
pkg/component/llmutil/llmutil.go.

deepinfra://¶