> ## Documentation Index > Fetch the complete documentation index at: https://docs.comfy.org/llms.txt > Use this file to discover all available pages before exploring further. # Use Chat Completions with Comfy Router > Call openrouter/chat-completions through Comfy Router: endpoint, request shape and the response Router returns. API Reference for `openrouter/chat-completions`, served by Comfy Router from Openrouter. ## Request setup Create a key in [your Comfy workspace](https://platform.comfy.org/profile/api-keys) and export it as `COMFY_API_KEY`. For Python, run `pip install comfy-sdk`. For TypeScript, run `npm install @comfyorg/sdk`. cURL uses raw HTTP. **Model ID:** `openrouter/chat-completions` **Endpoint:** `POST https://api.comfy.org/v2/models/openrouter/chat-completions` This model has no runnable request example. Build the body from the input documentation below, then use it with the [Router quickstart](/development/comfy-router/quickstart). ## Schema ### Input Enable automatic prompt caching. When set at the top level, the system automatically applies cache breakpoints to the last cacheable block in the request. Currently supported for Anthropic Claude models. Possible values: `5m`, `1h` Possible values: `ephemeral` Debug options for inspecting request transformations (streaming only) If true, includes the transformed upstream request body in a debug chunk at the start of the stream. Only works with streaming mode. Frequency penalty (-2.0 to 2.0) Format: `double` Token logit bias adjustments Return log probabilities Maximum tokens in completion Maximum tokens (deprecated, use max\_completion\_tokens). Note: some providers enforce a minimum of 16. List of messages for the conversation Key-value pairs for additional object information (max 16 pairs, 64 char keys, 512 char values) Output modalities for the response. Supported values are "text", "image", and "audio". Model to use for completion Models to use for completion Whether to enable parallel function calling during tool use. When true, the model may generate multiple tool calls in a single response. Plugins you want to enable for this request, including their settings. Presence penalty (-2.0 to 2.0) Format: `double` When multiple model providers are available, optionally indicate your routing preference. Whether to allow backup providers to serve requests * true: (default) when the primary provider (or your custom providers in "order") is unavailable, use the next best provider. * false: use only the primary/custom provider, and return the upstream error if it's unavailable. Data collection setting. If no available model provider meets the requirement, your request will return an error. * allow: (default) allow providers which store user data non-transiently and may train on it * deny: use only providers which do not collect user data. Whether to restrict routing to only models that allow text distillation. When true, only models where the author has allowed distillation will be used. List of provider slugs to ignore. If provided, this list is merged with your account-wide ignored provider settings for this request. The object specifying the maximum price you want to pay for this request. USD price per million tokens, for prompt and completion. Price per million prompt tokens Price per million prompt tokens Price per million prompt tokens Price per million prompt tokens Price per million prompt tokens List of provider slugs to allow. If provided, this list is merged with your account-wide allowed provider settings for this request. An ordered list of provider slugs. The router will attempt to use the first provider in the subset of this list that supports your requested model, and fall back to the next if it is unavailable. If no providers are available, the request will fail with an error message. Preferred maximum latency (in seconds). Can be a number (applies to p50) or an object with percentile-specific cutoffs. Endpoints above the threshold(s) may still be used, but are deprioritized in routing. When using fallback models, this may cause a fallback model to be used instead of the primary model if it meets the threshold. Preferred minimum throughput (in tokens per second). Can be a number (applies to p50) or an object with percentile-specific cutoffs. Endpoints below the threshold(s) may still be used, but are deprioritized in routing. When using fallback models, this may cause a fallback model to be used instead of the primary model if it meets the threshold. A list of quantization levels to filter the provider by. Whether to filter providers to only those that support the parameters you've provided. If this setting is omitted or set to false, then providers will receive only the parameters they support, and ignore the rest. The sorting strategy to use for this request, if "order" is not specified. When set, no load balancing is performed. Whether to restrict routing to only ZDR (Zero Data Retention) endpoints. When true, only endpoints that do not retain prompts will be used. Configuration options for reasoning models Constrains effort on reasoning for reasoning models Possible values: `auto`, `concise`, `detailed` Response format configuration Any type Random seed for deterministic outputs The service tier to use for processing this request. A unique identifier for grouping related requests (e.g., a conversation or agent workflow) for observability. If provided in both the request body and the x-session-id header, the body value takes precedence. Maximum of 256 characters. Stop sequences (up to 4) Stop conditions for the server-tool agent loop. Any condition firing halts the loop (OR logic). When set, this overrides `max_tool_calls`. Enable streaming response Streaming configuration options Deprecated: This field has no effect. Full usage details are always included. Sampling temperature (0-2) Format: `double` Tool choice configuration Available tools for function calling Number of top log probabilities to return (0-20) Nucleus sampling parameter (0-1) Format: `double` Metadata for observability and tracing. Known keys (trace\_id, trace\_name, span\_name, generation\_name, parent\_span\_id) have special handling. Additional keys are passed through as custom metadata to configured broadcast destinations. Unique user identifier Generated from the schema Router serves at `GET /v2/models/openrouter/chat-completions/openapi.json`, the same document it validates a call against before the request reaches the provider. ### Output List of completion choices Possible values: `tool_calls`, `stop`, `length`, `content_filter`, `error` Choice index Log probabilities for the completion Log probabilities for content tokens UTF-8 bytes of the token Log probability of the token Format: `double` The token Top alternative tokens with probabilities Format: `double` Log probabilities for refusal tokens UTF-8 bytes of the token Log probability of the token Format: `double` The token Top alternative tokens with probabilities Format: `double` Assistant message for requests and responses Audio output data or reference Base64 encoded audio data Audio expiration timestamp Audio output identifier Audio transcript Assistant message content Generated images from image generation models URL or base64-encoded data of the generated image Optional name for the assistant Reasoning output Reasoning details for extended thinking models Refusal message if content was refused Tool calls made by the assistant Function arguments as JSON string Function name to call Tool call identifier Possible values: `function` Unix timestamp of creation Unique completion identifier Model used for completion Possible values: `chat.completion` Format: `double` Format: `double` Format: `double` Categorical kind of a pipeline stage. Multiple plugins can share a type (e.g. all guardrail-level plugins emit `guardrail`); the `name` field disambiguates which plugin emitted it. Possible values: `guardrail`, `plugin`, `server_tools`, `response_healing`, `context_compression` Possible values: `direct`, `auto`, `free`, `latest`, `alias`, `fallback`, `pareto`, `bodybuilder`, `fusion` The service tier used by the upstream provider for this request System fingerprint Token usage statistics Number of tokens in the completion Detailed completion token usage Cost of the completion Format: `double` Breakdown of upstream inference costs Format: `double` Format: `double` Format: `double` Whether a request was made using a Bring Your Own Key configuration Number of tokens in the prompt Detailed prompt token usage Total number of tokens ## Examples ### Output ```json theme={null} { "choices": [ { "finish_reason": "stop", "index": 0, "message": {} } ], "created": 1750000000, "id": "gen-0000000000-examplecompletion", "model": "anthropic/claude-sonnet-4.5", "object": "chat.completion", "system_fingerprint": null, "usage": { "completion_tokens": 128, "cost": 0.00123, "prompt_tokens": 42, "total_tokens": 170 } } ``` ## Before you ship The SDKs create an `Idempotency-Key` and reuse it for automatic retries. For manual retries, reuse the original key. Router can hold the connection for up to 10 minutes. When a request fails, Router sends an `X-Comfy-Error-Type` response header explaining why. A `422` means Router rejected the input before calling the provider, and a `413` means the request body was larger than Router accepts. Download generated assets promptly because [result URLs can expire](/development/comfy-router/reference#result-assets). Any size limit named in a field description above is the provider's own bound on that field, quoted from the provider's specification. Router applies a separate cap to the whole request body, which base64-encoded media counts against: see [request body size](/development/comfy-router/limitations#request-bodies-are-capped). Authentication, idempotency, request IDs, error buckets, retry pacing, spend limits. Model discovery, validation errors, retries, and billing. What Router does not do today, and what to use instead.