## Create response
`POST /v1/responses`
```bash cURL theme={null}
curl --request POST \
--url https://www.anyfast.ai/v1/responses \
--header 'Authorization: Bearer YOUR_API_KEY' \
--header 'Content-Type: application/json' \
--data '{
"model": "gpt-6-astra",
"instructions": "Act as a senior reliability engineer.",
"input": "Review this retry design and propose a safe implementation plan.",
"reasoning": { "effort": "high", "summary": "auto" },
"text": { "verbosity": "medium" },
"max_output_tokens": 4096,
"store": true
}'
```
## Request headers
Bearer authentication in the form `Bearer YOUR_API_KEY`.
Data exchange format.
## Request body
AnyFast model ID. Use `gpt-6-astra`.
Text or an ordered list of message, image, file, and tool-result items.
System or developer instructions inserted into the model context.
Reasoning configuration.
`low`, `medium`, `high`, `xhigh`, or `max`. `none` and `minimal` are not supported.
`auto`, `concise`, or `detailed`.
`auto`, `current_turn`, or `all_turns`.
Use `standard`. The current AnyFast route does not expose GPT-6 Astra Pro mode.
Maximum visible-output and reasoning tokens. Maximum: `128000`.
`low`, `medium`, or `high`.
Additional response data. Use `reasoning.encrypted_content` for encrypted reasoning items. Do not use `message.output_text.logprobs`.
Stable key that improves cache matching for related requests.
Prompt-cache options. `mode` selects implicit or explicit breakpoints; `ttl` currently supports `30m`; `comparison_response_id` requests comparison diagnostics.
`implicit` or `explicit`.
Currently only `30m` is supported.
Marks an explicit reusable-prefix boundary. Use `{ "mode": "explicit" }`.
Application-defined string key-value pairs.
Stream response events over SSE.
Controls event obfuscation for streaming responses.
Controls upstream response storage. AnyFast does not expose response retrieval.
Use `disabled` on the current AnyFast route.
Do not send `temperature`, `top_p`, or `top_logprobs`. GPT-6 Astra also does not support `message.output_text.logprobs` in `include`.
AnyFast does not expose Responses retrieval, Conversations, or Vector Stores management. Do not use `background`, `conversation`, `previous_response_id`, stored prompt references, or `configuration_update`.
## Response
```json 200 theme={null}
{
"id": "resp_01astraexample",
"object": "response",
"created_at": 1788460800,
"completed_at": 1788460802,
"model": "gpt-6-astra",
"status": "completed",
"output": [
{ "id": "rs_01reasoning", "type": "reasoning", "summary": [] },
{
"id": "msg_01astraexample",
"type": "message",
"status": "completed",
"role": "assistant",
"content": [{
"type": "output_text",
"text": "Use bounded exponential backoff, idempotency keys, and a single retry owner.",
"annotations": []
}]
}
],
"output_text": "Use bounded exponential backoff, idempotency keys, and a single retry owner.",
"usage": {
"input_tokens": 38,
"input_tokens_details": { "cached_tokens": 0, "cache_write_tokens": 0 },
"output_tokens": 124,
"output_tokens_details": { "reasoning_tokens": 82 },
"total_tokens": 162
},
"error": null,
"incomplete_details": null
}
```
Response ID.
Always `response`.
`queued`, `in_progress`, `completed`, `incomplete`, `failed`, or `cancelled`.
Ordered reasoning, message, and tool-call output items.
Convenience aggregation of text output when available.
Input, output, cached, cache-write, and reasoning token usage.
Why a response ended incomplete.
Error details for a failed response.
## Create response
`POST /v1/responses`
```bash cURL theme={null}
curl --request POST \
--url https://www.anyfast.ai/v1/responses \
--header 'Authorization: Bearer YOUR_API_KEY' \
--header 'Content-Type: application/json' \
--data '{
"model": "gpt-6-astra",
"input": [{
"role": "user",
"content": [
{ "type": "input_text", "text": "Compare the diagram with the deployment notes and list conflicts." },
{ "type": "input_image", "image_url": "data:image/png;base64,", "detail": "high" },
{ "type": "input_file", "filename": "deployment-notes.pdf", "file_data": "data:application/pdf;base64," }
]
}],
"reasoning": { "effort": "high" }
}'
```
## Request body
Use `gpt-6-astra`.
Ordered Responses API input items.
`user`, `assistant`, `system`, or `developer`.
Text, image, or file content parts.
`input_text`, `input_image`, or `input_file`.
Text for an `input_text` part.
Public URL or data URL for an `input_image` part.
`auto`, `low`, or `high`.
Filename for inline file data.
Data URL containing Base64-encoded inline file data.
Public image URLs must allow server-side downloads. Data URLs and inline files are supported. Audio and video are not supported GPT-6 Astra input modalities.
## Response
```json 200 theme={null}
{
"id": "resp_01visionexample",
"object": "response",
"model": "gpt-6-astra",
"status": "completed",
"output": [{
"id": "msg_01visionexample",
"type": "message",
"status": "completed",
"role": "assistant",
"content": [{
"type": "output_text",
"text": "The diagram uses two retry owners, while the notes require one.",
"annotations": []
}]
}],
"usage": { "input_tokens": 842, "output_tokens": 67, "total_tokens": 909 }
}
```
Output item type, such as `message` or `reasoning`.
Content type, such as `output_text`.
Generated text.
Citations and annotations when present.
## Create response
`POST /v1/responses`
```bash cURL theme={null}
curl --request POST \
--url https://www.anyfast.ai/v1/responses \
--header 'Authorization: Bearer YOUR_API_KEY' \
--header 'Content-Type: application/json' \
--data '{
"model": "gpt-6-astra",
"input": "Classify the deployment risk.",
"text": {
"format": {
"type": "json_schema",
"name": "deployment_risk",
"strict": true,
"schema": {
"type": "object",
"properties": {
"risk": { "type": "string", "enum": ["low", "medium", "high"] },
"reason": { "type": "string" }
},
"required": ["risk", "reason"],
"additionalProperties": false
}
}
}
}'
```
## Request body
Use `gpt-6-astra`.
Prompt and optional context.
Use `json_schema` or `text`.
Schema name.
Description of the expected output.
JSON Schema that defines the response.
Enforce strict schema adherence.
## Response
```json 200 theme={null}
{
"id": "resp_01structuredexample",
"object": "response",
"model": "gpt-6-astra",
"status": "completed",
"output": [{
"id": "msg_01structuredexample",
"type": "message",
"status": "completed",
"role": "assistant",
"content": [{
"type": "output_text",
"text": "{\"risk\":\"medium\",\"reason\":\"Retries are not idempotent.\"}",
"annotations": []
}]
}]
}
```
JSON text that conforms to the supplied schema.
## Create response
`POST /v1/responses`
```bash cURL theme={null}
curl --request POST \
--url https://www.anyfast.ai/v1/responses \
--header 'Authorization: Bearer YOUR_API_KEY' \
--header 'Content-Type: application/json' \
--data '{
"model": "gpt-6-astra",
"input": "Check the deployment status for service api-gateway.",
"tools": [{
"type": "function",
"name": "get_deployment_status",
"description": "Return the current deployment status for a service.",
"parameters": {
"type": "object",
"properties": { "service": { "type": "string" } },
"required": ["service"],
"additionalProperties": false
},
"strict": true,
"async": true
}],
"tool_choice": "auto",
"parallel_tool_calls": true,
"max_tool_calls": 4
}'
```
## Request body
Use `gpt-6-astra`. Tool calling requires the Responses API.
Function, custom, or supported built-in tool definitions.
`function`, `custom`, `web_search`, `code_interpreter`, `shell`, `apply_patch`, `computer`, or `mcp`.
Function or custom-tool name.
When and how to use the tool.
JSON Schema for function arguments.
Strict function-argument validation.
Allow independent work while your application runs a function or custom tool.
Use `direct` for direct model tool calls.
Optional JSON Schema for Function output.
Web Search context size, such as `low`.
Domains allowed for Web Search.
Code Interpreter container configuration. Use `{ "type": "auto" }`; `memory_limit` is supported.
Shell environment configuration. Use `{ "type": "container_auto" }`; `memory_limit` is supported.
Label for an MCP server.
HTTPS endpoint for an MCP server.
MCP approval policy, such as `never`.
`none`, `auto`, or a forced-tool object.
Allow multiple tool calls in one turn.
Maximum total built-in tool calls.
Tool details to return, such as `web_search_call.action.sources` or `code_interpreter_call.outputs`.
Your application executes Function and Custom Tool calls. File Search requires a Vector Store created in the same upstream resource, but AnyFast does not currently expose the Vector Stores management endpoint.
## Tool-call response
```json 200 theme={null}
{
"id": "resp_01toolexample",
"object": "response",
"model": "gpt-6-astra",
"status": "completed",
"output": [{
"id": "fc_01deployment",
"type": "function_call",
"call_id": "call_01deployment",
"name": "get_deployment_status",
"arguments": "{\"service\":\"api-gateway\"}",
"status": "completed",
"async": true
}]
}
```
`function_call` for this workflow.
Tool-call correlation ID.
Function name.
JSON-encoded arguments.
Whether the call can complete asynchronously.
Do not rely on cross-request response or tool-item references. Requests routed to different upstream resources can reject those references.
## Create text completion
`POST /v1/chat/completions`
```bash cURL theme={null}
curl --request POST \
--url https://www.anyfast.ai/v1/chat/completions \
--header 'Authorization: Bearer YOUR_API_KEY' \
--header 'Content-Type: application/json' \
--data '{
"model": "gpt-6-astra",
"messages": [{ "role": "user", "content": "Return a concise deployment checklist." }],
"reasoning_effort": "low",
"max_completion_tokens": 1024
}'
```
## Request body
Use `gpt-6-astra`.
Ordered Chat Completions messages.
Message role.
Text message content.
`low`, `medium`, `high`, `xhigh`, or `max`.
Maximum completion and reasoning tokens.
Use Chat Completions for text requests only. GPT-6 Astra tool calling requires Responses API. Do not send `temperature`, `top_p`, `top_logprobs`, or `logprobs`.
## Response
Chat completion ID.
Generated text.
Why generation stopped.
Prompt, completion, and total token usage.
## Common errors