Chat Completions

The primary Serverless API: OpenAI-compatible chat completions with streaming, tools, structured outputs, and reasoning.

POST https://api.abhigya.ai/v1/chat/completions follows the OpenAI Chat Completions format, so OpenAI SDKs and compatible tools work by setting the base URL to https://api.abhigya.ai/v1.

Request

curl -sS https://api.abhigya.ai/v1/chat/completions \
  -H "Authorization: Bearer $ABHIGYA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen2.5-7b-instruct",
    "messages": [
      {"role": "system", "content": "Answer in one sentence."},
      {"role": "user", "content": "What is Abhigya Serverless?"}
    ],
    "max_tokens": 1024
  }'

Request fields

Field	Notes
`model`	Required. A model ID from `GET /v1/models`.
`messages`	Required. `system`, `user`, `assistant`, and `tool` messages.
`max_tokens`	Maximum generated tokens, including reasoning tokens. `max_completion_tokens` is also accepted.
`temperature`, `top_p`	Sampling controls.
`top_k`, `min_p`, `repetition_penalty`	Additional sampling controls.
`stop`	Stop sequences: at most 16, each at most 512 characters.
`seed`	Sampling seed.
`n`	Number of choices. Not every model supports `n > 1`.
`stream`	Stream server-sent events. See Streaming.
`tools`, `tool_choice`	Function tools. See Tools.
`response_format`	JSON mode or JSON Schema. See Structured outputs.
`reasoning_effort`	Reasoning effort. See Reasoning.

Output length

max_tokens bounds generated tokens, including reasoning tokens.
When you omit max_tokens, a default applies: the model's output cap, otherwise 65,536 tokens or half the context window, whichever is smaller.
A max_tokens above the model's cap is lowered to that cap.
If the prompt plus max_tokens does not fit the context window, the request returns 400 with code context_length_exceeded.

Response

{
  "id": "c3a7ac6de84a4e9eade77ab63a27a1ca",
  "object": "chat.completion",
  "created": 1790793243,
  "model": "qwen/qwen2.5-7b-instruct",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Abhigya Serverless is pay-per-token access to hosted models.",
        "reasoning_content": null,
        "tool_calls": null
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 22,
    "completion_tokens": 16,
    "total_tokens": 38,
    "prompt_tokens_details": {"cached_tokens": 0},
    "completion_tokens_details": {"reasoning_tokens": 0}
  }
}

finish_reason is stop, length, or tool_calls. See Usage and Billing for the usage fields.

Streaming

Set "stream": true to receive server-sent events. Each event is a data: line with a chat.completion.chunk object, followed by a blank line. The stream ends with data: [DONE].

curl -N -sS https://api.abhigya.ai/v1/chat/completions \
  -H "Authorization: Bearer $ABHIGYA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen2.5-7b-instruct",
    "messages": [{"role": "user", "content": "Count from 1 to 5."}],
    "max_tokens": 256,
    "stream": true
  }'

The first chunk sets delta.role.
Text arrives in delta.content. Reasoning text arrives in delta.reasoning_content.
A non-null finish_reason ends the choice.
A final chunk with an empty choices array carries usage.

Tools

Define tools with a JSON Schema for their parameters. parameters must be an object schema with "type": "object".

curl -sS https://api.abhigya.ai/v1/chat/completions \
  -H "Authorization: Bearer $ABHIGYA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen2.5-7b-instruct",
    "messages": [{"role": "user", "content": "What is the weather in Paris?"}],
    "tools": [{
      "type": "function",
      "function": {
        "name": "get_weather",
        "description": "Get the current weather for a city.",
        "parameters": {
          "type": "object",
          "properties": {"city": {"type": "string"}},
          "required": ["city"]
        }
      }
    }],
    "tool_choice": "auto",
    "max_tokens": 1024
  }'

When the model calls a tool, finish_reason is tool_calls and the message carries tool_calls. The arguments value is a JSON string. Run the tool, then send the assistant message and a tool message with the matching tool_call_id.

Structured outputs

Use response_format to constrain output to JSON. JSON mode returns a valid JSON object. JSON Schema constrains the output to your schema.

curl -sS https://api.abhigya.ai/v1/chat/completions \
  -H "Authorization: Bearer $ABHIGYA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen2.5-7b-instruct",
    "messages": [{"role": "user", "content": "Invent a fictional person."}],
    "response_format": {
      "type": "json_schema",
      "json_schema": {
        "name": "person",
        "strict": true,
        "schema": {
          "type": "object",
          "properties": {"name": {"type": "string"}, "age": {"type": "integer"}},
          "required": ["name", "age"],
          "additionalProperties": false
        }
      }
    },
    "max_tokens": 512
  }'

The JSON is returned as a string in message.content. Parse it before use. Schema keywords that cannot be enforced during decoding are dropped from the constraint; validate the output if you rely on them.

Reasoning

Reasoning models can think before answering. Reasoning text is returned separately in message.reasoning_content, and reasoning tokens are reported in usage.completion_tokens_details.reasoning_tokens.

Control reasoning with reasoning_effort: none, low, medium, high, or max. An effort the model does not list runs at a nearby tier. An unknown value returns 400 with code unsupported_value.

curl -sS https://api.abhigya.ai/v1/chat/completions \
  -H "Authorization: Bearer $ABHIGYA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek/deepseek-v4.1-flash",
    "messages": [{"role": "user", "content": "Is 391 prime?"}],
    "reasoning_effort": "high",
    "max_tokens": 4096
  }'

Reasoning tokens count toward max_tokens and bill as output tokens. If finish_reason is length with empty content, raise max_tokens or lower the effort.

Zero Data Retention

Zero Data Retention is enabled by default. No header is needed. See Zero Data Retention.