Chat Completions

POST /v1/chat/completions is the primary FlowAPI inference endpoint. It follows the familiar OpenAI-compatible chat format and is the recommended interface for most language and multimodal generation workflows.

Endpoint

POST https://api.flowapi.net/v1/chat/completions

Headers

Authorization: Bearer YOUR_FLOW_API_KEY
Content-Type: application/json

You can also provide X-Request-ID for tracing.

Request Body

Use one of the following examples to send a request in your preferred language.

from openai import OpenAI

client = OpenAI(
  api_key="YOUR_FLOWAPI_API_KEY",
  base_url="https://api.flowapi.net/v1",
)

response = client.chat.completions.create(
  model="deepseek-ai/DeepSeek-V3.2",
  messages=[
      {
          "role": "system",
          "content": "You are a concise technical assistant.",
      },
      {
          "role": "user",
          "content": "Explain why embeddings are useful in RAG systems.",
      },
  ],
  enable_thinking=False,
  thinking_budget=4096,
  temperature=0.7,
  top_p=0.7,
  max_tokens=1024,
  stream=False,
)

print(response.choices[0].message.content)

Parameters

model

string required

The model ID to call.

Example:

"deepseek-ai/DeepSeek-V3.2"

messages

array required

A list of messages comprising the conversation so far in OpenAI-compatible format.

Each message may use plain string content or multimodal content parts such as text and image_url when the selected model supports them.

Example:

[
  {
    "role": "system",
    "content": "You are a concise technical assistant."
  },
  {
    "role": "user",
    "content": [
      {
        "type": "text",
        "text": "Explain what is shown in this image."
      },
      {
        "type": "image_url",
        "image_url": {
          "url": "https://example.com/image.png",
          "detail": "high"
        }
      }
    ]
  }
]

stream

boolean

If set, tokens are returned as Server-Sent Events as they are made available.

Example:

false

max_tokens

integer

Maximum number of output tokens to generate.

Example:

1024

enable_thinking

boolean

Enables thinking mode for supported reasoning models.

This is a provider-specific passthrough field. It may be accepted by compatible upstream models and ignored by models that do not support thinking mode.

Example:

false

thinking_budget

integer default: 4096

Maximum number of tokens for chain-of-thought output. This field applies to supported reasoning models when thinking mode is enabled.

Required range: 128 <= x <= 32768

This is a provider-specific passthrough field. Support depends on the selected upstream model.

Example:

4096

stop

string | string[]

Up to 4 sequences where the API will stop generating further tokens. The returned text will not contain the stop sequence.

Example:

["</answer>"]

temperature

number

Determines the degree of randomness in the response.

Typical range: 0.0 to 2.0

Example:

0.7

top_p

number default: 0.7

The top_p nucleus sampling parameter dynamically adjusts the number of token choices considered during decoding.

Example:

0.7

frequency_penalty

number

Penalizes repeated tokens to reduce repetition in the output.

Example:

0.5

presence_penalty

number

Encourages the model to introduce new topics or less-repeated tokens.

Example:

0

response_format

object

An object specifying the format that the model must output.

For basic JSON mode, use { "type": "json_object" }.

Example:

{
  "type": "json_object"
}

n

integer

Number of completions to generate.

Example:

1

tools

object[]

A list of tools the model may call. Currently, function-style tools are the most portable option across OpenAI-compatible providers.

Example:

[
  {
    "type": "function",
    "function": {
      "name": "get_weather",
      "description": "Get the weather for a city",
      "parameters": {
        "type": "object",
        "properties": {
          "city": {
            "type": "string"
          }
        },
        "required": ["city"]
      }
    }
  }
]

Response Examples

Use the following examples to understand the most common response shapes returned by this endpoint.

{
"id": "chatcmpl-flowapi-123",
"object": "chat.completion",
"created": 1713000000,
"model": "deepseek-ai/DeepSeek-V3.2",
"choices": [
  {
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "Embeddings convert text into dense vectors, making semantic retrieval possible in RAG systems."
    },
    "finish_reason": "stop"
  }
],
"usage": {
  "prompt_tokens": 24,
  "completion_tokens": 18,
  "total_tokens": 42
}
}

When stream=true, the endpoint returns SSE chunks with object: "chat.completion.chunk" and incremental delta payloads.

See Streaming for detailed SSE handling guidance.

Notes

• Model availability may change over time as models are added, adjusted, or retired.
• Some advanced parameters are provider-specific and may be ignored by unsupported models.
• Multimodal input is model-dependent. Use a model that explicitly supports vision or mixed content when sending image parts.
• Request compatibility depends on the selected upstream provider and model. FlowAPI keeps the main request shape stable, while supported provider-specific fields may be forwarded when available.