Create Chat Completion

post

{base_url}/chat/completions

Create a chat completion. Supports synchronous response (default), streaming SSE response (stream: true), function calling (tools array), and persistent conversations (conversation_id extension).

Model Field Formats

Format	Example	Description
Bot ULID	`01KQ07QSSST2TSZV5WT2YQ6BPM`	Targets a specific bot with its system prompt, KB, and guardrails
Bot ULID + override	`01KQ07QSSST2TSZV5WT2YQ6BPM/claude-v4.6-sonnet`	Bot config + specific underlying model
Raw model	`claude-v4.6-sonnet`	Direct Bedrock model, no bot context
OpenAI alias	`gpt-4`	Maps to nearest Claude tier

Available Raw Models: claude-v4.6-sonnet, claude-v4.6-opus, claude-v4.5-haiku, claude-v3.7-sonnet, amazon-nova-pro, amazon-nova-lite, amazon-nova-micro, llama-4-maverick-17b-instruct, llama-4-scout-17b-instruct, deepseek-r1, qwen3-32b, openai-gpt-oss-20b, openai-gpt-oss-120b, mistral-large-2

OpenAI Aliases: gpt-4 / gpt-4o / gpt-4-turbo → claude-v4.6-sonnet, gpt-3.5-turbo → claude-v4.5-haiku

Recent Requests

Time	Status	User Agent
Retrieving recent requests…

Loading…

Body Params

model

string

required

Model identifier. Accepts: Bot ULID (01KQ07QSSST2TSZV5WT2YQ6BPM), Bot ULID + override (01KQ07.../claude-v4.6-sonnet), raw model literal (claude-v4.6-sonnet), or OpenAI alias (gpt-4).

messages

array of objects

required

Conversation history. Must contain at least one user message.

messages*

temperature

number

0 to 2

Sampling temperature (0.0-2.0). Normalized internally to Bedrock range 0.0-1.0 by dividing by 2.

max_tokens

integer

≥ 1

Maximum number of output tokens. Response truncated with finish_reason length if hit.

top_p

number

0 to 1

Nucleus sampling — considers only the top top_p probability mass.

top_k

integer

≥ 0

Top-K sampling — limits the model to K most probable next tokens.

stream

boolean

Defaults to false

If true, returns Server-Sent Events (SSE) stream of incremental chunks ending with data: [DONE].

tools

array of objects

List of function definitions for client-side tool calling.

tools

tool_choice

enum

Controls tool selection. Values: auto (model decides), required (must call a tool), none (no tools), or object to force a specific function.

conversation_id

string

^[0-9A-HJKMNP-TV-Za-z]{26}$

Non-standard extension. A 26-character ULID to continue an existing conversation. Omit to start a new conversation. Returns 400 if malformed or if the conversation was started with a different bot.

Responses

200Chat completion response. When stream is false (default): returns a ChatCompletionResponse JSON object. When stream is true: returns text/event-stream of incremental chunks ending with data: [DONE].

400Bad request — malformed fields, empty messages, invalid conversation_id

401Missing or invalid API key

403Insufficient permissions

404Model or bot not found, or caller lacks access

429Monthly token quota exceeded

500Internal server error