GET

/v1/schemas/mistral/chat/v1/agents/completions

Accept:

kind inputactivity chathash a71bcc97…extracted Nov 5, 2026version currentsource upstream spec ↗

field	recorded source
schema	Source linked above. upstream-spec Source trace `7b344504336785172c86cd826ce4b1cbfd12be228a8432cf046b2b8fdafacf72` { "fetchedAt": 1793856071, "extractorVersion": "1", "verifiedAt": null }

Request body

27 properties · 2 required · 46 $defs

property	type	description	constraints
messages*	SystemMessage \| UserMessage \| AssistantMessage[]	The prompt(s) to generate completions for, encoded as a list of dict with role and content.
agent_id*	string	The ID of the agent to use for this completion.
max_tokens	integer	The maximum number of tokens to generate in the completion. The token count of your prompt plus `max_tokens` cannot exceed the model's context length.	n ≥ 0 · nullable
stream	boolean	Whether to stream back partial progress. If set, tokens will be sent as data-only server-side events as they become available, with the stream terminated by a data: [DONE] message. Otherwise, the server will hold the request open until the timeout or until com…	default: false
stop	string \| string[]	Stop generation if this token is detected. Or if one of these tokens is detected when providing an array	nullable
random_seed	integer	The seed to use for random sampling. If set, different calls will generate deterministic results.	n ≥ 0 · nullable
metadata	object		nullable
response_format	ResponseFormat	Specify the format that the model must output. By default it will use `{ "type": "text" }`. Setting to `{ "type": "json_object" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the m…
tools	Tool \| ImageGenerationTool \| DocumentLibraryTool[]		nullable
tool_choice	ToolChoice \| enum		default: "auto"
presence_penalty	number	The `presence_penalty` determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.	-2 ≤ n ≤ 2 · nullable
frequency_penalty	number	The `frequency_penalty` penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.	-2 ≤ n ≤ 2 · nullable
min_tokens	integer		n ≥ 0 · nullable
repetition_penalty	number		0 < n ≤ 2 · nullable
top_k	integer		n ≥ -1 · nullable
logprobs	boolean		nullable
top_logprobs	integer		nullable
prompt_logprobs	boolean		nullable
top_prompt_logprobs	integer		nullable
n	integer	Number of completions to return for each request, input tokens are only billed once.	n ≥ 1 · nullable
prediction	Prediction	Enable users to specify expected results, optimizing response times by leveraging known or predictable content. This approach is especially effective for updating text documents or code files with minimal changes, reducing latency while maintaining high-qualit…
parallel_tool_calls	boolean		default: true
reasoning_effort	enum		none \| minimal \| low \| medium \| high \| xhigh · +1 more · nullable
prompt_mode	enum	Available options to the prompt_mode argument on the chat completion endpoint. Values represent high-level intent. Assignment to actual SPs is handled internally. System prompt may include knowledge cutoff date, model capabilities, tone to use, safety guidelin…	reasoning · nullable
guardrails	GuardrailConfig[]		nullable
prompt_cache_key	string		nullable
service_tier	enum	Determines whether to serve the request using priority or standard capacity.	auto \| standard_only · nullable

Types

46 $defs

APIKeyAuthobject

type	enum		api-key · default: "api-key"
value*	string

AssistantMessageobject

role	const		"assistant" · default: "assistant"
content	string \| ContentChunk[]		nullable
tool_calls	ToolCall[]		nullable
prefix	boolean	Set this to `true` when adding an assistant message as prefix to condition the model response. The role of the prefix message is to force the model to start its answer by the content of the message.	default: false

AudioChunkobject

type	const		"input_audio" · default: "input_audio"
input_audio*	string \| string

BuiltInConnectorsenum

ContentChunkTextChunk | ImageURLChunk | DocumentURLChunk | …

CustomConnectorobject

type	enum		connector · default: "connector"
connector_id*	string
authorization	OAuth2TokenAuth \| APIKeyAuth	Deprecated: per-request connector credentials will be removed. Authenticate the connector through the Connectors API instead.	nullable · deprecated
tool_configuration	ToolConfiguration		nullable

DocumentLibraryToolobject

tool_configuration	ToolConfiguration		nullable
type	enum		document_library · default: "document_library"
library_ids*	string[]	Ids of the library in which to search.	items ≥ 1

DocumentURLChunkobject

type	const		"document_url" · default: "document_url"
document_url*	string
document_name	string	The filename of the document	nullable

FileChunkobject

type	const		"file" · default: "file"
file_id*	string		format: uuid

Functionobject

name*	string
description	string	default: ""
strict	boolean	default: false
parameters*	object

FunctionCallobject

name*	string
arguments*	object \| string

FunctionNameobject

this restriction of `Function` is used to select a specific function to call

name*

string

GuardrailConfigobject

block_on_error	boolean	If true, return HTTP 403 and block request in the event of a server-side error	default: false
moderation_llm_v1	ModerationLLMV1Config		nullable
moderation_llm_v2	ModerationLLMV2Config		nullable

ImageGenerationToolobject

tool_configuration	ToolConfiguration		nullable
type	enum		image_generation · default: "image_generation"

ImageURLobject

url*	string
detail	enum		low \| auto \| high · nullable

ImageURLChunkobject

{"type":"image_url","image_url":"data:image/png;base64,iVBORw0"}

type	const		"image_url" · default: "image_url"
image_url*	ImageURL \| string

JsonSchemaobject

name*	string
description	string	nullable
schema*	object
strict	boolean	default: false

Available options to the prompt_mode argument on the chat completion endpoint. Values represent high-level intent. Assignment to actual SPs is handled internally. System prompt may include knowledge cutoff date, model capabilities, tone to use, safety guidelin…

reasoning

ModerationLLMActionenum

none | block

ModerationLLMV1CategoryThresholdsobject

sexual	number	nullable
hate_and_discrimination	number	nullable
violence_and_threats	number	nullable
dangerous_and_criminal_content	number	nullable
selfharm	number	nullable
health	number	nullable
financial	number	nullable
law	number	nullable
pii	number	nullable

ModerationLLMV1Configobject

model_name	string	Override model name. Should be omitted in general.	default: "mistral-moderation-2411"
custom_category_thresholds	ModerationLLMV1CategoryThresholds		nullable
ignore_other_categories	boolean	If true, only evaluate categories in custom_category_thresholds; others are ignored.	default: false
action	enum	Action to take if any score is above the threshold for any category.	none \| block

ModerationLLMV2CategoryThresholdsobject

sexual	number	nullable
hate_and_discrimination	number	nullable
violence_and_threats	number	nullable
dangerous	number	nullable
criminal	number	nullable
selfharm	number	nullable
health	number	nullable
financial	number	nullable
law	number	nullable
pii	number	nullable
jailbreaking	number	nullable

ModerationLLMV2Configobject

model_name	string	Override model name. Should be omitted in general.	default: "mistral-moderation-2603"
custom_category_thresholds	ModerationLLMV2CategoryThresholds		nullable
ignore_other_categories	boolean	If true, only evaluate categories in custom_category_thresholds; others are ignored.	default: false
action	enum	Action to take if any score is above the threshold for any category.	none \| block

OAuth2TokenAuthobject

type	enum		oauth2-token · default: "oauth2-token"
value*	string

Predictionobject

Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content.

type	const		"content" · default: "content"
content	string		default: ""

ReasoningEffortenum

ReferenceChunkobject

type	const		"reference" · default: "reference"
reference_ids*	integer \| string[]

RequestedServiceTierenum

auto | standard_only

ResourceChunkobject

type	const	"resource" · default: "resource"
uri*	string
content*	TextChunk \| ImageURLChunk

ResourceLinkChunkobject

type	const	"resource_link" · default: "resource_link"
uri*	string
metadata	object	nullable

ResponseFormatobject

Specify the format that the model must output. By default it will use `{ "type": "text" }`. Setting to `{ "type": "json_object" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the m…

type	enum		text \| json_object \| json_schema
json_schema	JsonSchema		nullable

ResponseFormatsenum

text | json_object | json_schema

SystemMessageobject

role	const		"system" · default: "system"
content*	string \| SystemMessageContentChunks[]

SystemMessageContentChunksTextChunk | ThinkChunk

TextChunkobject

type	const		"text" · default: "text"
text*	string

ThinkChunkobject

type	const		"thinking" · default: "thinking"
thinking*	TextChunk \| ToolReferenceChunk \| ReferenceChunk[]
signature	string	Signature to replay some reasoning blocks across turns.	nullable
closed	boolean	Whether the thinking chunk is closed or not. Currently only used for prefixing.	default: true

Toolobject

type	enum		function
function*	Function

ToolCallobject

id	string	default: "null"
type	enum	function
function*	FunctionCall
index	integer	default: 0

ToolChoiceobject

ToolChoice is either a ToolChoiceEnum or a ToolChoice

type	enum		function
function*	FunctionName	this restriction of `Function` is used to select a specific function to call

ToolChoiceEnumenum

auto | none | any | required

ToolConfigurationobject

exclude	string[]	nullable
include	string[]	nullable
requires_confirmation	string[]	nullable

ToolMessageobject

role	const	"tool" · default: "tool"
content*	string \| ContentChunk[]	nullable
tool_call_id	string	nullable
name	string	nullable

ToolReferenceChunkobject

type	const	"tool_reference" · default: "tool_reference"
tool*	enum \| string
title*	string
url	string	nullable
favicon	string	nullable
description	string	nullable

ToolTypesenum

function

UserMessageobject

role	const		"user" · default: "user"
content*	string \| ContentChunk[]		nullable

Validate a payload

before you spend tokens

shell

$ curl -X POST https://modelschemas.com/v1/validate \
    -d '{"provider":"mistral","endpointId":"v1/agents/completions","payload":{…}}'
→ {"valid": false, "errors": [{"path": "#", "keyword": "required", …}]}