> ## Documentation Index
> Fetch the complete documentation index at: https://composio-preview.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Cohere

> Cohere provides enterprise language and multimodal AI models for chat, embeddings, reranking, classification, and related workflows.

![Cohere logo](https://mintcdn.com/composio-preview/NPf8agM6UrvZD-Iu/images/toolkits/cohere.svg?fit=max\&auto=format\&n=NPf8agM6UrvZD-Iu\&q=85\&s=0393b6460c4bf86ef6c0c63b7461c236)

* **Category:** artificial intelligence
* **Auth:** API\_KEY
* **Composio-managed OAuth available?** N/A
* **Tools:** 22
* **Triggers:** 0
* **Slug:** `COHERE`
* **Version:** 20261030\_00

## Tools

> [accordion] Cancel Batch
>
> **Slug:** `COHERE_CANCEL_BATCH`
>
> Cancel one in-progress Cohere batch. This requires a production Cohere API key, stops the batch irreversibly, and cannot be undone.
>
> #### Input Parameters
>
> **id** `string` *required*
>
> Cohere ID of the in-progress batch to cancel.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Cancel Embed Job
>
> **Slug:** `COHERE_CANCEL_EMBED_JOB`
>
> Irreversibly cancel one active Cohere asynchronous embed job by ID. Cohere accrued charges for work processed before cancellation still apply, and partial results are not available. Cancel only a caller-owned job. This action sends one request and does not poll for the cancelled state.
>
> #### Input Parameters
>
> **id** `string` *required*
>
> Unique identifier of the active, caller-owned embed job to cancel.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Chat
>
> **Slug:** `COHERE_CHAT`
>
> Generate one non-streaming Cohere chat response from an ordered conversation, optionally using documents, tools, structured output, or reasoning controls.
>
> #### Input Parameters
>
> **seed** `integer`
>
> Best-effort deterministic sampling seed.
>
> **model** `string` *required*
>
> Cohere chat model name; use COHERE\_LIST\_MODELS with endpoint=chat to discover current choices.
>
> **tools** `array`
>
> Functions the model may call.
>
> **top\_k** `integer`
>
> Top-k sampling value; 0 disables top-k sampling.
>
> **top\_p** `number`
>
> Top-p probability mass from 0.01 to 0.99.
>
> **logprobs** `boolean`
>
> Include generated-token log probabilities in the response.
>
> **messages** `array` *required*
>
> Conversation messages in chronological order.
>
> **priority** `integer`
>
> Scheduling priority; lower numbers are handled earlier.
>
> **thinking** `object`
>
> Reasoning enablement and optional token budget.
>
> **documents** `array`
>
> Short relevant text or structured documents used to ground the answer.
>
> **max\_tokens** `integer`
>
> Maximum output tokens; low values may produce finish\_reason=MAX\_TOKENS.
>
> **safety\_mode** `string ("CONTEXTUAL" | "STRICT" | "OFF")`
>
> Safety instruction mode; model compatibility varies.
>
> **temperature** `number`
>
> Generation randomness; lower values are more deterministic.
>
> **tool\_choice** `string ("REQUIRED" | "NONE")`
>
> REQUIRED asks the model to call at least one supplied tool. NONE asks the model not to call tools, but this is best-effort and model-dependent.
>
> **strict\_tools** `boolean`
>
> Force generated tool calls to follow the supplied function schemas strictly.
>
> **stop\_sequences** `array`
>
> Up to five strings that stop generation before the matching sequence.
>
> **response\_format** `object`
>
> Plain-text or structured JSON output configuration.
>
> **citation\_options** `object`
>
> Citation generation configuration.
>
> **presence\_penalty** `number`
>
> Presence-based repetition penalty from 0 to 1.
>
> **frequency\_penalty** `number`
>
> Frequency-based repetition penalty from 0 to 1.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Check API Key
>
> **Slug:** `COHERE_CHECK_API_KEY`
>
> Check whether the connected Cohere API key is valid and return its organization and owner identifiers when available.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Create Batch
>
> **Slug:** `COHERE_CREATE_BATCH`
>
> Start one asynchronous Cohere batch from an existing validated batch request dataset and return immediately without polling. Cohere requires a production API key for the Batch API.
>
> #### Input Parameters
>
> **name** `string` *required*
>
> Human-readable name for the new batch.
>
> **model** `string` *required*
>
> Cohere model that will process the batch dataset.
>
> **input\_dataset\_id** `string` *required*
>
> Identifier of a validated Cohere batch request dataset.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Create Dataset
>
> **Slug:** `COHERE_CREATE_DATASET`
>
> Upload a Cohere dataset and start asynchronous validation. Returns immediately without waiting for validation to finish.
>
> #### Input Parameters
>
> **data** `object` *required*
>
> Primary CSV or JSONL dataset file to upload.
>
> **name** `string` *required*
>
> Dataset name; must contain fewer than 50 characters.
>
> **type** `string ("embed-input" | "embed-result" | "cluster-result" | "cluster-outliers" | "reranker-finetune-input" | "single-label-classification-finetune-input" | "chat-finetune-input" | "multi-label-classification-finetune-input" | "batch-chat-input" | "batch-openai-chat-input" | "batch-embed-v2-input" | "batch-chat-v2-input")` *required*
>
> Dataset type used by Cohere to validate the uploaded records.
>
> **eval\_data** `object`
>
> Optional evaluation CSV or JSONL dataset file to upload.
>
> **keep\_fields** `array`
>
> Source fields that must exist and should be retained in the dataset.
>
> **csv\_delimiter** `string`
>
> Single-character delimiter used to parse CSV uploads.
>
> **text\_separator** `string`
>
> Separator Cohere should use to split raw text uploads into records.
>
> **optional\_fields** `array`
>
> Source fields to retain when present without requiring them.
>
> **keep\_original\_file** `boolean`
>
> Whether Cohere should retain the original uploaded file.
>
> **skip\_malformed\_input** `boolean`
>
> Whether to drop malformed rows instead of failing validation.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Create Embed Job
>
> **Slug:** `COHERE_CREATE_EMBED_JOB`
>
> Start one asynchronous Cohere embedding job for a validated embed-input dataset and return immediately without polling.
>
> #### Input Parameters
>
> **name** `string`
>
> Optional human-readable name for the embed job.
>
> **model** `string` *required*
>
> Cohere embedding model to use for the job.
>
> **truncate** `string ("START" | "END")`
>
> Overlength behavior: START discards the beginning; END discards the end.
>
> **dataset\_id** `string` *required*
>
> Validated Cohere embed-input dataset to process.
>
> **input\_type** `string ("search_document" | "search_query" | "classification" | "clustering" | "image")` *required*
>
> How the resulting embeddings will be used: search\_document, search\_query, classification, clustering, or image.
>
> **embedding\_types** `array`
>
> Optional embedding encodings to produce: float, int8, uint8, binary, ubinary, or base64.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Delete Dataset
>
> **Slug:** `COHERE_DELETE_DATASET`
>
> Permanently delete a Cohere dataset by ID. This deletion cannot be undone.
>
> #### Input Parameters
>
> **id** `string` *required*
>
> Cohere dataset ID to permanently delete.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Detokenize Tokens
>
> **Slug:** `COHERE_DETOKENIZE_TOKENS`
>
> Convert ordered Cohere token IDs back to text using the tokenizer for an explicit matching model.
>
> #### Input Parameters
>
> **model** `string` *required*
>
> Cohere model whose tokenizer produced the token IDs; it must match the tokenization model.
>
> **tokens** `array` *required*
>
> Cohere token IDs to detokenize in order; use the exact order returned by tokenization.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Embed Inputs
>
> **Slug:** `COHERE_EMBED_INPUTS`
>
> Create v2 embeddings from exactly one input form: text strings, image data URIs, or structured multimodal content.
>
> #### Input Parameters
>
> **model** `string` *required*
>
> Cohere embedding model name, such as embed-v4.0; discover choices with COHERE\_LIST\_MODELS endpoint=embed.
>
> **texts** `array`
>
> One to 96 text strings to embed. Provide exactly one of texts, images, or inputs.
>
> **images** `array`
>
> Image data URIs to embed. JPEG, PNG, WebP, and GIF are supported. Embed v3 accepts one image up to 5 MB; v4 and newer accept up to 20 MB of image data combined.
>
> **inputs** `array`
>
> One to 96 structured multimodal inputs. Each input contains a non-empty ordered list of text or image\_url blocks.
>
> **priority** `integer`
>
> Scheduling priority; lower numbers are handled earlier.
>
> **truncate** `string ("NONE" | "START" | "END")`
>
> Overlength behavior: NONE errors, START discards the beginning, and END discards the end.
>
> **input\_type** `string ("search_document" | "search_query" | "classification" | "clustering" | "image")` *required*
>
> Purpose of the input: search\_document, search\_query, classification, clustering, or image.
>
> **max\_tokens** `integer`
>
> Maximum tokens embedded per input before applying truncate.
>
> **embedding\_types** `array`
>
> Vector encodings to return: float, int8, uint8, binary, ubinary, or base64.
>
> **output\_dimension** `integer ("256" | "512" | "1024" | "1536")`
>
> For embed-v4 and newer, output dimensions: 256, 512, 1024, or 1536.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Get Batch
>
> **Slug:** `COHERE_GET_BATCH`
>
> Retrieve one asynchronous Cohere batch's status and counters by batch ID. A usable batch ID depends on a production-key CREATE\_BATCH lifecycle; this action performs one read only, and callers control any bounded polling.
>
> #### Input Parameters
>
> **id** `string` *required*
>
> Batch ID returned by CREATE\_BATCH or LIST\_BATCHES.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Get Dataset
>
> **Slug:** `COHERE_GET_DATASET`
>
> Retrieve one Cohere dataset by ID, including its asynchronous validation state, schema, files, parsing details, and metrics. Callers control any polling and should bound it.
>
> #### Input Parameters
>
> **id** `string` *required*
>
> Cohere dataset ID returned by a dataset create or list operation.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Get Dataset Usage
>
> **Slug:** `COHERE_GET_DATASET_USAGE`
>
> Read the organization's Cohere dataset-storage consumption in bytes.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Get Embed Job
>
> **Slug:** `COHERE_GET_EMBED_JOB`
>
> Retrieve one Cohere asynchronous embedding job's current status, model, and input and output dataset identifiers by job ID.
>
> #### Input Parameters
>
> **id** `string` *required*
>
> Unique identifier of the embedding job to retrieve.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Get Model
>
> **Slug:** `COHERE_GET_MODEL`
>
> Retrieve a Cohere model's endpoint compatibility, context size, features, tokenizer, and default sampling parameters by model name.
>
> #### Input Parameters
>
> **model** `string` *required*
>
> Cohere model name to retrieve.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] List Batches
>
> **Slug:** `COHERE_LIST_BATCHES`
>
> Return one page of asynchronous Cohere batches and their statuses without creating or cancelling work.
>
> #### Input Parameters
>
> **order\_by** `string ("created_at" | "updated_at")`
>
> Sort batches by creation time or last-updated time.
>
> **page\_size** `integer`
>
> Maximum batches to return on this page (1-1000).
>
> **next\_cursor** `string`
>
> Continuation token returned by a prior call. Omit for the first page; pass the returned value back unchanged to fetch the next page.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] List Datasets
>
> **Slug:** `COHERE_LIST_DATASETS`
>
> Return one offset-based page of Cohere datasets, optionally filtered by type, validation status, or creation time. Cohere provides no total or continuation metadata, so callers control offset explicitly.
>
> #### Input Parameters
>
> **after** `string`
>
> Return datasets created after this timestamp.
>
> **limit** `integer`
>
> Maximum datasets to return on this page; must be positive.
>
> **before** `string`
>
> Return datasets created before this timestamp.
>
> **offset** `integer`
>
> Zero-based number of datasets to skip. Increase it by returned\_count to request the following page.
>
> **dataset\_type** `string`
>
> Filter datasets by the type recorded on the dataset.
>
> **validation\_status** `string ("unknown" | "queued" | "processing" | "failed" | "validated" | "skipped")`
>
> Filter datasets by their current validation status.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] List Embed Jobs
>
> **Slug:** `COHERE_LIST_EMBED_JOBS`
>
> List Cohere asynchronous embedding jobs and their current statuses.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] List Models
>
> **Slug:** `COHERE_LIST_MODELS`
>
> Discover Cohere model names and capabilities, optionally filtered to an endpoint or its default model. Returns one page without auto-pagination.
>
> #### Input Parameters
>
> **endpoint** `string ("chat" | "embed" | "classify" | "summarize" | "rerank" | "rate" | "generate")`
>
> Filter to models compatible with chat, embed, classify, summarize, rerank, rate, or generate.
>
> **page\_size** `integer`
>
> Requested models on this page (1-1000). The default of 1000 reduces exposure to Cohere's observed missing continuation token.
>
> **next\_cursor** `string`
>
> Continuation token from a prior response. Repeat the same endpoint and default\_only filters when continuing; omit for the first page.
>
> **default\_only** `boolean`
>
> Return only the endpoint's default model. Requires endpoint.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Rerank Documents
>
> **Slug:** `COHERE_RERANK_DOCUMENTS`
>
> Rank text documents by relevance to a query and return the highest-scoring source documents.
>
> #### Input Parameters
>
> **model** `string` *required*
>
> Cohere rerank model name; discover current choices with COHERE\_LIST\_MODELS endpoint=rerank.
>
> **query** `string` *required*
>
> Search query used to score the documents.
>
> **top\_n** `integer`
>
> Maximum highest-ranked results to return; must not exceed the number of documents.
>
> **priority** `integer`
>
> Scheduling priority; lower numbers are handled earlier.
>
> **documents** `array` *required*
>
> Non-empty text documents to rank; result indexes refer to this zero-based array.
>
> **max\_tokens\_per\_doc** `integer`
>
> Maximum tokens considered from each document.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Tokenize Text
>
> **Slug:** `COHERE_TOKENIZE_TEXT`
>
> Convert text into token IDs with an explicit Cohere model. Detokenization must use the same model because token IDs differ between model tokenizers.
>
> #### Input Parameters
>
> **text** `string` *required*
>
> Text to tokenize, from 1 to 65,536 characters.
>
> **model** `string` *required*
>
> Model whose tokenizer to use. Use the same model when detokenizing these token IDs.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Transcribe Audio
>
> **Slug:** `COHERE_TRANSCRIBE_AUDIO`
>
> Transcribe an uploaded audio file to text with a Cohere transcription model.
>
> #### Input Parameters
>
> **file** `object` *required*
>
> Audio file to transcribe. Supported file extensions are .flac, .mp3, .mpeg, .mpga, .ogg, and .wav.
>
> **model** `string` *required*
>
> Cohere transcription model ID.
>
> **language** `string` *required*
>
> Language of the audio in ISO-639-1 format.
>
> **temperature** `number`
>
> Sampling temperature from 0 to 1; lower values are more deterministic.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not
