Built by Metorial, the integration platform for agentic AI.
Generate embeddings using Cohere's Embed models for text, image data URIs, or mixed text/image inputs. Returns vector representations useful for semantic search, classification, clustering, and similarity comparisons. Supports configurable dimensionality and multiple output formats.
Generate text responses using Cohere's Command family of models. Supports multi-turn conversations with system prompts, tool use for calling external APIs, and retrieval augmented generation (RAG) with inline citations. Can be configured for reasoning tasks with adjustable thinking budgets.
Split text into tokens using byte-pair encoding (BPE) for a specific Cohere model. Useful for estimating costs, understanding how a model processes input, and checking token limits before making API calls.
Get the current status and details of a specific embed job by its ID.
Retrieve details about a specific dataset by its ID, including its type, validation status, and metadata.
View total Cohere hosted dataset storage usage for the organization.
Launch an asynchronous batch embedding job to embed a large dataset (100K+ documents). Results are stored as a new hosted dataset. Best suited for encoding large corpora for retrieval use cases.
List datasets stored in your Cohere account. Datasets are used for batch embedding jobs and can be filtered by type, date, and validation status.
List available Cohere models with their capabilities. Filter by endpoint type (chat, embed, rerank, etc.) to find models compatible with a specific use case.
Transcribe speech from an uploaded audio file using Cohere Transcribe. Supports the current Cohere audio transcription endpoint for automatic speech recognition.
Retrieve detailed metadata for a Cohere model, including supported endpoints, context length, tokenizer URL, feature flags, and default sampling parameters.
Cancel an active embed job. You will be charged for embeddings processed up to the cancellation point.
Create a Cohere hosted dataset by uploading a CSV, JSONL, or text file. Embed-input datasets can be used later with batch embed jobs.
Delete a dataset from your Cohere account by its ID. Datasets are automatically deleted after 30 days, but this allows immediate removal.
Rerank a list of documents by semantic relevance to a query using Cohere's Rerank models. Useful for improving search quality by re-ordering results from any existing search system based on meaning rather than keyword matching.
Convert an array of token IDs back into text using a specific Cohere model's tokenizer. The inverse of tokenization.
List all embed jobs in your Cohere account, including their status, model, and associated datasets.