promptfoo--promptfoo/site/docs/providers/cohere.md

sidebar_label, description

sidebar_label	description
Cohere	Configure Cohere chat models for RAG-optimized inference, with current Command A, Aya, and Command R variants plus flexible prompt truncation controls

Cohere

The cohere provider is an interface to Cohere AI's chat inference API, with models such as Command R that are optimized for RAG and tool usage.

Setup

First, set the COHERE_API_KEY environment variable with your Cohere API key.

Next, edit the promptfoo configuration file to point to the Cohere provider.

• cohere:<model name> - uses the specified Cohere model (for example, command-a-03-2025).

The following models are confirmed supported. For an up-to-date list of supported models, see Cohere Models.

• command-a-03-2025
• command-r7b-12-2024
• command-a-translate-08-2025
• command-a-reasoning-08-2025
• command-a-vision-07-2025
• command-r-08-2024
• command-r-plus-08-2024
• tiny-aya-global
• tiny-aya-earth
• tiny-aya-fire
• tiny-aya-water
• c4ai-aya-expanse-32b
• c4ai-aya-vision-32b

Legacy aliases such as command, command-r, and command-r-plus are still accepted for existing configs, but new configurations should use the current IDs above.

Here's an example configuration:

providers:
  - id: cohere:command-a-03-2025
    config:
      temperature: 0.5
      max_tokens: 256
      prompt_truncation: 'AUTO'
      connectors:
        - id: web-search

Control over prompting

By default, a regular string prompt will be automatically wrapped in the appropriate chat format and sent to the Cohere API via the message field:

prompts:
  - 'Write a tweet about {{topic}}'

providers:
  - cohere:command-a-03-2025

tests:
  - vars:
      topic: bananas

If desired, your prompt can reference a YAML or JSON file that has a more complex set of API parameters. For example:

prompts:
  - file://prompt1.yaml

providers:
  - cohere:command-a-03-2025

tests:
  - vars:
      question: What year was he born?
  - vars:
      question: What did he like eating for breakfast?

And in prompt1.yaml:

chat_history:
  - role: USER
    message: 'Who discovered gravity?'
  - role: CHATBOT
    message: 'Isaac Newton'
message: '{{question}}'
connectors:
  - id: web-search

Embedding Configuration

Cohere provides embedding capabilities that can be used for various natural language processing tasks, including similarity comparisons. To use Cohere's embedding model in your evaluations, you can configure it as follows:

1. In your promptfooconfig.yaml file, add the embedding configuration under the defaultTest section:

defaultTest:
  options:
    provider:
      embedding:
        id: cohere:embedding:embed-english-v3.0

This configuration sets the default embedding provider for all tests that require embeddings (such as similarity assertions) to use Cohere's embed-english-v3.0 model.

2. You can also specify the embedding provider for individual assertions:

assert:
  - type: similar
    value: Some reference text
    provider:
      embedding:
        id: cohere:embedding:embed-english-v3.0

3. Additional configuration options can be passed to the embedding provider:

defaultTest:
  options:
    provider:
      embedding:
        id: cohere:embedding:embed-english-v3.0
        config:
          apiKey: your_api_key_here # If not set via environment variable
          truncate: NONE # Options: NONE, START, END

Displaying searches and documents

When the Cohere API is called, the provider can optionally include the search queries and documents in the output. This is controlled by the showSearchQueries and showDocuments config parameters. If true, the content will be appending to the output.

Configuration

Cohere parameters

Parameter	Description
apiKey	Your Cohere API key if not using an environment variable.
chatHistory	An array of chat history objects with role, message, and optionally user_name and conversation_id.
connectors	An array of connector objects for integrating with external systems.
documents	An array of document objects for providing reference material to the model.
frequency_penalty	Penalizes new tokens based on their frequency in the text so far.
k	Controls the diversity of the output via top-k sampling.
max_tokens	The maximum length of the generated text.
modelName	The model name to use for the chat completion.
p	Controls the diversity of the output via nucleus (top-p) sampling.
preamble_override	A string to override the default preamble used by the model.
presence_penalty	Penalizes new tokens based on their presence in the text so far.
prompt_truncation	Controls how prompts are truncated ('AUTO' or 'OFF').
search_queries_only	If true, only search queries are processed.
temperature	Controls the randomness of the output.

Special parameters

Parameter	Description
showSearchQueries	If true, includes the search queries used in the output.
showDocuments	If true, includes the documents used in the output.