Cohere
Peeklogic AI Orchestrator’s Cohere integration brings enterprise-grade natural language processing to your Salesforce organization. Cohere’s Command models are optimized for business tasks, reasoning, and chat, allowing you to build powerful conversational agents and automation flows.
Chat
Use Cohere’s Command models to generate text, answer questions, and interact conversationally.
Required fields:
- Name: Friendly name for your configuration.
- Model: Select the specific Cohere model (e.g., command-r, command-light, etc.).
Optional fields:
- Frequency Penalty: penalizes tokens that have already appeared in the generated text, reducing exact word repetition.
- Presence Penalty: Penalizes tokens that have already appeared, encouraging the model to talk about new topics.
- K (Top K): Limits the next token selection to the top K most probable tokens. Lower values make the output more focused; higher values make it more diverse.
- P (Top P): Uses nucleus sampling. The model considers the results of the tokens with top_p probability mass.
- Temperature: Controls the randomness of the output. Lower values mean more deterministic responses; higher values mean more creative ones.
- Max Tokens: The maximum number of tokens the model should generate in the response.
- Logprobs: If checked, returns the log probabilities of the generated tokens. Critical Restriction: You must disable Logprobs if you are using Function Calls or Chat with Documents (RAG). Enabling Logprobs with these features will cause the request to fail.
- Safety Mode: Defines the strictness of content filtering (e.g., CONTEXTUAL).
Seed: A number used to ensure reproducible results. If you use the same seed and parameters, the model will generate the same response
