swarmauri_llm_cohere 0.11.0.dev1

Provider-specific Swarmauri imports for Cohere Command chat, streaming, async, batch, preamble, and tool-use workflows.

Swarmauri Cohere LLM

swarmauri_llm_cohere provides provider-specific imports for Cohere Command models in Swarmauri. It re-exports CohereModel for standard chat workflows and CohereToolModel for tool-assisted conversations, both backed by the maintained implementations in swarmauri_standard.

Why Swarmauri Cohere LLM?

Use this package when a Swarmauri application should depend explicitly on Cohere chat support while preserving Swarmauri conversations, messages, toolkits, usage records, streaming, and batch workflows. The adapters call Cohere API v1 /chat, map Swarmauri system messages to Cohere preamble, pass prior messages as chat_history, and support Command model families used for enterprise chat, RAG, multilingual, and tool-use workloads.

FAQ

Q: Which Cohere API does this package use?

A: The current adapters call Cohere API v1 at https://api.cohere.ai/v1/chat. They do not yet target the newer /v2/chat endpoint.

Q: Which classes are exported?

A: The package exports CohereModel through the swarmauri.llms entry point and CohereToolModel through the swarmauri.tool_llms entry point.

Q: Which model names are configured locally?

A: The configured list includes 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-plus, command-r, command, and related dated aliases. The default is command-a-03-2025.

Q: How does tool use work?

A: CohereToolModel converts Swarmauri toolkit tools into Cohere-compatible tool schemas, sends an initial /chat request with tools, executes returned tool calls with the Swarmauri toolkit, then sends a second /chat request with tool_results.

Features

• Provider-specific imports for CohereModel and CohereToolModel.
• Swarmauri Conversation, message, toolkit, and usage-data integration.
• Sync predict() and async apredict() chat calls.
• Sync stream() and async astream() response streaming.
• Sync batch() and async abatch() helpers.
• Cohere v1 preamble support from Swarmauri system messages.
• Cohere v1 chat_history construction from previous Swarmauri messages.
• Cohere tool schema conversion and tool result submission in CohereToolModel.
• Retry handling for rate-limit and overloaded-provider status codes.
• Python 3.10, 3.11, 3.12, 3.13, and 3.14 support.

Prerequisites

• Cohere API key from the Cohere dashboard.
• Network access to api.cohere.ai.
• Swarmauri conversations built from Swarmauri message classes.
• Swarmauri Toolkit objects when using CohereToolModel.

Installation

Install with uv:

uv add swarmauri_llm_cohere

Install with pip:

pip install swarmauri_llm_cohere

Usage

Run a standard Cohere chat request:

import os

from swarmauri_llm_cohere import CohereModel
from swarmauri_standard.conversations.Conversation import Conversation
from swarmauri_standard.messages.HumanMessage import HumanMessage
from swarmauri_standard.messages.SystemMessage import SystemMessage

conversation = Conversation()
conversation.add_message(SystemMessage(content="Answer with concise technical prose."))
conversation.add_message(HumanMessage(content="What is Cohere Command A?"))

model = CohereModel(
    api_key=os.environ["COHERE_API_KEY"],
    name="command-a-03-2025",
)
result = model.predict(conversation=conversation, max_tokens=256, temperature=0.2)

print(result.get_last().content)

Stream a response:

import os

from swarmauri_llm_cohere import CohereModel
from swarmauri_standard.conversations.Conversation import Conversation
from swarmauri_standard.messages.HumanMessage import HumanMessage

conversation = Conversation()
conversation.add_message(HumanMessage(content="Draft a short support response."))

model = CohereModel(api_key=os.environ["COHERE_API_KEY"])

for token in model.stream(conversation=conversation, max_tokens=200):
    print(token, end="")

Use the tool model with a Swarmauri toolkit:

import os

from swarmauri_llm_cohere import CohereToolModel
from swarmauri_standard.conversations.Conversation import Conversation
from swarmauri_standard.messages.HumanMessage import HumanMessage
from swarmauri_standard.toolkits.Toolkit import Toolkit

toolkit = Toolkit()
# toolkit.add_tool(...)

conversation = Conversation()
conversation.add_message(HumanMessage(content="Use an available tool if needed."))

model = CohereToolModel(api_key=os.environ["COHERE_API_KEY"])
result = model.predict(conversation=conversation, toolkit=toolkit)

print(result.get_last().content)

Related Packages

LLM provider packages:

• swarmauri_llm_ai21
• swarmauri_llm_anthropic
• swarmauri_llm_cerebras
• swarmauri_llm_gemini
• swarmauri_llm_groq
• swarmauri_llm_mistral
• swarmauri_llm_openai

Foundational packages:

• swarmauri_core defines core interfaces.
• swarmauri_base provides base component classes.
• swarmauri_standard contains the shared Cohere implementations.
• swarmauri provides namespace imports and plugin discovery.

Best Practices

• Store COHERE_API_KEY in a secret manager or environment variable.
• Set name explicitly for reproducible model behavior.
• Use CohereToolModel only with tool schemas that have been tested against Cohere's v1 tool format.
• Use async methods for high-concurrency service workloads.
• Capture Swarmauri UsageData for cost and latency monitoring when Cohere returns usage fields.

License

Apache-2.0