Implementation:Cohere ai Cohere python ToolCallDelta Model

Knowledge Sources	Cohere Python SDK
Domains	SDK, Streaming, Function Calling
Last Updated	2026-02-15 14:00 GMT

Overview

ToolCallDelta is a Pydantic model representing a streaming chunk of a tool call being generated, containing partial data for the tool name, index, parameters, and plan text.

Description

The ToolCallDelta model encapsulates an incremental chunk of a tool call during streaming chat responses. When the Cohere Chat API is used with stream=True and the model generates tool calls, the tool call data arrives in pieces via ToolCallDelta objects embedded within ToolCallsChunkStreamedChatResponse events.

Each delta may contain partial data for:

• name -- A chunk of the tool call name (typically sent in the first delta).
• index -- The index of the tool call being generated (identifies which tool call in a multi-call scenario).
• parameters -- A chunk of the tool parameters as a string (JSON fragments that must be concatenated).
• text -- A chunk of the tool plan text the model is generating alongside the tool call.

All fields are optional since each streaming chunk contains only the data that has changed.

Usage

Use ToolCallDelta when consuming streaming chat responses that involve function calling. Accumulate the chunks by index to reconstruct the complete tool call name, parameters, and plan text.

Code Reference

Source Location

• Repository: Cohere Python SDK
• File: src/cohere/types/tool_call_delta.py

Signature

class ToolCallDelta(UncheckedBaseModel):
    name: typing.Optional[str] = pydantic.Field(default=None)
    index: typing.Optional[float] = pydantic.Field(default=None)
    parameters: typing.Optional[str] = pydantic.Field(default=None)
    text: typing.Optional[str] = pydantic.Field(default=None)

Import

from cohere.types import ToolCallDelta

I/O Contract

Fields

Field	Type	Required	Description
name	Optional[str]	No	Chunk of the tool call name.
index	Optional[float]	No	Index of the tool call being generated (for multi-call scenarios).
parameters	Optional[str]	No	Chunk of the tool parameters as a JSON string fragment.
text	Optional[str]	No	Chunk of the tool plan text.

Usage Examples

import json
from cohere.types import ToolCallDelta

# Accumulate tool call deltas during streaming
stream = client.chat_stream(
    message="What is the weather in London and Paris?",
    tools=[weather_tool]
)

tool_calls = {}  # index -> {name, parameters_str, text}

for event in stream:
    if event.event_type == "tool-calls-chunk":
        delta = event.tool_call_delta
        idx = delta.index

        if idx is not None:
            if idx not in tool_calls:
                tool_calls[idx] = {"name": "", "parameters_str": "", "text": ""}

            if delta.name:
                tool_calls[idx]["name"] += delta.name
            if delta.parameters:
                tool_calls[idx]["parameters_str"] += delta.parameters
            if delta.text:
                tool_calls[idx]["text"] += delta.text

    elif event.event_type == "stream-end":
        break

# Parse the accumulated tool calls
for idx, call_data in tool_calls.items():
    name = call_data["name"]
    parameters = json.loads(call_data["parameters_str"])
    print(f"Tool call {idx}: {name}({parameters})")

Related Pages

• Environment:Cohere_ai_Cohere_python_Python_SDK_Runtime