Skip to content

coze-dev/coze-py

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Coze Python API SDK

PyPI version PyPI - Downloads codecov

Introduction

The Coze Python SDK is your comprehensive solution for seamlessly integrating Coze's powerful open APIs into Python applications.

  • Complete API coverage: All Coze open APIs and authentication methods supported
  • Dual interface: Both synchronous and asynchronous SDK calls available
  • Stream-optimized: Native Stream and AsyncStream objects for real-time data
  • Pagination made easy: Iterator-based Page objects for efficient list operations
  • Developer-first: Intuitive API design for rapid integration

Requirements

Python 3.7 or higher.

Install

pip install cozepy

Usage

Examples

Module Example File
Auth JWT OAuth examples/auth_oauth_jwt.py
Web OAuth examples/auth_oauth_web.py
PKCE OAuth examples/auth_oauth_pkce.py
Device OAuth examples/auth_oauth_device.py
Personal Access Token examples/auth_pat.py
Websockets Audio Speech by Websocket examples/websockets_audio_speech.py
Audio Transcription by Websocket examples/websockets_audio_transcriptions.py
Audio Chat by Websocket examples/websockets_chat.py
Audio Chat Realtime by Websocket examples/websockets_chat_realtime_gui.py
Bot Chat Bot Chat with Stream examples/chat_stream.py
Bot Chat without Stream examples/chat_no_stream.py
Bot Chat with Conversation examples/chat_conversation_stream.py
Bot Chat with Image examples/chat_multimode_stream.py
Bot Chat with Audio examples/chat_simple_audio.py
Bot Chat with Audio One-One examples/chat_oneonone_audio.py
Bot Chat With Local Plugin examples/chat_local_plugin.py
Workflow Run & Chat Workflow Run with Stream examples/workflow_stream.py
Workflow Run without Stream examples/workflow_no_stream.py
Workflow Async Run & Fetch examples/workflow_async.py
Workflow Chat with Stream examples/workflow_chat_stream.py
Workflow chat with Image and Stream examples/workflow_chat_multimode_stream.py
List Workflow Versions examples/workflow_version_list.py
Conversation Conversation examples/conversation.py
List Conversation examples/conversation_list.py
Dataset Create Dataset examples/dataset_create.py
Audio Audio Example by HTTP examples/audio.py
Bot Publish Bot examples/bot_publish.py
File Upload File examples/files_upload.py
template Duplicate Template examples/template_duplicate.py
User Get Current User examples/users_me.py
Workspace List Workspaces examples/workspace.py
Variable Retrieve Variable examples/variable_retrieve.py
Update Variable examples/variable_update.py
Other Timeout Config examples/timeout.py
Log Config examples/log.py
Exception Usage examples/exception.py

Initialize client

Start by obtaining your access token from the Coze platform:

Create a new personal access token by specifying a name, expiration period, and required permissions. Store this token securely—never expose it in code or version control.

import os

from cozepy import Coze, TokenAuth, COZE_CN_BASE_URL, AsyncCoze, AsyncTokenAuth

# Get an access_token through personal access token or oauth.
coze_api_token = os.getenv("COZE_API_TOKEN")
# The default access is api.coze.com, but if you need to access api.coze.cn,
# please use base_url to configure the api endpoint to access
coze_api_base = os.getenv("COZE_API_BASE") or COZE_CN_BASE_URL

# init coze with token and base_url
coze = Coze(auth=TokenAuth(coze_api_token), base_url=coze_api_base)
async_coze = AsyncCoze(auth=AsyncTokenAuth(coze_api_token), base_url=coze_api_base)

coze api access_token can also be generated via the OAuth App. For details, refer to:

Bot Chat

Create your bot in Coze and extract the bot ID from the URL (the final numeric segment).

Use coze.chat.stream for real-time streaming conversations. This method returns a Chat Iterator that yields events as they occur—simply iterate through the stream to process each event.

import os

from cozepy import Coze, TokenAuth, Message, ChatEventType, COZE_CN_BASE_URL

# initialize client
coze_api_token = os.getenv("COZE_API_TOKEN")
coze_api_base = os.getenv("COZE_API_BASE") or COZE_CN_BASE_URL
coze = Coze(auth=TokenAuth(coze_api_token), base_url=coze_api_base)

# The return values of the streaming interface can be iterated immediately.
for event in coze.chat.stream(
        # id of bot
        bot_id='bot_id',
        # id of user, Note: The user_id here is specified by the developer, for example, it can be the
        # business id in the developer system, and does not include the internal attributes of coze.
        user_id='user_id',
        # user input
        additional_messages=[Message.build_user_question_text("How are you?")]
        # conversation id, for Chaining conversation context
        # conversation_id='<conversation_id>',
):
    if event.event == ChatEventType.CONVERSATION_MESSAGE_DELTA:
        print(event.message.content, end="")

    if event.event == ChatEventType.CONVERSATION_CHAT_COMPLETED:
        print()
        print("token usage:", event.chat.usage.token_count)

Workflow Chat

Execute workflows directly through the SDK for powerful automation capabilities.

import os

from cozepy import TokenAuth, ChatEventType, Coze, COZE_CN_BASE_URL, Message

# Get the workflow id
workflow_id = os.getenv("COZE_WORKFLOW_ID") or "workflow id"
# Get the bot id
bot_id = os.getenv("COZE_BOT_ID")

# initialize client
coze_api_token = os.getenv("COZE_API_TOKEN")
coze_api_base = os.getenv("COZE_API_BASE") or COZE_CN_BASE_URL
coze = Coze(auth=TokenAuth(coze_api_token), base_url=coze_api_base)

conversation = coze.conversations.create()

# Call the coze.chat.stream method to create a chat. The create method is a streaming
# chat and will return a Chat Iterator. Developers should iterate the iterator to get
# chat event and handle them.
for event in coze.workflows.chat.stream(
        workflow_id=workflow_id,
        bot_id=bot_id,
        conversation_id=conversation.id,
        additional_messages=[
            Message.build_user_question_text("How are you?"),
        ],
):
    if event.event == ChatEventType.CONVERSATION_MESSAGE_DELTA:
        print(event.message.content, end="", flush=True)

    if event.event == ChatEventType.CONVERSATION_CHAT_COMPLETED:
        print()
        print("token usage:", event.chat.usage.token_count)

Real-time Audio Chat with WebSocket

import asyncio
import os

from cozepy import (
    AsyncTokenAuth,
    COZE_CN_BASE_URL,
    AsyncCoze,
    AsyncWebsocketsChatClient,
    AsyncWebsocketsChatEventHandler,
    AudioFormat,
    ConversationAudioDeltaEvent,
    ConversationChatCompletedEvent,
    ConversationChatCreatedEvent,
    ConversationMessageDeltaEvent,
    InputAudioBufferAppendEvent,
)
from cozepy.log import log_info
from cozepy.util import write_pcm_to_wav_file


class AsyncWebsocketsChatEventHandlerSub(AsyncWebsocketsChatEventHandler):
    delta = []

    async def on_conversation_chat_created(self, cli: AsyncWebsocketsChatClient, event: ConversationChatCreatedEvent):
        log_info("[examples] asr completed, logid=%s", event.detail.logid)

    async def on_conversation_message_delta(self, cli: AsyncWebsocketsChatClient, event: ConversationMessageDeltaEvent):
        print("Received:", event.data.content)

    async def on_conversation_audio_delta(self, cli: AsyncWebsocketsChatClient, event: ConversationAudioDeltaEvent):
        self.delta.append(event.data.get_audio())

    async def on_conversation_chat_completed(
        self, cli: "AsyncWebsocketsChatClient", event: ConversationChatCompletedEvent
    ):
        log_info("[examples] Saving audio data to output.wav")
        write_pcm_to_wav_file(b"".join(self.delta), "output.wav")


def wrap_coze_speech_to_iterator(coze: AsyncCoze, text: str):
    async def iterator():
        voices = await coze.audio.voices.list()
        content = await coze.audio.speech.create(
            input=text,
            voice_id=voices.items[0].voice_id,
            response_format=AudioFormat.WAV,
            sample_rate=24000,
        )
        for data in content._raw_response.iter_bytes(chunk_size=1024):
            yield data

    return iterator


async def main():
    coze_api_token = os.getenv("COZE_API_TOKEN")
    coze_api_base = os.getenv("COZE_API_BASE") or COZE_CN_BASE_URL
    coze = AsyncCoze(auth=AsyncTokenAuth(coze_api_token), base_url=coze_api_base)

    bot_id = os.getenv("COZE_BOT_ID")
    text = os.getenv("COZE_TEXT") or "How Are you?"

    # Initialize Audio
    speech_stream = wrap_coze_speech_to_iterator(coze, text)

    chat = coze.websockets.chat.create(
        bot_id=bot_id,
        on_event=AsyncWebsocketsChatEventHandlerSub(),
    )

    # Create and connect WebSocket client
    async with chat() as client:
        # Read and send audio data
        async for delta in speech_stream():
            await client.input_audio_buffer_append(
                InputAudioBufferAppendEvent.Data.model_validate(
                    {
                        "delta": delta,
                    }
                )
            )
        await client.input_audio_buffer_complete()
        await client.wait()


asyncio.run(main())

Request Debugging with LogID

Every SDK request includes a unique log ID for debugging purposes. Retrieve this ID from any response object to troubleshoot issues with Coze support.

import os
from cozepy import Coze, TokenAuth

coze = Coze(auth=TokenAuth(os.getenv("COZE_API_TOKEN")))

bot = coze.bots.retrieve(bot_id='bot id')
print(bot.response.logid) # support for CozeModel

stream = coze.chat.stream(bot_id='bot id', user_id='user id')
print(stream.response.logid) # support for stream

workspaces = coze.workspaces.list()
print(workspaces.response.logid) # support for paged

messages = coze.chat.messages.list(conversation_id='conversation id', chat_id='chat id')
print(messages.response.logid) # support for list(simple list, not paged)

Async Usage

Leverage full async/await support via httpx.AsyncClient. Simply swap Coze for AsyncCoze to enable non-blocking API calls across all Coze endpoints.

import os
import asyncio

from cozepy import AsyncTokenAuth, Message, AsyncCoze, COZE_CN_BASE_URL

# Get an access_token through personal access token or oauth.
coze_api_token = os.getenv("COZE_API_TOKEN")
# The default access is api.coze.com, but if you need to access api.coze.cn,
# please use base_url to configure the api endpoint to access
coze_api_base = os.getenv("COZE_API_BASE") or COZE_CN_BASE_URL

# init coze with token and base_url
coze = AsyncCoze(auth=AsyncTokenAuth(coze_api_token), base_url=coze_api_base)

async def main() -> None:
    chat = await coze.chat.create(
        bot_id='bot id',
        # id of user, Note: The user_id here is specified by the developer, for example, it can be the business id in the developer system, and does not include the internal attributes of coze.
        user_id='user id',
        additional_messages=[
            Message.build_user_question_text('how are you?'),
            Message.build_assistant_answer('I am fine, thank you.')
        ],
    )
    print('chat', chat)


asyncio.run(main())

Streaming Support

Both bot conversations and workflow executions offer real-time streaming capabilities for responsive user experiences.

chat streaming example:

import os
from cozepy import Coze, TokenAuth, ChatEventType, Message

coze = Coze(auth=TokenAuth(os.getenv("COZE_API_TOKEN")))

stream = coze.chat.stream(
    bot_id='bot id',
    # id of user, Note: The user_id here is specified by the developer, for example, it can be the
    # business id in the developer system, and does not include the internal attributes of coze.
    user_id='user id',
    additional_messages=[
        Message.build_user_question_text('how are you?'),
        Message.build_assistant_answer('I am fine, thank you.')
    ],
)
for event in stream:
    if event.event == ChatEventType.CONVERSATION_MESSAGE_DELTA:
        print('got message delta:', event.message.content)

workflow streaming example:

import os
from cozepy import Coze, TokenAuth, Stream, WorkflowEvent, WorkflowEventType

coze = Coze(auth=TokenAuth(os.getenv("COZE_API_TOKEN")))


def handle_workflow_iterator(stream: Stream[WorkflowEvent]):
    for event in stream:
        if event.event == WorkflowEventType.MESSAGE:
            print('got message', event.message)
        elif event.event == WorkflowEventType.ERROR:
            print('got error', event.error)
        elif event.event == WorkflowEventType.INTERRUPT:
            handle_workflow_iterator(coze.workflows.runs.resume(
                workflow_id='workflow id',
                event_id=event.interrupt.interrupt_data.event_id,
                resume_data='hey',
                interrupt_type=event.interrupt.interrupt_data.type,
            ))


handle_workflow_iterator(coze.workflows.runs.stream(
    workflow_id='workflow id',
    parameters={
        'input_key': 'input value',
    }
))

Asynchronous calls also support streaming mode:

import os
import asyncio

from cozepy import AsyncTokenAuth, ChatEventType, Message, AsyncCoze

coze = AsyncCoze(auth=AsyncTokenAuth(os.getenv("COZE_API_TOKEN")))


async def main():
    stream = coze.chat.stream(
        bot_id='bot id',
        # id of user, Note: The user_id here is specified by the developer, for example, it can be the
        # business id in the developer system, and does not include the internal attributes of coze.
        user_id='user id',
        additional_messages=[
            Message.build_user_question_text('how are you?'),
            Message.build_assistant_answer('I am fine, thank you.')
        ],
    )
    async for event in stream:
        if event.event == ChatEventType.CONVERSATION_MESSAGE_DELTA:
            print('got message delta:', event.message.content)


asyncio.run(main())

Smart Pagination

All list endpoints return intelligent paginators that support multiple iteration patterns. Explore three flexible approaches using bot listing as an example:

1. Not using iterators

import os
from cozepy import Coze, TokenAuth

coze = Coze(auth=TokenAuth(os.getenv("COZE_API_TOKEN")))

# open your workspace, browser url will be https://www.coze.com/space/<this is workspace id>/develop
# copy <this is workspace id> as workspace id
bots_page = coze.bots.list(space_id='workspace id', page_size=10)
bots = bots_page.items
total = bots_page.total
has_more = bots_page.has_more

2. Iterate over the paginator, getting T

import os
from cozepy import Coze, TokenAuth

coze = Coze(auth=TokenAuth(os.getenv("COZE_API_TOKEN")))

# open your workspace, browser url will be https://www.coze.com/space/<this is workspace id>/develop
# copy <this is workspace id> as workspace id
bots_page = coze.bots.list(space_id='workspace id', page_size=10)
for bot in bots_page:
    print('got bot:', bot)

Asynchronous methods also support:

import os
import asyncio

from cozepy import AsyncTokenAuth, AsyncCoze

coze = AsyncCoze(auth=AsyncTokenAuth(os.getenv("COZE_API_TOKEN")))


async def main():
    # open your workspace, browser url will be https://www.coze.com/space/<this is workspace id>/develop
    # copy <this is workspace id> as workspace id
    bots_page = await coze.bots.list(space_id='workspace id', page_size=10)
    async for bot in bots_page:
        print('got bot:', bot)


asyncio.run(main())

3. Iterate over the paginator iter_pages to get the next page paginator

import os
from cozepy import Coze, TokenAuth

coze = Coze(auth=TokenAuth(os.getenv("COZE_API_TOKEN")))

# open your workspace, browser url will be https://www.coze.com/space/<this is workspace id>/develop
# copy <this is workspace id> as workspace id
bots_page = coze.bots.list(space_id='workspace id', page_size=10)
for page in bots_page.iter_pages():
    print('got page:', page.page_num)
    for bot in page.items:
        print('got bot:', bot)

Asynchronous methods also support:

import asyncio
import os

from cozepy import AsyncTokenAuth, AsyncCoze

coze = AsyncCoze(auth=AsyncTokenAuth(os.getenv("COZE_API_TOKEN")))


async def main():
    # open your workspace, browser url will be https://www.coze.com/space/<this is workspace id>/develop
    # copy <this is workspace id> as workspace id
    bots_page = await coze.bots.list(space_id='workspace id', page_size=10)
    async for page in bots_page.iter_pages():
        print('got page:', page.page_num)
        for bot in page.items:
            print('got bot:', bot)


asyncio.run(main())

Configuration Options

Logging Configuration

Fine-tune SDK logging to match your debugging needs

import logging

from cozepy import setup_logging

# Enable debug logging (default: warning)
setup_logging(level=logging.DEBUG)

Timeout Configuration

Customize HTTP timeouts using the underlying httpx client for optimal performance in your environment.

import os

import httpx

from cozepy import COZE_COM_BASE_URL, Coze, TokenAuth, SyncHTTPClient

# Coze client is built on httpx, and supports passing a custom httpx.Client when initializing
# Coze, and setting a timeout on the httpx.Client
http_client = SyncHTTPClient(timeout=httpx.Timeout(
    # 600s timeout on elsewhere
    timeout=600.0,
    # 5s timeout on connect
    connect=5.0
))

# Init the Coze client through the access_token and custom timeout http client.
coze = Coze(auth=TokenAuth(token=os.getenv("COZE_API_TOKEN")),
            base_url=COZE_COM_BASE_URL,
            http_client=http_client
            )