Migrating to 1.0

Guide to upgrading your Pipecat Flows application from 0.0.x to 1.0

Pipecat Flows 1.0 removes deprecated APIs from the 0.0.x series. If you were already using the non-deprecated replacements, most of these changes won’t affect you.

Before upgrading, search your codebase for the deprecated imports and patterns listed below to identify what needs to change.

Pipecat Flows 1.0 requires Python 3.11 or later and pipecat-ai >= 1.0.0. If you’re also upgrading Pipecat, see the Pipecat 1.0 migration guide first.

1. FlowManager Initialization

The tts parameter was removed from FlowManager.__init__(). The built-in tts_say action now uses Pipecat’s TTSSpeakFrame directly and no longer needs a reference to your TTS service.

Before (1.0)

flow_manager = FlowManager(
    task=task,
    llm=llm,
    context_aggregator=context_aggregator,
    tts=tts,  # No longer accepted
)

After (1.0)

flow_manager = FlowManager(
    task=task,
    llm=llm,
    context_aggregator=context_aggregator,
)

What was removed

Removed	Replacement
FlowManager(tts=...)	Remove the parameter; tts_say works without it

2. Node Transitions

Three deprecated APIs for controlling node transitions were removed. All three are replaced by consolidated handlers that return a (result, next_node) tuple, or by set_node_from_config() for imperative transitions.

Before (1.0)

transition_to on FlowsFunctionSchema:

FlowsFunctionSchema(
    name="collect_name",
    description="Collect the user's name",
    properties={"name": {"type": "string"}},
    required=["name"],
    handler=handle_name,
    transition_to="next_node",  # No longer accepted
)

transition_callback on FlowsFunctionSchema:

FlowsFunctionSchema(
    name="collect_name",
    description="Collect the user's name",
    properties={"name": {"type": "string"}},
    required=["name"],
    handler=handle_name,
    transition_callback=decide_next_node,  # No longer accepted
)

set_node() on FlowManager:

# In a transition callback or handler
await flow_manager.set_node("next_node", next_node_config)  # No longer available

After (1.0)

Consolidated handler (preferred) — return a tuple of (result, next_node):

async def handle_name(args: FlowArgs, flow_manager: FlowManager):
    name = args["name"]
    flow_manager.state["name"] = name
    return {"name": name}, create_next_node()

FlowsFunctionSchema(
    name="collect_name",
    description="Collect the user's name",
    properties={"name": {"type": "string"}},
    required=["name"],
    handler=handle_name,
)

Direct function alternative:

async def collect_name(flow_manager: FlowManager, name: str):
    """Collect the user's name.

    Args:
        name: The user's name.
    """
    flow_manager.state["name"] = name
    return {"name": name}, create_next_node()

Imperative transition with set_node_from_config():

await flow_manager.set_node_from_config(next_node_config)

What was removed

Removed	Replacement
FlowsFunctionSchema(transition_to=...)	Return (result, next_node) from handler
FlowsFunctionSchema(transition_callback=...)	Return (result, next_node) from handler
flow_manager.set_node(id, config)	flow_manager.set_node_from_config(config)

3. Static Flows

The FlowConfig type and flow_config parameter were removed. Static flows defined all nodes upfront in a dictionary and transitioned by node name. Dynamic flows replaced them. You create NodeConfig objects at runtime and return them from handlers.

Before (1.0)

from pipecat_flows import FlowManager, FlowConfig

flow_config: FlowConfig = {
    "initial_node": "greeting",
    "nodes": {
        "greeting": {
            "role_messages": [...],
            "task_messages": [...],
            "functions": [...]
        },
        "collect_info": {
            "task_messages": [...],
            "functions": [...]
        }
    }
}

flow_manager = FlowManager(
    task=task,
    llm=llm,
    context_aggregator=context_aggregator,
    flow_config=flow_config,  # No longer accepted
)
await flow_manager.initialize()

After (1.0)

from pipecat_flows import FlowManager, NodeConfig

def create_greeting_node() -> NodeConfig:
    return {
        "role_message": "You are a helpful assistant.",
        "task_messages": [{"role": "developer", "content": "Greet the user."}],
        "functions": [collect_info_func],
    }

flow_manager = FlowManager(
    task=task,
    llm=llm,
    context_aggregator=context_aggregator,
)
await flow_manager.initialize(create_greeting_node())

What was removed

Removed	Replacement
FlowConfig type	NodeConfig returned from functions
FlowManager(flow_config=...)	flow_manager.initialize(initial_node)
Transition by node name string	Return NodeConfig from handler

4. Function Definition Format

Provider-specific function definition dicts (OpenAI, Anthropic, Gemini formats) are no longer accepted in node configs. Use FlowsFunctionSchema or direct functions instead.

Before (1.0)

# OpenAI-style dict format
node_config = {
    "task_messages": [...],
    "functions": [
        {
            "type": "function",
            "function": {
                "name": "get_weather",
                "description": "Get the weather",
                "parameters": {
                    "type": "object",
                    "properties": {
                        "location": {"type": "string"}
                    },
                    "required": ["location"]
                }
            },
            "handler": get_weather_handler,
        }
    ]
}

After (1.0)

Using FlowsFunctionSchema:

from pipecat_flows import FlowsFunctionSchema

node_config = {
    "task_messages": [...],
    "functions": [
        FlowsFunctionSchema(
            name="get_weather",
            description="Get the weather",
            properties={"location": {"type": "string"}},
            required=["location"],
            handler=get_weather_handler,
        )
    ]
}

Using a direct function:

async def get_weather(flow_manager: FlowManager, location: str):
    """Get the weather for a location.

    Args:
        location: The city to check weather for.
    """
    result = await fetch_weather(location)
    return {"weather": result}, None

node_config = {
    "task_messages": [...],
    "functions": [get_weather]
}

What was removed

Removed	Replacement
OpenAI-style function dicts	FlowsFunctionSchema
Anthropic-style function dicts	FlowsFunctionSchema
Gemini function_declarations dicts	FlowsFunctionSchema
Any provider-specific dict format	FlowsFunctionSchema or direct functions

Still-Active Deprecations

The following APIs still work in 1.0 but emit deprecation warnings. They will be removed in 2.0.

Deprecated	Since	Replacement
role_messages (list of dicts in NodeConfig)	v0.0.24	role_message (plain string)
RESET_WITH_SUMMARY context strategy	v0.0.25	Pipecat’s native LLMSummarizeContextFrame

role_messages: Replace the list-of-dicts role_messages field with the role_message string field. The string is sent as the LLM’s system instruction and persists across node transitions until a new node explicitly sets it again.

# Before
node = {
    "role_messages": [{"role": "system", "content": "You are a helpful assistant."}],
    "task_messages": [...]
}

# After
node = {
    "role_message": "You are a helpful assistant.",
    "task_messages": [...]
}

RESET_WITH_SUMMARY: Instead of using the RESET_WITH_SUMMARY context strategy, push an LLMSummarizeContextFrame in a pre-action to trigger on-demand summarization during a node transition. See the context summarization guide for details.