qtype 0.0.12__py3-none-any.whl → 0.1.7__py3-none-any.whl

qtype/interpreter/stream/chat/vercel.py

@@ -0,0 +1,609 @@
+"""
+Pydantic models for Vercel AI SDK UI types.
+
+This module reproduces the exact TypeScript type shapes from the AI SDK UI
+as Pydantic models for use in Python implementations.
+
+Based on Vercel AI SDK v5.0.2:
+https://github.com/vercel/ai/tree/ai@5.0.2/packages/ai/src/ui
+
+## Streaming Protocol
+
+The Vercel AI SDK uses Server-Sent Events (SSE) to stream UIMessageChunks
+from the server to the client. Each chunk must be sent as:
+
+    data: {chunk_json}\\n\\n
+
+## Understanding Steps
+
+Steps are visual grouping markers in the UI. The StartStepChunk and
+FinishStepChunk are added to the UIMessage.parts array alongside content
+chunks (text, reasoning, tools) to create visual groupings.
+
+How it works:
+1. Send StartStepChunk - adds a 'step-start' marker to message.parts
+2. Send content chunks - text/reasoning/tool chunks are added to message.parts
+3. Send FinishStepChunk - adds a boundary AND resets activeTextParts/activeReasoningParts
+
+The final UIMessage.parts array will contain ALL parts in sequence:
+    [
+        { type: 'step-start' },
+        { type: 'text', text: 'Hello world', state: 'done' },
+        { type: 'step-start' },
+        { type: 'text', text: 'Step 2', state: 'done' },
+    ]
+
+Streaming Example - LLM generating a response as a step:
+    StartStepChunk()                                    # Marker added to parts
+    TextStartChunk(id="llm-response")                   # Text part added to parts
+    TextDeltaChunk(id="llm-response", delta="The ")     # Updates text in parts
+    TextDeltaChunk(id="llm-response", delta="answer ")  # Updates text in parts
+    TextDeltaChunk(id="llm-response", delta="is 42")    # Updates text in parts
+    TextEndChunk(id="llm-response")                     # Marks text as done
+    FinishStepChunk()                                   # Resets active parts
+
+Streaming Example - File writer status as a step:
+    StartStepChunk()                                    # Marker added to parts
+    TextStartChunk(id="file-status")                    # Text part added to parts
+    TextDeltaChunk(id="file-status", delta="Writing 3 records...")
+    TextEndChunk(id="file-status")                      # Marks text as done
+    FinishStepChunk()                                   # Resets active parts
+
+Streaming Example - Multiple steps in sequence:
+    StartStepChunk()                                    # Step 1 marker
+    TextStartChunk(id="step1")
+    TextDeltaChunk(id="step1", delta="Step 1 content")
+    TextEndChunk(id="step1")
+    FinishStepChunk()                                   # Resets for next step
+
+    StartStepChunk()                                    # Step 2 marker
+    TextStartChunk(id="step2")
+    TextDeltaChunk(id="step2", delta="Step 2 content")
+    TextEndChunk(id="step2")
+    FinishStepChunk()                                   # Resets for next step
+
+## Tool Execution
+
+Tools have a multi-stage lifecycle:
+
+1. ToolInputStartChunk - Begin receiving tool input
+2. ToolInputDeltaChunk - Incremental input text (JSON being parsed)
+3. ToolInputAvailableChunk - Complete input ready, tool can execute
+4. ToolOutputAvailableChunk - Tool completed successfully
+   OR ToolOutputErrorChunk - Tool execution failed
+
+## Complete Message Stream Pattern
+
+    StartChunk(messageId="msg-123")
+    StartStepChunk()
+    TextStartChunk(id="text-1")
+    TextDeltaChunk(id="text-1", delta="Hello")
+    TextDeltaChunk(id="text-1", delta=" world")
+    TextEndChunk(id="text-1")
+    FinishStepChunk()
+    FinishChunk()
+"""
+
+from __future__ import annotations
+
+from typing import Any, Literal, Union
+
+from pydantic import BaseModel, Field
+
+
+# Provider metadata
+class ProviderMetadata(BaseModel):
+    """Provider-specific metadata.
+
+    Reproduces: ProviderMetadata
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/types/provider-metadata.ts
+    """
+
+    model_config = {"extra": "allow"}
+
+
+# UI Message Parts (final state in UIMessage.parts)
+# https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui/ui-messages.ts
+class TextUIPart(BaseModel):
+    """A text part of a message.
+
+    Reproduces: TextUIPart
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui/ui-messages.ts
+    """
+
+    type: Literal["text"] = "text"
+    text: str
+    state: Literal["streaming", "done"] | None = None
+    provider_metadata: ProviderMetadata | None = Field(
+        default=None, alias="providerMetadata"
+    )
+
+
+class ReasoningUIPart(BaseModel):
+    """A reasoning part of a message.
+
+    Reproduces: ReasoningUIPart
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui/ui-messages.ts
+    """
+
+    type: Literal["reasoning"] = "reasoning"
+    text: str
+    state: Literal["streaming", "done"] | None = None
+    provider_metadata: ProviderMetadata | None = Field(
+        default=None, alias="providerMetadata"
+    )
+
+
+class SourceUrlUIPart(BaseModel):
+    """A source URL part of a message.
+
+    Reproduces: SourceUrlUIPart
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui/ui-messages.ts
+    """
+
+    type: Literal["source-url"] = "source-url"
+    source_id: str = Field(alias="sourceId")
+    url: str
+    title: str | None = None
+    provider_metadata: ProviderMetadata | None = Field(
+        default=None, alias="providerMetadata"
+    )
+
+
+class SourceDocumentUIPart(BaseModel):
+    """A document source part of a message.
+
+    Reproduces: SourceDocumentUIPart
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui/ui-messages.ts
+    """
+
+    type: Literal["source-document"] = "source-document"
+    source_id: str = Field(alias="sourceId")
+    media_type: str = Field(alias="mediaType")
+    title: str
+    filename: str | None = None
+    provider_metadata: ProviderMetadata | None = Field(
+        default=None, alias="providerMetadata"
+    )
+
+
+class FileUIPart(BaseModel):
+    """A file part of a message.
+
+    Reproduces: FileUIPart
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui/ui-messages.ts
+    """
+
+    type: Literal["file"] = "file"
+    media_type: str = Field(alias="mediaType")
+    filename: str | None = None
+    url: str
+    provider_metadata: ProviderMetadata | None = Field(
+        default=None, alias="providerMetadata"
+    )
+
+
+class StepStartUIPart(BaseModel):
+    """A step boundary part of a message.
+
+    Reproduces: StepStartUIPart
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui/ui-messages.ts
+    """
+
+    type: Literal["step-start"] = "step-start"
+
+
+# Union type for UI message parts
+UIMessagePart = Union[
+    TextUIPart,
+    ReasoningUIPart,
+    SourceUrlUIPart,
+    SourceDocumentUIPart,
+    FileUIPart,
+    StepStartUIPart,
+]
+
+
+# UI Message
+class UIMessage(BaseModel):
+    """AI SDK UI Message.
+
+    Reproduces: UIMessage
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui/ui-messages.ts
+    """
+
+    id: str
+    role: Literal["system", "user", "assistant"]
+    metadata: dict[str, Any] | None = None
+    parts: list[UIMessagePart]
+
+
+# Chat Request (the request body sent from frontend)
+class ChatRequest(BaseModel):
+    """Chat request format sent from AI SDK UI/React.
+
+    Reproduces: ChatRequest
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui/chat-transport.ts
+    """
+
+    id: str  # chatId
+    messages: list[UIMessage]
+    trigger: Literal["submit-message", "regenerate-message"]
+    message_id: str | None = Field(default=None, alias="messageId")
+
+
+class CompletionRequest(BaseModel):
+    """Completion request format sent from AI SDK UI/React useCompletion hook.
+
+    The useCompletion hook sends { prompt: string, ...body } where body can
+    contain any additional fields needed by the flow.
+
+    Reproduces: Request body from useCompletion
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui/call-completion-api.ts
+    """
+
+    prompt: str
+    model_config = {"extra": "allow"}  # Allow arbitrary additional fields
+
+
+# UI Message Chunks (streaming events)
+# https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+
+
+class TextStartChunk(BaseModel):
+    """Text start chunk - begins a text content section.
+
+    Reproduces: UIMessageChunk (text-start variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["text-start"] = "text-start"
+    id: str
+    provider_metadata: ProviderMetadata | None = Field(
+        default=None, alias="providerMetadata"
+    )
+
+
+class TextDeltaChunk(BaseModel):
+    """Text delta chunk - incremental text content.
+
+    Reproduces: UIMessageChunk (text-delta variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["text-delta"] = "text-delta"
+    id: str
+    delta: str
+    provider_metadata: ProviderMetadata | None = Field(
+        default=None, alias="providerMetadata"
+    )
+
+
+class TextEndChunk(BaseModel):
+    """Text end chunk - completes a text content section.
+
+    Reproduces: UIMessageChunk (text-end variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["text-end"] = "text-end"
+    id: str
+    provider_metadata: ProviderMetadata | None = Field(
+        default=None, alias="providerMetadata"
+    )
+
+
+class ReasoningStartChunk(BaseModel):
+    """Reasoning start chunk - begins a reasoning section.
+
+    Reproduces: UIMessageChunk (reasoning-start variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["reasoning-start"] = "reasoning-start"
+    id: str
+    provider_metadata: ProviderMetadata | None = Field(
+        default=None, alias="providerMetadata"
+    )
+
+
+class ReasoningDeltaChunk(BaseModel):
+    """Reasoning delta chunk - incremental reasoning content.
+
+    Reproduces: UIMessageChunk (reasoning-delta variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["reasoning-delta"] = "reasoning-delta"
+    id: str
+    delta: str
+    provider_metadata: ProviderMetadata | None = Field(
+        default=None, alias="providerMetadata"
+    )
+
+
+class ReasoningEndChunk(BaseModel):
+    """Reasoning end chunk - completes a reasoning section.
+
+    Reproduces: UIMessageChunk (reasoning-end variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["reasoning-end"] = "reasoning-end"
+    id: str
+    provider_metadata: ProviderMetadata | None = Field(
+        default=None, alias="providerMetadata"
+    )
+
+
+class ToolInputStartChunk(BaseModel):
+    """Tool input start chunk - begins tool input streaming.
+
+    Reproduces: UIMessageChunk (tool-input-start variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["tool-input-start"] = "tool-input-start"
+    tool_call_id: str = Field(alias="toolCallId")
+    tool_name: str = Field(alias="toolName")
+    provider_executed: bool | None = Field(
+        default=None, alias="providerExecuted"
+    )
+    dynamic: bool | None = None
+
+
+class ToolInputDeltaChunk(BaseModel):
+    """Tool input delta chunk - incremental tool input text.
+
+    Reproduces: UIMessageChunk (tool-input-delta variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["tool-input-delta"] = "tool-input-delta"
+    tool_call_id: str = Field(alias="toolCallId")
+    input_text_delta: str = Field(alias="inputTextDelta")
+
+
+class ToolInputAvailableChunk(BaseModel):
+    """Tool input available chunk - complete tool input ready.
+
+    Reproduces: UIMessageChunk (tool-input-available variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["tool-input-available"] = "tool-input-available"
+    tool_call_id: str = Field(alias="toolCallId")
+    tool_name: str = Field(alias="toolName")
+    input: Any
+    provider_executed: bool | None = Field(
+        default=None, alias="providerExecuted"
+    )
+    provider_metadata: ProviderMetadata | None = Field(
+        default=None, alias="providerMetadata"
+    )
+    dynamic: bool | None = None
+
+
+class ToolOutputAvailableChunk(BaseModel):
+    """Tool output available chunk - tool execution completed.
+
+    Reproduces: UIMessageChunk (tool-output-available variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["tool-output-available"] = "tool-output-available"
+    tool_call_id: str = Field(alias="toolCallId")
+    output: Any
+    provider_executed: bool | None = Field(
+        default=None, alias="providerExecuted"
+    )
+    dynamic: bool | None = None
+
+
+class ToolOutputErrorChunk(BaseModel):
+    """Tool output error chunk - tool execution failed.
+
+    Reproduces: UIMessageChunk (tool-output-error variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["tool-output-error"] = "tool-output-error"
+    tool_call_id: str = Field(alias="toolCallId")
+    error_text: str = Field(alias="errorText")
+    provider_executed: bool | None = Field(
+        default=None, alias="providerExecuted"
+    )
+    dynamic: bool | None = None
+
+
+class SourceUrlChunk(BaseModel):
+    """Source URL chunk - references a URL source.
+
+    Reproduces: UIMessageChunk (source-url variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["source-url"] = "source-url"
+    source_id: str = Field(alias="sourceId")
+    url: str
+    title: str | None = None
+    provider_metadata: ProviderMetadata | None = Field(
+        default=None, alias="providerMetadata"
+    )
+
+
+class SourceDocumentChunk(BaseModel):
+    """Source document chunk - references a document source.
+
+    Reproduces: UIMessageChunk (source-document variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["source-document"] = "source-document"
+    source_id: str = Field(alias="sourceId")
+    media_type: str = Field(alias="mediaType")
+    title: str
+    filename: str | None = None
+    provider_metadata: ProviderMetadata | None = Field(
+        default=None, alias="providerMetadata"
+    )
+
+
+class FileChunk(BaseModel):
+    """File chunk - includes a file in the message.
+
+    Reproduces: UIMessageChunk (file variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["file"] = "file"
+    url: str
+    media_type: str = Field(alias="mediaType")
+    provider_metadata: ProviderMetadata | None = Field(
+        default=None, alias="providerMetadata"
+    )
+
+
+class ErrorChunk(BaseModel):
+    """Error chunk - signals an error occurred.
+
+    Reproduces: UIMessageChunk (error variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["error"] = "error"
+    error_text: str = Field(alias="errorText")
+
+
+class StartStepChunk(BaseModel):
+    """Start step chunk - marks the beginning of a step boundary.
+
+    This is a boundary marker with NO fields. Content between
+    start-step and finish-step is grouped as a single step in the UI.
+
+    Reproduces: UIMessageChunk (start-step variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+
+    Example usage:
+        StartStepChunk()
+        TextStartChunk(id="step-1-text")
+        TextDeltaChunk(id="step-1-text", delta="Processing...")
+        TextEndChunk(id="step-1-text")
+        FinishStepChunk()
+    """
+
+    type: Literal["start-step"] = "start-step"
+
+
+class FinishStepChunk(BaseModel):
+    """Finish step chunk - marks the end of a step boundary.
+
+    This is a boundary marker with NO fields. When received, it
+    resets activeTextParts and activeReasoningParts in the processor.
+
+    Reproduces: UIMessageChunk (finish-step variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["finish-step"] = "finish-step"
+
+
+class StartChunk(BaseModel):
+    """Start chunk - marks the beginning of message streaming.
+
+    Reproduces: UIMessageChunk (start variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["start"] = "start"
+    message_id: str | None = Field(default=None, alias="messageId")
+    message_metadata: dict[str, Any] | None = Field(
+        default=None, alias="messageMetadata"
+    )
+
+
+# temp
+class ToolStarted(BaseModel):
+    """Start chunk.
+
+    Reproduces: Tool Started from ui/ui-message-chunks.ts
+    """
+
+    type: Literal["tool_started"] = "tool_started"
+    message_id: str | None = Field(default=None, alias="messageId")
+    message_metadata: dict[str, Any] | None = Field(
+        default=None, alias="messageMetadata"
+    )
+
+
+class ToolResultReceived(BaseModel):
+    """Start chunk.
+
+    Reproduces: Tool Result Received from ui/ui-message-chunks.ts
+    """
+
+    type: Literal["tool_result_received"] = "tool_result_received"
+    message_id: str | None = Field(default=None, alias="messageId")
+    message_metadata: dict[str, Any] | None = Field(
+        default=None, alias="messageMetadata"
+    )
+
+
+class FinishChunk(BaseModel):
+    """Finish chunk - marks the completion of message streaming.
+
+    Reproduces: UIMessageChunk (finish variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["finish"] = "finish"
+    message_metadata: dict[str, Any] | None = Field(
+        default=None, alias="messageMetadata"
+    )
+
+
+class AbortChunk(BaseModel):
+    """Abort chunk - signals streaming was aborted.
+
+    Reproduces: UIMessageChunk (abort variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["abort"] = "abort"
+
+
+class MessageMetadataChunk(BaseModel):
+    """Message metadata chunk - updates message metadata during stream.
+
+    Reproduces: UIMessageChunk (message-metadata variant)
+    https://github.com/vercel/ai/blob/ai@5.0.2/packages/ai/src/ui-message-stream/ui-message-chunks.ts
+    """
+
+    type: Literal["message-metadata"] = "message-metadata"
+    message_metadata: dict[str, Any] = Field(alias="messageMetadata")
+
+
+# Union type for all UI message chunks
+UIMessageChunk = Union[
+    TextStartChunk,
+    TextDeltaChunk,
+    TextEndChunk,
+    ReasoningStartChunk,
+    ReasoningDeltaChunk,
+    ReasoningEndChunk,
+    ToolInputStartChunk,
+    ToolInputDeltaChunk,
+    ToolInputAvailableChunk,
+    ToolOutputAvailableChunk,
+    ToolOutputErrorChunk,
+    SourceUrlChunk,
+    SourceDocumentChunk,
+    FileChunk,
+    ErrorChunk,
+    StartStepChunk,
+    FinishStepChunk,
+    StartChunk,
+    FinishChunk,
+    AbortChunk,
+    MessageMetadataChunk,
+]

qtype/interpreter/stream/utils/__init__.py

@@ -0,0 +1,15 @@
+from __future__ import annotations
+
+from .build_vercel_ai_formatter import build_vercel_ai_formatter
+from .callback_to_stream import callback_to_async_iterator
+from .create_streaming_response import create_streaming_response
+from .default_chat_extract_text import default_chat_extract_text
+from .error_streaming_response import error_streaming_response
+
+__all__ = [
+    "build_vercel_ai_formatter",
+    "callback_to_async_iterator",
+    "create_streaming_response",
+    "error_streaming_response",
+    "default_chat_extract_text",
+]

qtype/interpreter/stream/utils/build_vercel_ai_formatter.py

@@ -0,0 +1,74 @@
+from __future__ import annotations
+
+import logging
+import uuid
+from collections.abc import Callable, Generator, Iterable
+from concurrent.futures import Future
+from typing import Any
+
+from qtype.interpreter.stream.chat.vercel import (
+    ErrorChunk,
+    FinishChunk,
+    StartChunk,
+    TextDeltaChunk,
+    TextEndChunk,
+    TextStartChunk,
+)
+
+logger = logging.getLogger(__name__)
+
+
+def build_vercel_ai_formatter(
+    stream_generator: Iterable[tuple[Any, Any]],
+    result_future: Future,
+    extract_text: Callable[[Any], str],
+) -> Generator[str, None, None]:
+    """
+    Convert a low-level stream of (step, message) into Vercel AI UI protocol SSE.
+
+    Args:
+        stream_generator: Iterable yielding (step, message) pairs.
+        result_future: Future representing completion of the flow.
+        extract_text: Function to extract textual content from a message object.
+
+    Yields:
+        Lines formatted as 'data: {...}\\n\\n' for SSE.
+    """
+    start_chunk = StartChunk(messageId=str(uuid.uuid4()))  # type: ignore[arg-type]
+    yield f"data: {start_chunk.model_dump_json(by_alias=True, exclude_none=True)}\n\n"
+
+    text_id = str(uuid.uuid4())
+    text_started = False
+
+    for _step, message in stream_generator:
+        try:
+            content = extract_text(message)
+        except Exception as exc:  # Defensive; continue streaming
+            logger.debug(
+                "Failed extracting text from message: %s", exc, exc_info=True
+            )
+            continue
+
+        if not content or not content.strip():
+            continue
+
+        if not text_started:
+            text_start = TextStartChunk(id=text_id)
+            yield f"data: {text_start.model_dump_json(by_alias=True, exclude_none=True)}\n\n"
+            text_started = True
+
+        text_delta = TextDeltaChunk(id=text_id, delta=content)
+        yield f"data: {text_delta.model_dump_json(by_alias=True, exclude_none=True)}\n\n"
+
+    if text_started:
+        text_end = TextEndChunk(id=text_id)
+        yield f"data: {text_end.model_dump_json(by_alias=True, exclude_none=True)}\n\n"
+
+    try:
+        result_future.result(timeout=5.0)
+        finish_chunk = FinishChunk()
+        yield f"data: {finish_chunk.model_dump_json(by_alias=True, exclude_none=True)}\n\n"
+    except Exception as exc:
+        logger.error("Error finalizing flow execution: %s", exc, exc_info=True)
+        error_chunk = ErrorChunk(errorText=str(exc))
+        yield f"data: {error_chunk.model_dump_json(by_alias=True, exclude_none=True)}\n\n"