agentengine-sdk-python 0.7.0__py3-none-any.whl → 0.8.1__py3-none-any.whl

ksadk/server/routes/models.py

@@ -0,0 +1,451 @@
+"""Runtime HTTP request models and pure projection helpers."""
+
+from __future__ import annotations
+
+import os
+from datetime import datetime, timezone
+from typing import Any, Dict, List, Mapping, Optional
+
+from fastapi import HTTPException
+from pydantic import BaseModel, Field, field_validator
+
+from ksadk.conversations.run_kinds import RUN_MODE_UNKNOWN, RUN_TRIGGER_UNKNOWN
+from ksadk.conversations.run_status import RUN_STATUS_ACTIVE, RUN_STATUS_TERMINAL
+from ksadk.runtime.launch import RuntimeLaunchContext
+from ksadk.sessions import SessionEvent
+
+_RUN_TERMINAL_STATUSES = RUN_STATUS_TERMINAL
+_RUN_ACTIVE_STATUSES = RUN_STATUS_ACTIVE
+_EVENT_SCAN_PAGE_SIZE = 500
+_MAX_PREVIEW_TOOL_RECEIPTS = 500
+_RUNTIME_RUN_STATUS_BY_EVENT_TYPE = {
+    "run.started": "in_progress",
+    "run.progress": "in_progress",
+    "run.interrupted": "interrupted",
+    "run.completed": "completed",
+    "run.failed": "failed",
+    "run.canceled": "cancelled",
+}
+
+
+def _parse_iso_datetime(value: Any) -> datetime | None:
+    raw = str(value or "").strip()
+    if not raw:
+        return None
+    if raw.endswith("Z"):
+        raw = f"{raw[:-1]}+00:00"
+    try:
+        parsed = datetime.fromisoformat(raw)
+    except ValueError:
+        return None
+    if parsed.tzinfo is None:
+        return parsed.replace(tzinfo=timezone.utc)
+    return parsed.astimezone(timezone.utc)
+
+
+class UiBootstrapRequest(BaseModel):
+    AgentId: Optional[str] = None
+    SessionId: Optional[str] = None
+
+
+class CreateSessionActionRequest(BaseModel):
+    AgentId: str
+    UserId: Optional[str] = "user"
+    SessionId: Optional[str] = None
+
+
+class ListSessionsActionRequest(BaseModel):
+    AgentId: str
+    UserId: Optional[str] = None
+    Page: int = Field(1, ge=1)
+    PageSize: int = Field(20, ge=1, le=200)
+
+
+class SessionIdRequest(BaseModel):
+    SessionId: str
+    AgentId: Optional[str] = None
+    UserId: Optional[str] = None
+
+
+class ListSessionEventsActionRequest(BaseModel):
+    AgentId: Optional[str] = None
+    SessionId: Optional[str] = None
+    UserId: Optional[str] = None
+    Offset: Optional[int] = Field(None, ge=0)
+    Limit: int = Field(200, ge=1, le=2000)
+    AfterSeqId: Optional[int] = Field(None, ge=0)
+    BeforeSeqId: Optional[int] = Field(None, ge=1)
+
+
+class ListSessionMessagesActionRequest(BaseModel):
+    AgentId: Optional[str] = None
+    UserId: Optional[str] = None
+    SessionId: str
+    AfterSeqId: Optional[int] = Field(None, ge=0)
+    BeforeSeqId: Optional[int] = Field(None, ge=1)
+    Limit: int = Field(50, ge=1, le=200)
+    IncludeReasoning: bool = False
+    IncludeToolEvents: bool = False
+    IncludeAttachments: bool = True
+
+
+class ListSessionCheckpointsActionRequest(BaseModel):
+    AgentId: str
+    SessionId: Optional[str] = None
+    UserId: Optional[str] = None
+    RunId: Optional[str] = None
+    OnlyResumable: bool = False
+    Framework: Optional[str] = None
+    Offset: Optional[int] = Field(None, ge=0)
+    Limit: int = Field(100, ge=1, le=500)
+
+
+class ListToolReceiptsActionRequest(BaseModel):
+    AgentId: str
+    UserId: Optional[str] = None
+    SessionId: str
+    RunId: Optional[str] = None
+    CheckpointId: Optional[str] = None
+    Offset: int = Field(0, ge=0)
+    Limit: int = Field(200, ge=1, le=500)
+
+
+class ResumeRunActionRequest(BaseModel):
+    AgentId: str
+    UserId: Optional[str] = None
+    SessionId: str
+    RunId: str
+    CheckpointId: str
+    ResumeAttemptId: Optional[str] = None
+    InvocationId: Optional[str] = None
+    Stream: bool = False
+    Model: Optional[str] = None
+    ModelMetadata: Optional[Dict[str, Any]] = None
+    ModelOptions: Optional[Dict[str, Any]] = None
+    Metadata: Optional[Dict[str, Any]] = None
+    ResumeInstructionEnabled: bool = False
+    ResumeInstruction: Optional[str] = None
+
+
+class GetCheckpointResumePreviewActionRequest(BaseModel):
+    AgentId: str
+    UserId: Optional[str] = None
+    SessionId: str
+    RunId: str
+    CheckpointId: str
+
+
+class RunAgentActionRequest(BaseModel):
+    AgentId: str
+    Messages: List[Dict[str, Any]] = Field(default_factory=list)
+    UserId: Optional[str] = "user"
+    AccountId: Optional[str] = None
+    SessionId: Optional[str] = None
+    InvocationId: Optional[str] = None
+    ApiFormat: str = "responses"
+    Stream: bool = False
+    Background: bool = False  # 立即返回 job 句柄，后台执行，进度走 SubscribeRunEvents
+    Model: Optional[str] = None
+    ModelMetadata: Optional[Dict[str, Any]] = None
+    ModelOptions: Optional[Dict[str, Any]] = None
+    Metadata: Optional[Dict[str, Any]] = None
+    ResponsesInput: Optional[Any] = None
+    PreviousResponseId: Optional[str] = None
+
+
+class ResponseFeedbackRefActionRequest(BaseModel):
+    AgentId: str
+    SessionId: str
+    ResponseId: str
+
+
+class UpsertResponseFeedbackActionRequest(ResponseFeedbackRefActionRequest):
+    Rating: str
+    Comment: Optional[str] = ""
+    EventId: Optional[str] = None
+    TraceId: Optional[str] = None
+    RootSpanId: Optional[str] = None
+
+
+class ResponsesRequest(BaseModel):
+    input: Any
+    model: Optional[str] = None
+    model_metadata: Optional[Dict[str, Any]] = None
+    model_options: Optional[Dict[str, Any]] = None
+    instructions: Optional[str] = None
+    metadata: Optional[Dict[str, Any]] = None
+    conversation: Optional[Any] = None
+    safety_identifier: Optional[str] = None
+    prompt_cache_key: Optional[str] = None
+    user: Optional[str] = None
+    account_id: Optional[str] = None
+    store: Optional[bool] = None
+    previous_response_id: Optional[str] = None
+    stream: bool = False
+    session_id: Optional[str] = None
+
+    @field_validator("metadata")
+    @classmethod
+    def validate_metadata(cls, value: Optional[Dict[str, Any]]) -> Optional[Dict[str, Any]]:
+        if value is None:
+            return None
+        public_items = [(key, item) for key, item in value.items() if key != "agentengine"]
+        if len(public_items) > 16:
+            raise ValueError("Responses metadata supports at most 16 key-value pairs")
+        for key, item in public_items:
+            if len(key) > 64:
+                raise ValueError("Responses metadata keys must be at most 64 characters")
+            if not isinstance(item, str):
+                raise ValueError("Responses metadata values must be strings")
+            if len(item) > 512:
+                raise ValueError("Responses metadata values must be at most 512 characters")
+        return value
+
+
+class WorkspaceListActionRequest(BaseModel):
+    AgentId: Optional[str] = None
+    Path: str = "."
+    Recursive: bool = False
+
+
+def _clean_optional_string(value: Any) -> str | None:
+    if value is None:
+        return None
+    text = str(value).strip()
+    return text or None
+
+
+def _resolve_responses_conversation_id(conversation_value: Any) -> str | None:
+    if conversation_value is None:
+        return None
+    if isinstance(conversation_value, str):
+        return _clean_optional_string(conversation_value)
+    if isinstance(conversation_value, Mapping):
+        return _clean_optional_string(conversation_value.get("id"))
+    raise HTTPException(
+        status_code=400,
+        detail="Responses field 'conversation' must be a string or an object with an 'id'.",
+    )
+
+
+def _resolve_responses_session_and_user(request: ResponsesRequest) -> tuple[str | None, str]:
+    conversation_id = _resolve_responses_conversation_id(request.conversation)
+    legacy_session_id = _clean_optional_string(request.session_id)
+
+    if conversation_id and legacy_session_id and conversation_id != legacy_session_id:
+        raise HTTPException(
+            status_code=400,
+            detail=(
+                "Responses field 'conversation' conflicts with ksadk legacy field "
+                "'session_id'. Use 'conversation' for OpenAI-compatible calls."
+            ),
+        )
+    if conversation_id and request.previous_response_id:
+        raise HTTPException(
+            status_code=400,
+            detail=(
+                "Responses fields 'conversation' and 'previous_response_id' cannot be "
+                "used together."
+            ),
+        )
+
+    resolved_session_id = conversation_id or legacy_session_id
+    resolved_user_id = (
+        _clean_optional_string(request.safety_identifier)
+        or _clean_optional_string(request.user)
+        or "user"
+    )
+    return resolved_session_id, resolved_user_id
+
+
+def _runtime_agent_id(launch_context: RuntimeLaunchContext) -> str:
+    runtime_id = _clean_optional_string(os.getenv("AGENT_RUNTIME_ID"))
+    if runtime_id:
+        return runtime_id
+    return str(getattr(launch_context.detection, "name", "") or "agent")
+
+
+def _metadata_invocation_id(metadata: Mapping[str, Any] | None) -> str | None:
+    if not isinstance(metadata, Mapping):
+        return None
+    agentengine_metadata = metadata.get("agentengine")
+    if not isinstance(agentengine_metadata, Mapping):
+        return None
+    return _clean_optional_string(agentengine_metadata.get("invocation_id"))
+
+
+def _split_custom_metadata(
+    metadata: Mapping[str, Any] | None,
+) -> tuple[dict[str, Any], dict[str, Any]]:
+    public_metadata = dict(metadata or {})
+    runtime_metadata: dict[str, Any] = {}
+    agentengine_metadata = public_metadata.get("agentengine")
+    public_metadata.pop("agentengine", None)
+    if isinstance(agentengine_metadata, Mapping):
+        runtime_metadata["agentengine"] = dict(agentengine_metadata)
+        # The approval profile is a small, validated-by-the-agent runtime control.
+        # It must not be mixed into caller-visible public metadata.
+        tool_approval_mode = agentengine_metadata.get("tool_approval_mode")
+        if isinstance(tool_approval_mode, str):
+            runtime_metadata["tool_approval_mode"] = tool_approval_mode
+    return public_metadata, runtime_metadata
+
+
+def _run_agent_response_metadata(
+    custom_metadata: Mapping[str, Any] | None,
+    result: Mapping[str, Any] | None,
+) -> dict[str, Any]:
+    response_metadata = dict(custom_metadata or {})
+    result_metadata = result.get("metadata") if isinstance(result, Mapping) else None
+    if isinstance(result_metadata, Mapping):
+        agentengine_metadata = result_metadata.get("agentengine")
+        if isinstance(agentengine_metadata, Mapping):
+            response_metadata["agentengine"] = dict(agentengine_metadata)
+    return response_metadata
+
+
+def _event_text(event: SessionEvent) -> str:
+    parts = (event.content or {}).get("parts")
+    if isinstance(parts, list):
+        text_parts: list[str] = []
+        for part in parts:
+            if isinstance(part, Mapping):
+                text_parts.append(str(part.get("text") or ""))
+            else:
+                text_parts.append(str(part or ""))
+        text = "".join(text_parts).strip()
+        if text:
+            return text
+    return str((event.content or {}).get("text") or "").strip()
+
+
+def _truncate_session_text(text: str, limit: int = 512) -> str:
+    normalized = " ".join(str(text or "").strip().split())
+    if len(normalized) <= limit:
+        return normalized
+    return f"{normalized[: max(limit - 1, 0)].rstrip()}…"
+
+
+def _session_user_prompt_from_event(event: SessionEvent) -> str:
+    metadata = event.metadata or {}
+    content = event.content or {}
+    return str(
+        metadata.get("agent_input")
+        or metadata.get("user_input")
+        or content.get("agent_input")
+        or _event_text(event)
+        or ""
+    ).strip()
+
+
+def _run_status_payload_status(event: SessionEvent) -> str:
+    legacy_status = str(
+        (event.metadata or {}).get("status")
+        or (event.metadata or {}).get("run_status")
+        or (event.content or {}).get("status")
+        or ""
+    ).strip()
+    if legacy_status:
+        return legacy_status
+    runtime_payload = (event.content or {}).get("payload")
+    runtime_status = (
+        str(runtime_payload.get("status") or "").strip()
+        if isinstance(runtime_payload, Mapping)
+        else ""
+    )
+    return runtime_status or _RUNTIME_RUN_STATUS_BY_EVENT_TYPE.get(event.event_type, "")
+
+
+def _is_run_lifecycle_event(event: SessionEvent) -> bool:
+    return event.event_type == "run_status" or event.event_type in _RUNTIME_RUN_STATUS_BY_EVENT_TYPE
+
+
+def _event_run_id(event: SessionEvent) -> str:
+    return str(
+        (event.metadata or {}).get("run_id")
+        or (event.metadata or {}).get("invocation_id")
+        or event.invocation_id
+        or ""
+    ).strip()
+
+
+def _session_topic_from_events(events: list[SessionEvent]) -> str:
+    for event in reversed(events):
+        metadata = event.metadata or {}
+        tool_output = metadata.get("tool_output")
+        if isinstance(tool_output, Mapping):
+            topic = str(tool_output.get("topic") or tool_output.get("research_title") or "").strip()
+            if topic:
+                return topic
+        topic = str(metadata.get("research_title") or metadata.get("task_title") or "").strip()
+        if topic:
+            return topic
+    return ""
+
+
+def _latest_session_run_status(events: list[SessionEvent]) -> tuple[str, str]:
+    latest_by_invocation: dict[str, tuple[str, SessionEvent]] = {}
+    for event in reversed(events):
+        if not _is_run_lifecycle_event(event):
+            continue
+        status = _run_status_payload_status(event)
+        invocation_id = _event_run_id(event)
+        if status or invocation_id:
+            latest_by_invocation.setdefault(invocation_id, (status, event))
+    for invocation_id, (status, _) in latest_by_invocation.items():
+        if status in _RUN_ACTIVE_STATUSES:
+            return invocation_id, status
+    for invocation_id, (status, _) in latest_by_invocation.items():
+        if status not in _RUN_TERMINAL_STATUSES:
+            return invocation_id, status
+    if latest_by_invocation:
+        invocation_id, (status, _) = next(iter(latest_by_invocation.items()))
+        return invocation_id, status
+    for event in reversed(events):
+        invocation_id = _event_run_id(event)
+        if invocation_id:
+            return invocation_id, ""
+    return "", ""
+
+
+def _latest_session_run_metadata(
+    events: list[SessionEvent],
+) -> tuple[str, str, str, str]:
+    """返回 (invocation_id, status, run_mode, run_trigger)。
+
+    与 _latest_session_run_status 同语义，但额外从最新 run_status 事件的 metadata
+    读取 run_mode/run_trigger。旧事件缺字段降级 unknown。原 _latest_session_run_status
+    不动，保护现有 ActiveInvocationId/ActiveRunStatus 契约。
+    """
+    invocation_id, status = _latest_session_run_status(events)
+    run_mode = RUN_MODE_UNKNOWN
+    run_trigger = RUN_TRIGGER_UNKNOWN
+    if invocation_id:
+        for event in reversed(events):
+            if not _is_run_lifecycle_event(event) or _event_run_id(event) != invocation_id:
+                continue
+            metadata = event.metadata or {}
+            candidate_mode = str(metadata.get("run_mode") or RUN_MODE_UNKNOWN)
+            candidate_trigger = str(metadata.get("run_trigger") or RUN_TRIGGER_UNKNOWN)
+            # Canonical RuntimeEvents and their legacy ``run_status`` projection
+            # coexist in the session log. RuntimeEvents intentionally carry only
+            # the transport-neutral marker, while the projection owns run kind
+            # metadata. Keep scanning past an unclassified RuntimeEvent so the UI
+            # does not downgrade a known background/foreground run to ``unknown``.
+            if candidate_mode != RUN_MODE_UNKNOWN or candidate_trigger != RUN_TRIGGER_UNKNOWN:
+                run_mode = candidate_mode
+                run_trigger = candidate_trigger
+                break
+    return invocation_id, status, run_mode, run_trigger
+
+
+class WorkspaceDeleteActionRequest(BaseModel):
+    AgentId: Optional[str] = None
+    Path: str
+
+
+class CancelRunActionRequest(BaseModel):
+    AgentId: Optional[str] = None
+    UserId: Optional[str] = None
+    SessionId: Optional[str] = None
+    InvocationId: str

ksadk/server/routes/openai_compat.py

@@ -0,0 +1,229 @@
+"""OpenAI Responses and Chat Completions compatibility routes."""
+
+from __future__ import annotations
+
+import uuid
+from collections.abc import Mapping
+from typing import Any, Dict, List, Optional
+
+from fastapi.responses import StreamingResponse
+from pydantic import BaseModel
+
+from ksadk.conversations.normalize import normalize_kop_messages, normalize_responses_input
+from ksadk.conversations.run_kinds import RUN_MODE_FOREGROUND, trigger_from_resume_input
+from ksadk.conversations.runtime_payloads import (
+    build_chat_completions_payload,
+    build_responses_payload,
+    extract_responses_resume_input,
+)
+from ksadk.conversations.runtime_streaming import (
+    stream_runtime_conversation_turn,
+    stream_runtime_responses_conversation_turn,
+)
+from ksadk.runtime.conversation_execution import invoke_runtime_conversation_once
+from ksadk.server.factory import get_runtime_execution
+
+from . import dependencies as deps
+from .checkpoint_resolution import _resolve_checkpoint_resume_input_from_session
+from .models import (
+    ResponsesRequest,
+    _clean_optional_string,
+    _metadata_invocation_id,
+    _resolve_responses_session_and_user,
+    _runtime_agent_id,
+    _split_custom_metadata,
+)
+from .routers import openai_compat_router
+from .streaming import (
+    _detached_resume_key_from_input,
+    _reject_if_detached_resume_active,
+)
+from .workspace import _build_models_payload
+
+
+class ChatCompletionRequest(BaseModel):
+    messages: List[Dict[str, Any]]
+    model: Optional[str] = None
+    model_metadata: Optional[Dict[str, Any]] = None
+    model_options: Optional[Dict[str, Any]] = None
+    stream: bool = False
+    session_id: Optional[str] = None
+    user: Optional[str] = None
+    account_id: Optional[str] = None
+    metadata: Optional[Dict[str, Any]] = None
+    temperature: Optional[float] = 0.7
+    max_tokens: Optional[int] = None
+
+
+@openai_compat_router.get("/v1/models")
+async def list_openai_models():
+    """Expose the current model catalog through the OpenAI-compatible path."""
+
+    payload = await _build_models_payload()
+    return {
+        "object": "list",
+        "data": payload.get("data", []),
+        "current": payload.get("current"),
+        "source": payload.get("source", ""),
+    }
+
+
+@openai_compat_router.post("/v1/responses")
+async def responses(request: ResponsesRequest):
+    """OpenAI Responses 兼容接口。"""
+    executor, launch_context = get_runtime_execution()
+    resolved_session_id, resolved_user_id = _resolve_responses_session_and_user(request)
+    agent_id = _runtime_agent_id(launch_context)
+
+    resume_input = extract_responses_resume_input(request.input)
+    resume_input = await _resolve_checkpoint_resume_input_from_session(
+        service=deps.resolve_session_service(),
+        agent_id=agent_id,
+        session_id=resolved_session_id,
+        resume_input=resume_input,
+    )
+    messages = (
+        []
+        if resume_input is not None
+        else normalize_responses_input(request.input)
+    )
+    custom_metadata, request_metadata = _split_custom_metadata(request.metadata)
+    if request.previous_response_id:
+        request_metadata["previous_response_id"] = request.previous_response_id
+    if request.prompt_cache_key:
+        request_metadata["prompt_cache_key"] = request.prompt_cache_key
+    if request.safety_identifier:
+        request_metadata["safety_identifier"] = request.safety_identifier
+    if request.user:
+        request_metadata["user"] = request.user
+    if request.conversation is not None:
+        request_metadata["conversation"] = request.conversation
+    if request.store is not None:
+        request_metadata["store"] = request.store
+    account_id = _clean_optional_string(request.account_id)
+    invocation_id = _metadata_invocation_id(request_metadata)
+
+    if request.stream:
+        runtime_preparation = (
+            None
+            if resume_input is not None
+            else await executor.prepare_start(launch_context)
+        )
+        resume_key = _detached_resume_key_from_input(resolved_session_id, resume_input)
+        _reject_if_detached_resume_active(resume_key)
+        return deps.detached_streaming_response(
+            stream_runtime_responses_conversation_turn(
+                executor=executor,
+                launch_context=launch_context,
+                agent_id=agent_id,
+                user_id=resolved_user_id,
+                messages=messages,
+                session_id=resolved_session_id,
+                model=request.model,
+                model_metadata=request.model_metadata,
+                model_options=request.model_options,
+                instructions=request.instructions,
+                request_metadata=request_metadata,
+                custom_metadata=custom_metadata,
+                resume_input=resume_input,
+                account_id=account_id,
+                invocation_id=invocation_id,
+                session_service_provider=deps.resolve_session_service,
+                run_mode=RUN_MODE_FOREGROUND,
+                runtime_preparation=runtime_preparation,
+            ),
+            invocation_id=invocation_id,
+            resume_key=resume_key,
+            run_mode=RUN_MODE_FOREGROUND,
+            run_trigger=trigger_from_resume_input(resume_input),
+        )
+
+    response_id = f"resp_{uuid.uuid4().hex}"
+    resolved_session_id, result = await invoke_runtime_conversation_once(
+        executor=executor,
+        launch_context=launch_context,
+        agent_id=agent_id,
+        user_id=resolved_user_id,
+        messages=messages,
+        session_id=resolved_session_id,
+        model=request.model,
+        model_metadata=request.model_metadata,
+        model_options=request.model_options,
+        instructions=request.instructions,
+        request_metadata=request_metadata,
+        custom_metadata=custom_metadata,
+        resume_input=resume_input,
+        response_id=response_id,
+        account_id=account_id,
+        invocation_id=invocation_id,
+        session_service_provider=deps.resolve_session_service,
+        run_mode=RUN_MODE_FOREGROUND,
+    )
+    return build_responses_payload(
+        output_text=result["output_text"],
+        model=request.model,
+        session_id=resolved_session_id,
+        response_id=response_id,
+        metadata=custom_metadata,
+        usage=result.get("usage") if isinstance(result.get("usage"), Mapping) else None,
+    )
+
+
+@openai_compat_router.post("/v1/chat/completions")
+async def chat_completions(request: ChatCompletionRequest):
+    """OpenAI 兼容的聊天补全接口 (支持流式和非流式)"""
+    executor, launch_context = get_runtime_execution()
+    messages = normalize_kop_messages(request.messages)
+    agent_id = _runtime_agent_id(launch_context)
+    resolved_user_id = _clean_optional_string(request.user) or "user"
+    account_id = _clean_optional_string(request.account_id)
+    custom_metadata, request_metadata = _split_custom_metadata(request.metadata)
+    invocation_id = _metadata_invocation_id(request_metadata)
+
+    if request.stream:
+        runtime_preparation = await executor.prepare_start(launch_context)
+        return StreamingResponse(
+            stream_runtime_conversation_turn(
+                executor=executor,
+                launch_context=launch_context,
+                agent_id=agent_id,
+                user_id=resolved_user_id,
+                messages=messages,
+                session_id=request.session_id,
+                model=request.model,
+                model_metadata=request.model_metadata,
+                model_options=request.model_options,
+                request_metadata=request_metadata,
+                custom_metadata=custom_metadata,
+                invocation_id=invocation_id,
+                account_id=account_id,
+                session_service_provider=deps.resolve_session_service,
+                run_mode=RUN_MODE_FOREGROUND,
+                runtime_preparation=runtime_preparation,
+            ),
+            media_type="text/event-stream",
+        )
+
+    resolved_session_id, result = await invoke_runtime_conversation_once(
+        executor=executor,
+        launch_context=launch_context,
+        agent_id=agent_id,
+        user_id=resolved_user_id,
+        messages=messages,
+        session_id=request.session_id,
+        model=request.model,
+        model_metadata=request.model_metadata,
+        model_options=request.model_options,
+        request_metadata=request_metadata,
+        custom_metadata=custom_metadata,
+        invocation_id=invocation_id,
+        account_id=account_id,
+        session_service_provider=deps.resolve_session_service,
+        run_mode=RUN_MODE_FOREGROUND,
+    )
+    return build_chat_completions_payload(
+        output_text=result["output_text"],
+        model=request.model,
+        session_id=resolved_session_id,
+        metadata=result.get("metadata"),
+    )