by-framework 0.2.2.dev3__tar.gz → 0.2.2.dev4__tar.gz

{by_framework-0.2.2.dev3 → by_framework-0.2.2.dev4}/.github/workflows/publish.yml

@@ -8,6 +8,8 @@ on:
       - "by-framework-history-byclaw-v*"
       - "by-framework-trace-langfuse-v*"
       - "by-framework-trace-phoenix-v*"
+      - "by-framework-trace-query-v*"
+      - "by-framework-dashboard-v*"
       - "by-framework-langgraph-v*"
       - "by-framework-adk-v*"
 
@@ -54,6 +56,16 @@ jobs:
               echo "package_dir=libs/by-framework-trace-phoenix" >> "$GITHUB_OUTPUT"
               echo "version=${REF_NAME#by-framework-trace-phoenix-v}" >> "$GITHUB_OUTPUT"
               ;;
+            by-framework-trace-query-v*)
+              echo "package_name=by-framework-trace-query" >> "$GITHUB_OUTPUT"
+              echo "package_dir=libs/by-framework-trace-query" >> "$GITHUB_OUTPUT"
+              echo "version=${REF_NAME#by-framework-trace-query-v}" >> "$GITHUB_OUTPUT"
+              ;;
+            by-framework-dashboard-v*)
+              echo "package_name=by-framework-dashboard" >> "$GITHUB_OUTPUT"
+              echo "package_dir=libs/by-framework-dashboard" >> "$GITHUB_OUTPUT"
+              echo "version=${REF_NAME#by-framework-dashboard-v}" >> "$GITHUB_OUTPUT"
+              ;;
             by-framework-langgraph-v*)
               echo "package_name=by-framework-langgraph" >> "$GITHUB_OUTPUT"
               echo "package_dir=libs/by-framework-langgraph" >> "$GITHUB_OUTPUT"

{by_framework-0.2.2.dev3 → by_framework-0.2.2.dev4}/.gitignore

@@ -68,8 +68,6 @@ Thumbs.db
 .claude/settings.local.json
 .claude/settings.json
 
-core
-
 .coderfleet-uploads/
 
 .claude/skills

{by_framework-0.2.2.dev3 → by_framework-0.2.2.dev4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: by-framework
-Version: 0.2.2.dev3
+Version: 0.2.2.dev4
 Summary: 分布式 Agent 调度框架
 License-File: LICENSE
 Requires-Python: >=3.12

{by_framework-0.2.2.dev3 → by_framework-0.2.2.dev4}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "by-framework"
-version = "0.2.2.dev3"
+version = "0.2.2.dev4"
 description = "分布式 Agent 调度框架"
 readme = "README.md"
 requires-python = ">=3.12"

{by_framework-0.2.2.dev3 → by_framework-0.2.2.dev4}/src/by_framework/client/client.py

@@ -96,6 +96,16 @@ class GatewayClient:
         )
         self.interceptors = interceptors or []
         self.span_recorder = span_recorder or SpanRecorder(self.redis)
+        self._langfuse_dispatch_fn = self._resolve_langfuse_dispatch_fn()
+
+    @staticmethod
+    def _resolve_langfuse_dispatch_fn() -> Any:
+        try:
+            from by_framework_trace_langfuse import start_client_dispatch_observation
+
+            return start_client_dispatch_observation
+        except ImportError:
+            return None
 
     def add_interceptor(self, interceptor: GatewayInterceptor):
         self.interceptors.append(interceptor)
@@ -619,7 +629,7 @@ class GatewayClient:
             )
 
         langfuse_client_dispatch = None
-        if not params["parent_message_id"]:
+        if not langfuse_parent_observation_id:
             langfuse_client_dispatch = self._start_langfuse_client_dispatch_observation(
                 trace_id=trace_id,
                 message_id=message_id,
@@ -631,8 +641,7 @@ class GatewayClient:
                 metadata=metadata,
             )
             observation_id = getattr(langfuse_client_dispatch, "id", "")
-            if observation_id:
-                langfuse_parent_observation_id = observation_id
+            langfuse_parent_observation_id = observation_id or trace_parent_span_id
 
         header = MessageHeader(
             message_id=message_id,
@@ -860,10 +869,10 @@ class GatewayClient:
         content: Any,
         metadata: Dict[str, Any],
     ) -> Any:
+        if self._langfuse_dispatch_fn is None:
+            return None
         try:
-            from by_framework_trace_langfuse import start_client_dispatch_observation
-
-            return start_client_dispatch_observation(
+            return self._langfuse_dispatch_fn(
                 trace_id=trace_id,
                 message_id=message_id,
                 target_agent_type=target_agent_type,
@@ -877,7 +886,6 @@ class GatewayClient:
             logger.warning(
                 "Langfuse client.dispatch observation skipped: %s",
                 err,
-                exc_info=True,
             )
             return None
 

by_framework-0.2.2.dev4/src/by_framework/core/__init__.py

@@ -0,0 +1,69 @@
+"""
+Core module of the Gateway SDK.
+
+Provides core components including protocol definitions, worker registry,
+and workspace management.
+"""
+
+from .availability import (AvailabilityResult, AvailabilityRouter,
+                           AvailabilityStatus, DeliveryIntent, PendingDelivery,
+                           RoutePolicy, WakeupDecision, WakeupDecisionStatus,
+                           WakeupRequest)
+from .delivery_gate import DeliveryGate
+from .protocol import (ActionType, AgentState, ArtifactEvent, AskAgentCommand,
+                       AskUserEvent, BaiYingMessage, BaiYingMessageRole,
+                       BaseCommand, CancelTaskCommand, CancelTaskResponse,
+                       DataMessage, EventType, GatewayCommand, MessageContent,
+                       MessageFile, MessageHeader, Resource, ResumeCommand,
+                       SendMessageResponse, SseMessageType,
+                       SseReasonMessageType, StateChangeEvent,
+                       StreamChunkEvent, command_from_dict,
+                       get_registered_command, register_command,
+                       unregister_command)
+from .registry import WorkerRegistry
+from .wakeup_controller import WakeupController, WakeupProvider
+from .workspace import WorkspaceManager
+
+__all__ = [
+    "SendMessageResponse",
+    "CancelTaskResponse",
+    "BaiYingMessage",
+    "BaiYingMessageRole",
+    "MessageContent",
+    "MessageFile",
+    "Resource",
+    "DataMessage",
+    "ActionType",
+    "AgentState",
+    "EventType",
+    "StateChangeEvent",
+    "StreamChunkEvent",
+    "ArtifactEvent",
+    "AskUserEvent",
+    "SseMessageType",
+    "SseReasonMessageType",
+    "MessageHeader",
+    "BaseCommand",
+    "AskAgentCommand",
+    "ResumeCommand",
+    "CancelTaskCommand",
+    "GatewayCommand",
+    "command_from_dict",
+    "register_command",
+    "unregister_command",
+    "get_registered_command",
+    "WorkerRegistry",
+    "WorkspaceManager",
+    "AvailabilityRouter",
+    "AvailabilityStatus",
+    "AvailabilityResult",
+    "DeliveryIntent",
+    "RoutePolicy",
+    "PendingDelivery",
+    "WakeupRequest",
+    "WakeupDecision",
+    "WakeupDecisionStatus",
+    "WakeupController",
+    "WakeupProvider",
+    "DeliveryGate",
+]

by_framework-0.2.2.dev4/src/by_framework/core/availability.py

@@ -0,0 +1,495 @@
+"""Agent availability control-plane routing.
+
+This module centralizes online-worker checks and wakeup handshakes so client
+calls and inter-agent calls share the same offline routing behavior.
+"""
+
+from __future__ import annotations
+
+import json
+import time
+from dataclasses import asdict, dataclass, field
+from typing import Any, Optional
+
+from by_framework.common.constants import RedisKeys
+from by_framework.common.redis_client import Redis
+from by_framework.core.registry import WorkerRegistry, check_agent_type_online
+
+
+class RoutePolicy:
+    """Policy values for routing and availability checks."""
+
+    FAIL_FAST = "FAIL_FAST"
+    SEND_ANYWAY = "SEND_ANYWAY"
+    WAKE_AND_WAIT = "WAKE_AND_WAIT"
+    WAKE_AND_QUEUE = "WAKE_AND_QUEUE"
+    QUEUE_ONLY = "QUEUE_ONLY"
+
+
+class WakeupDecisionStatus:
+    """Wakeup controller decision status values."""
+
+    READY = "READY"
+    STARTING = "STARTING"
+    QUEUED = "QUEUED"
+    REJECTED = "REJECTED"
+    FAILED = "FAILED"
+    FALLBACK = "FALLBACK"
+
+
+class AvailabilityStatus:
+    """Availability router result values."""
+
+    DELIVER_NOW = "DELIVER_NOW"
+    WAIT_AND_DELIVER = "WAIT_AND_DELIVER"
+    QUEUE_PENDING = "QUEUE_PENDING"
+    REJECT = "REJECT"
+    FALLBACK_TO_OTHER_AGENT_TYPE = "FALLBACK_TO_OTHER_AGENT_TYPE"
+
+
+@dataclass(frozen=True)
+class DeliveryIntent:
+    """Framework-internal representation of a control-message delivery."""
+
+    execution_id: str
+    message_id: str
+    session_id: str
+    trace_id: str
+    source: str
+    target_agent_type: str
+    user_code: str = ""
+    region: str = ""
+    priority: int = 0
+    policy: str = RoutePolicy.FAIL_FAST
+    timeout_ms: int = 30000
+    command_payload: dict[str, Any] = field(default_factory=dict)
+    metadata: dict[str, Any] = field(default_factory=dict)
+
+
+@dataclass(frozen=True)
+class WakeupRequest:
+    """Redis management event emitted when a target agent type is unavailable."""
+
+    execution_id: str
+    target_agent_type: str
+    session_id: str
+    trace_id: str
+    message_id: str
+    source: str
+    policy: str
+    timeout_ms: int
+    user_code: str = ""
+    region: str = ""
+    priority: int = 0
+    metadata: dict[str, Any] = field(default_factory=dict)
+    command_payload: dict[str, Any] = field(default_factory=dict)
+
+    def to_redis_payload(self) -> dict[str, str]:
+        return {"data": json.dumps(asdict(self))}
+
+    @classmethod
+    def from_dict(cls, data: dict[str, Any]) -> "WakeupRequest":
+        return cls(
+            execution_id=str(data.get("execution_id", "")),
+            target_agent_type=str(data.get("target_agent_type", "")),
+            session_id=str(data.get("session_id", "")),
+            trace_id=str(data.get("trace_id", "")),
+            message_id=str(data.get("message_id", "")),
+            source=str(data.get("source", "")),
+            policy=str(data.get("policy", RoutePolicy.FAIL_FAST)),
+            timeout_ms=int(data.get("timeout_ms", 30000)),
+            user_code=str(data.get("user_code") or data.get("tenant_id", "")),
+            region=str(data.get("region", "")),
+            priority=int(data.get("priority", 0)),
+            metadata=dict(data.get("metadata", {})),
+            command_payload=dict(data.get("command_payload", {})),
+        )
+
+
+@dataclass(frozen=True)
+class PendingDelivery:
+    """Control command held until a wakeup decision allows dispatch."""
+
+    execution_id: str
+    message_id: str
+    session_id: str
+    trace_id: str
+    target_agent_type: str
+    delivery_stream: str
+    command_payload: dict[str, Any]
+    user_code: str = ""
+    region: str = ""
+    priority: int = 0
+    metadata: dict[str, Any] = field(default_factory=dict)
+
+    def to_redis_payload(self) -> dict[str, str]:
+        return {"data": json.dumps(asdict(self))}
+
+    @classmethod
+    def from_dict(cls, data: dict[str, Any]) -> "PendingDelivery":
+        return cls(
+            execution_id=str(data.get("execution_id", "")),
+            message_id=str(data.get("message_id", "")),
+            session_id=str(data.get("session_id", "")),
+            trace_id=str(data.get("trace_id", "")),
+            target_agent_type=str(data.get("target_agent_type", "")),
+            delivery_stream=str(data.get("delivery_stream", "")),
+            command_payload=dict(data.get("command_payload", {})),
+            user_code=str(data.get("user_code") or data.get("tenant_id", "")),
+            region=str(data.get("region", "")),
+            priority=int(data.get("priority", 0)),
+            metadata=dict(data.get("metadata", {})),
+        )
+
+
+@dataclass(frozen=True)
+class WakeupDecision:
+    """Decision returned by the manager-owned wakeup controller."""
+
+    execution_id: str = ""
+    target_agent_type: str = ""
+    status: str = WakeupDecisionStatus.FAILED
+    selected_agent_type: str = ""
+    worker_ids: list[str] = field(default_factory=list)
+    region: str = ""
+    retry_after_ms: Optional[int] = None
+    reason: str = ""
+
+    @classmethod
+    def from_dict(cls, data: dict[str, Any]) -> "WakeupDecision":
+        return cls(
+            execution_id=str(data.get("execution_id", "")),
+            target_agent_type=str(data.get("target_agent_type", "")),
+            status=str(data.get("status", WakeupDecisionStatus.FAILED)),
+            selected_agent_type=str(data.get("selected_agent_type", "")),
+            worker_ids=list(data.get("worker_ids", [])),
+            region=str(data.get("region", "")),
+            retry_after_ms=data.get("retry_after_ms"),
+            reason=str(data.get("reason", "")),
+        )
+
+
+@dataclass(frozen=True)
+class AvailabilityResult:
+    """Result returned by AvailabilityRouter before control-message delivery."""
+
+    status: str
+    stream_name: str = ""
+    target_worker_id: str = ""
+    error: str = ""
+    error_code: str = ""
+    selected_agent_type: str = ""
+    execution_id: str = ""
+
+
+class AvailabilityRouter:
+    """Shared availability and wakeup router for client and agent calls."""
+
+    def __init__(
+        self,
+        redis: Redis,
+        registry: Optional[WorkerRegistry] = None,
+    ):
+        self.redis = redis
+        self.registry = registry
+
+    async def prepare_delivery(
+        self,
+        intent: DeliveryIntent,
+    ) -> AvailabilityResult:
+        """Resolve whether a command can be delivered now or needs wakeup."""
+        policy_rejection = await self._check_control_plane_policy(intent)
+        if policy_rejection is not None:
+            return policy_rejection
+
+        if intent.policy == RoutePolicy.SEND_ANYWAY:
+            return AvailabilityResult(
+                status=AvailabilityStatus.DELIVER_NOW,
+                stream_name=RedisKeys.ctrl_stream(intent.target_agent_type),
+                execution_id=intent.execution_id,
+            )
+
+        has_online_worker = await self._has_online_agent_type(intent.target_agent_type)
+        if has_online_worker:
+            return AvailabilityResult(
+                status=AvailabilityStatus.DELIVER_NOW,
+                stream_name=RedisKeys.ctrl_stream(intent.target_agent_type),
+                execution_id=intent.execution_id,
+            )
+
+        fallback = await self._resolve_configured_fallback(intent)
+        if fallback is not None:
+            return fallback
+
+        if intent.policy == RoutePolicy.FAIL_FAST:
+            return self._unavailable(intent)
+
+        if intent.policy == RoutePolicy.WAKE_AND_WAIT:
+            return await self._wake_and_wait(intent)
+
+        if intent.policy == RoutePolicy.WAKE_AND_QUEUE:
+            return await self._wake_and_queue(intent)
+
+        if intent.policy == RoutePolicy.QUEUE_ONLY:
+            return await self._queue_pending(intent)
+
+        return AvailabilityResult(
+            status=AvailabilityStatus.REJECT,
+            error=f"Unsupported offline route policy '{intent.policy}'",
+            error_code="AGENT_TYPE_UNAVAILABLE",
+            execution_id=intent.execution_id,
+        )
+
+    async def _check_control_plane_policy(
+        self, intent: DeliveryIntent
+    ) -> Optional[AvailabilityResult]:
+        circuit = await self._read_json_key(
+            RedisKeys.control_plane_agent_circuit(intent.target_agent_type)
+        )
+        if circuit and str(circuit.get("state", "")).upper() == "OPEN":
+            return AvailabilityResult(
+                status=AvailabilityStatus.REJECT,
+                error=str(circuit.get("reason") or "agent circuit is open"),
+                error_code="AGENT_CIRCUIT_OPEN",
+                execution_id=intent.execution_id,
+            )
+
+        if intent.user_code:
+            quota = await self._read_json_key(
+                RedisKeys.control_plane_user_quota(intent.user_code)
+            )
+            if quota and quota.get("available") is False:
+                return AvailabilityResult(
+                    status=AvailabilityStatus.REJECT,
+                    error=str(quota.get("reason") or "tenant quota exceeded"),
+                    error_code="TENANT_QUOTA_EXCEEDED",
+                    execution_id=intent.execution_id,
+                )
+
+        return None
+
+    async def _resolve_configured_fallback(
+        self, intent: DeliveryIntent
+    ) -> Optional[AvailabilityResult]:
+        fallback = await self._read_json_key(
+            RedisKeys.control_plane_agent_fallback(intent.target_agent_type)
+        )
+        if not fallback:
+            return None
+
+        selected = str(
+            fallback.get("selected_agent_type")
+            or fallback.get("agent_type")
+            or fallback.get("target_agent_type")
+            or ""
+        )
+        if not selected:
+            return None
+
+        if await self._has_online_agent_type(selected):
+            return AvailabilityResult(
+                status=AvailabilityStatus.FALLBACK_TO_OTHER_AGENT_TYPE,
+                stream_name=RedisKeys.ctrl_stream(selected),
+                selected_agent_type=selected,
+                execution_id=intent.execution_id,
+            )
+
+        return None
+
+    async def _wake_and_queue(self, intent: DeliveryIntent) -> AvailabilityResult:
+        request = WakeupRequest(
+            execution_id=intent.execution_id,
+            target_agent_type=intent.target_agent_type,
+            session_id=intent.session_id,
+            trace_id=intent.trace_id,
+            message_id=intent.message_id,
+            source=intent.source,
+            policy=intent.policy,
+            timeout_ms=intent.timeout_ms,
+            user_code=intent.user_code,
+            region=intent.region,
+            priority=intent.priority,
+            metadata=intent.metadata,
+            command_payload=intent.command_payload,
+        )
+        await self.redis.xadd(
+            RedisKeys.control_plane_wakeup_stream(), request.to_redis_payload()
+        )
+        return await self._queue_pending(intent)
+
+    async def _queue_pending(self, intent: DeliveryIntent) -> AvailabilityResult:
+        pending = PendingDelivery(
+            execution_id=intent.execution_id,
+            message_id=intent.message_id,
+            session_id=intent.session_id,
+            trace_id=intent.trace_id,
+            target_agent_type=intent.target_agent_type,
+            delivery_stream=RedisKeys.ctrl_stream(intent.target_agent_type),
+            command_payload=intent.command_payload,
+            user_code=intent.user_code,
+            region=intent.region,
+            priority=intent.priority,
+            metadata=intent.metadata,
+        )
+        await self.redis.xadd(
+            RedisKeys.control_plane_delivery_pending_stream(),
+            pending.to_redis_payload(),
+        )
+        return AvailabilityResult(
+            status=AvailabilityStatus.QUEUE_PENDING,
+            stream_name=RedisKeys.control_plane_delivery_pending_stream(),
+            execution_id=intent.execution_id,
+        )
+
+    async def _wake_and_wait(self, intent: DeliveryIntent) -> AvailabilityResult:
+        request = WakeupRequest(
+            execution_id=intent.execution_id,
+            target_agent_type=intent.target_agent_type,
+            session_id=intent.session_id,
+            trace_id=intent.trace_id,
+            message_id=intent.message_id,
+            source=intent.source,
+            policy=intent.policy,
+            timeout_ms=intent.timeout_ms,
+            user_code=intent.user_code,
+            region=intent.region,
+            priority=intent.priority,
+            metadata=intent.metadata,
+            command_payload=intent.command_payload,
+        )
+        await self.redis.xadd(
+            RedisKeys.control_plane_wakeup_stream(), request.to_redis_payload()
+        )
+
+        decision = await self._wait_for_wakeup_decision(intent)
+        if decision is None:
+            return AvailabilityResult(
+                status=AvailabilityStatus.REJECT,
+                error=(
+                    f"Timed out waiting for worker wakeup for agent_type "
+                    f"'{intent.target_agent_type}'"
+                ),
+                error_code="AGENT_TYPE_UNAVAILABLE",
+                execution_id=intent.execution_id,
+            )
+
+        if decision.status == WakeupDecisionStatus.READY:
+            has_online_worker = await self._has_online_agent_type(
+                intent.target_agent_type
+            )
+            if has_online_worker:
+                return AvailabilityResult(
+                    status=AvailabilityStatus.WAIT_AND_DELIVER,
+                    stream_name=RedisKeys.ctrl_stream(intent.target_agent_type),
+                    execution_id=intent.execution_id,
+                )
+            return self._unavailable(intent)
+
+        if decision.status == WakeupDecisionStatus.FALLBACK:
+            selected = decision.selected_agent_type
+            return AvailabilityResult(
+                status=AvailabilityStatus.FALLBACK_TO_OTHER_AGENT_TYPE,
+                stream_name=RedisKeys.ctrl_stream(selected) if selected else "",
+                selected_agent_type=selected,
+                execution_id=intent.execution_id,
+            )
+
+        return AvailabilityResult(
+            status=AvailabilityStatus.REJECT,
+            error=decision.reason
+            or f"Wakeup rejected for agent_type '{intent.target_agent_type}'",
+            error_code="AGENT_TYPE_UNAVAILABLE",
+            execution_id=intent.execution_id,
+        )
+
+    async def _wait_for_wakeup_decision(
+        self, intent: DeliveryIntent
+    ) -> Optional[WakeupDecision]:
+        deadline = time.monotonic() + max(intent.timeout_ms, 0) / 1000
+        result_stream = RedisKeys.control_plane_wakeup_result_stream(
+            intent.execution_id
+        )
+        last_id = "0-0"
+
+        while True:
+            remaining_ms = int((deadline - time.monotonic()) * 1000)
+            if remaining_ms <= 0:
+                return None
+
+            messages = await self.redis.xread(
+                streams={result_stream: last_id},
+                count=1,
+                block=remaining_ms,
+            )
+            if not messages:
+                return None
+
+            for _, entries in messages:
+                for entry_id, fields in entries:
+                    last_id = (
+                        entry_id.decode("utf-8")
+                        if isinstance(entry_id, bytes)
+                        else str(entry_id)
+                    )
+                    decision = self._decode_decision(fields)
+                    if decision is None:
+                        continue
+                    if (
+                        decision.execution_id
+                        and decision.execution_id != intent.execution_id
+                    ):
+                        continue
+                    if decision.status in (
+                        WakeupDecisionStatus.STARTING,
+                        WakeupDecisionStatus.QUEUED,
+                    ):
+                        break
+                    return decision
+
+    async def _has_online_agent_type(self, agent_type: str) -> bool:
+        if self.registry is not None:
+            has_online_agent_type, _ = await self.registry.has_online_agent_type(
+                agent_type
+            )
+            return bool(has_online_agent_type)
+
+        has_online_agent_type, _ = await check_agent_type_online(
+            self.redis, agent_type, check_active=True
+        )
+        return bool(has_online_agent_type)
+
+    @staticmethod
+    def _decode_decision(fields: dict[Any, Any]) -> Optional[WakeupDecision]:
+        raw = fields.get(b"data") if isinstance(fields, dict) else None
+        if raw is None and isinstance(fields, dict):
+            raw = fields.get("data")
+        if raw is None:
+            return None
+        if isinstance(raw, bytes):
+            raw = raw.decode("utf-8")
+        return WakeupDecision.from_dict(json.loads(raw))
+
+    async def _read_json_key(self, key: str) -> Optional[dict[str, Any]]:
+        if not hasattr(self.redis, "get"):
+            return None
+        raw = await self.redis.get(key)
+        if raw is None:
+            return None
+        if isinstance(raw, bytes):
+            raw = raw.decode("utf-8")
+        if not isinstance(raw, str):
+            return None
+        try:
+            decoded = json.loads(raw)
+        except json.JSONDecodeError:
+            return None
+        return decoded if isinstance(decoded, dict) else None
+
+    @staticmethod
+    def _unavailable(intent: DeliveryIntent) -> AvailabilityResult:
+        return AvailabilityResult(
+            status=AvailabilityStatus.REJECT,
+            error=f"No online worker found for agent_type '{intent.target_agent_type}'",
+            error_code="AGENT_TYPE_UNAVAILABLE",
+            execution_id=intent.execution_id,
+        )