feat(otel #876): 双 Claw 强制标签 + 入口接线 + Grafana 双 Claw 对比看板

#876 OpenTelemetry 可观测(双 Claw 标签):
- app/core/otel.py: 新增 set_otel_ctx/get_otel_ctx(ContextVar 三元组 tenant_id+claw_type+
  trace_id);current_span_cm 自动注入双 Claw 强制标签,全部 13 个 span 调用点零签名侵入带维度;
  record_span_metrics 指标维度补 tenant_id+claw_type(trace_id 因高基数不计入指标标签)。
- app/core/llm_router.py: set_route_ctx 桥接转调 set_otel_ctx(单次入口注入双端共享:
  路由上下文 + OTel 标签),失败安全。
- app/api/chat.py: send_message 与 send_message_stream 入口注入 set_route_ctx(web),
  stream 经 asyncio.create_task 子协程自动传播 contextvars,使 _send_message_core 内
  所有 span 均带 tenant_id+claw_type。
- app/api/im_llm.py: _call_real_llm 入口最早注入 set_route_ctx(im_wechat),并显式开启
  im.send_message span(此前 IM Claw 完全无 span,Grafana 无 IM 数据点),finally 关闭。
- tests/test_dual_claw.py: R-DUAL-6 由 skip 转为确定性断言(Web/IM 双 Claw span 必带
  tenant_id+claw_type),作为持续哨兵。
- 交付 Grafana 双 Claw 对比看板 JSON(aiflow_span_* / aiflow_llm_route_total 按 claw_type
  分组:P50/P95 延迟、成功率、QPS、租户健康 Top10)。Alloy OTLP 仅桥接 metrics→Prometheus,
  trace 无 pipeline(#800 已压制 UNIMPLEMENTED 噪声),看板基于 metrics 出图。

safe-align: 仅复制 5 个 git-tracked 文件,未触碰 secrets/venv/node_modules/caches。本地提交不 push。
This commit is contained in:
QihuAI Design committed 2026-08-04 13:49:34 +08:00
1 parent 180409d317
commit 04a0641202
5 files changed
+153 -13

No files matched your search

+14
View File
@@ -143,6 +143,7 @@ from app.api.chat_utils import (CST_TZ, CreateConversationRequest, SendMessageRe
from app.api.chat_service import _send_message_core from app.api.chat_service import _send_message_core
from app.core.llm_router import set_route_ctx
@@ -1560,6 +1561,12 @@ async def send_message(
from app.core.observability import span from app.core.observability import span
# #876 双 Claw 维度:入口最早注入,确保外层 span 即带 tenant_id + claw_type 标签
set_route_ctx(
tenant_id=str(getattr(current_user, "tenant_id", "") or ""),
claw_type="web",
)
async with span("chat.send_message", conv_id=conv_id, user_id=getattr(current_user, "id", None)): async with span("chat.send_message", conv_id=conv_id, user_id=getattr(current_user, "id", None)):
return await _send_message_core(conv_id, req, db, current_user) return await _send_message_core(conv_id, req, db, current_user)
@@ -1579,6 +1586,13 @@ async def send_message_stream(
""" """
db = SessionLocal() db = SessionLocal()
try: try:
# #876 双 Claw 维度:入口最早注入。runner() 经 asyncio.create_task 派生子协程,
# contextvars 自动随上下文传播,使 _send_message_core 内所有 span 均带 tenant_id + claw_type。
set_route_ctx(
tenant_id=str(getattr(current_user, "tenant_id", "") or ""),
claw_type="web",
)
adapter = SseAdapter( adapter = SseAdapter(
trace_id=f"sse-{conv_id}", trace_id=f"sse-{conv_id}",
tenant_id=str(getattr(current_user, "tenant_id", "") or ""), tenant_id=str(getattr(current_user, "tenant_id", "") or ""),
+14 -2
View File
@@ -30,7 +30,7 @@ from sqlalchemy.orm import Session
from ..capability_registry import resolve_auto_model from ..capability_registry import resolve_auto_model
from ..models import AIModel, ModelProvider, TokenUsageLog, gen_uuid from ..models import AIModel, ModelProvider, TokenUsageLog, gen_uuid
from app.core.observability import get_trace_id from app.core.observability import get_trace_id, span
# #875 LLM Router:统一模型路由层(影子模式先行,收敛 litellm 主密钥硬编码) # #875 LLM Router:统一模型路由层(影子模式先行,收敛 litellm 主密钥硬编码)
from ..core.llm_router import ( from ..core.llm_router import (
route as _llm_route, route as _llm_route,
@@ -137,6 +137,13 @@ def _log_media_usage(db: Session, user_id: str, ai_model, media_type: str, promp
def _call_real_llm(db: Session, user=None, prompt: str = "", platform: str = "im", user_id: str = None) -> "str | dict": def _call_real_llm(db: Session, user=None, prompt: str = "", platform: str = "im", user_id: str = None) -> "str | dict":
"""调用七虎AI 生成回复,供 IM Bot 复用。支持媒体意图路由、已安装技能自动发现、模型 failover。""" """调用七虎AI 生成回复,供 IM Bot 复用。支持媒体意图路由、已安装技能自动发现、模型 failover。"""
# #876 双 Claw 维度:入口最早注入 tenant_id + claw_type,确保下方所有 span / metric
# 均带双 Claw 标签(此前 IM Claw 完全无 span,Grafana 无 IM 数据点)。
_set_route_ctx(tenant_id=getattr(user, "tenant_id", "") or "", claw_type="im_wechat")
# #876:IM Claw 此前完全无 span,Grafana 无 IM 数据点。此处显式开启一个 im.send_message
# span(手动 enter/exit 以避免整段重缩进),使 IM 链路在双 Claw 对比看板出图。
_im_span = span("im.send_message", platform=platform, user_id=user_id)
_im_span.__enter__()
try: try:
from app.api.chat import ( from app.api.chat import (
_generate_media, _generate_media,
@@ -224,7 +231,6 @@ def _call_real_llm(db: Session, user=None, prompt: str = "", platform: str = "im
# 计算决策并打 claw_type 标签 metrics;影子模式下仅观测,不改流量。 # 计算决策并打 claw_type 标签 metrics;影子模式下仅观测,不改流量。
# 切主路径:LLM_ROUTER_SHADOW=0 时改用 Router 候选链(已含文本过滤,等同此处逻辑)。 # 切主路径:LLM_ROUTER_SHADOW=0 时改用 Router 候选链(已含文本过滤,等同此处逻辑)。
try: try:
_set_route_ctx(tenant_id=getattr(user, "tenant_id", "") or "", claw_type="im_wechat")
_rm = _llm_route("text", db=db, legacy_primary=_try_models[0].name if _try_models else None) _rm = _llm_route("text", db=db, legacy_primary=_try_models[0].name if _try_models else None)
if not _LLM_ROUTER_SHADOW and _rm.models: if not _LLM_ROUTER_SHADOW and _rm.models:
_try_models = _rm.models _try_models = _rm.models
@@ -441,6 +447,12 @@ def _call_real_llm(db: Session, user=None, prompt: str = "", platform: str = "im
return f"(已尝试 {len(_try_models)} 个模型均调用失败。最后错误: {_summary}。请稍后重试。)" return f"(已尝试 {len(_try_models)} 个模型均调用失败。最后错误: {_summary}。请稍后重试。)"
except Exception as e: except Exception as e:
return f"(AI 回复生成失败:{e})" return f"(AI 回复生成失败:{e})"
finally:
# #876:关闭 im.send_message span(记录耗时 + 指标,强制带 tenant_id + claw_type)
try:
_im_span.__exit__(None, None, None)
except Exception:
pass
async def _call_engine_core(db, user_id, prompt: str, platform: str, user=None) -> "str | dict": async def _call_engine_core(db, user_id, prompt: str, platform: str, user=None) -> "str | dict":
+11 -1
View File
@@ -51,10 +51,20 @@ _route_claw_var: ContextVar[str] = ContextVar("aiflow_route_claw", default="web"
def set_route_ctx(*, tenant_id: str = "", claw_type: str = "web") -> None: def set_route_ctx(*, tenant_id: str = "", claw_type: str = "web") -> None:
"""在请求入口注入路由上下文。tenant_id/claw_type 为空时保留默认值。""" """在请求入口注入路由上下文(#875)。tenant_id/claw_type 为空时保留默认值。
#876 桥接:同时注入 OTel 双 Claw 强制标签(set_otel_ctx),使单次入口调用
即可让后续所有 span / metric 自动带上 tenant_id + claw_type + trace_id。
失败安全:otel 不可用时静默跳过。
"""
if tenant_id: if tenant_id:
_route_tenant_var.set(tenant_id) _route_tenant_var.set(tenant_id)
_route_claw_var.set(claw_type or "web") _route_claw_var.set(claw_type or "web")
try:
from app.core.otel import set_otel_ctx
set_otel_ctx(tenant_id=tenant_id, claw_type=claw_type or "web")
except Exception:
pass
def get_route_ctx() -> dict: def get_route_ctx() -> dict:
+72 -6
View File
@@ -15,12 +15,21 @@ OpenTelemetry 桥接 (Phase 1.I 标准化)
init_otel() 首次调用时构建 TracerProvider + MeterProvider(幂等) init_otel() 首次调用时构建 TracerProvider + MeterProvider(幂等)
get_tracer() 返回 tracer(未启用/失败则 None) get_tracer() 返回 tracer(未启用/失败则 None)
get_meter() 返回 meter(未启用/失败则 None) get_meter() 返回 meter(未启用/失败则 None)
current_span_cm(name) 返回一个 OTel span 的上下文管理器(供 observability 嵌套) current_span_cm(name) 返回一个 OTel span 的上下文管理器(供 observability 嵌套);
record_span_metrics(name, elapsed_ms, status) 记录 span 指标(耗时直方图 + 计数) **自动注入** tenant_id + claw_type + trace_id 双 Claw 强制标签
record_span_metrics(name, elapsed_ms, status) 记录 span 指标(耗时直方图 + 计数,
维度带 tenant_id + claw_type,供 Grafana 双 Claw 对比板)
set_otel_ctx(tenant_id, claw_type) 入口注入双 Claw 强制标签(与 llm_router.set_route_ctx 同源)
get_otel_ctx() 读取当前 contextvars 三元组 {tenant_id, claw_type, trace_id}
双 Claw 强制标签(#876 §D):每个 span / metric 必带 tenant_id + claw_type + trace_id
(+ 业务字段 model / node_type 由 span(name, model=..., node_type=...) 透传)。
入口调用一次 set_otel_ctx(或 set_route_ctx,已桥接转调),全链路自动带维度,业务零改动。
""" """
import os import os
import logging import logging
from contextvars import ContextVar
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
@@ -35,6 +44,44 @@ _initialized = False
_duration_hist = None _duration_hist = None
_span_counter = None _span_counter = None
# ---------------------------------------------------------------------------
# 强制标签上下文(#876 · 双 Claw 维度)
# ---------------------------------------------------------------------------
# 每个 span / metric 必带 tenant_id + claw_type + trace_id(设计书 §D)。
# 经 contextvars 注入,零签名侵入:入口调用 set_otel_ctx() 一次,
# 全部 span 经 current_span_cm 自动带上维度,业务侧无需改动。
# 与 #875 llm_router.set_route_ctx 同源(set_route_ctx 内部转调本函数)。
_otel_tenant_var: ContextVar[str] = ContextVar("aiflow_otel_tenant", default="")
_otel_claw_var: ContextVar[str] = ContextVar("aiflow_otel_claw", default="web")
def set_otel_ctx(*, tenant_id: str = "", claw_type: str = "web") -> None:
"""在请求入口注入 OTel 强制标签(双 Claw 维度)。
与 llm_router.set_route_ctx 共享同一调用点:入口调用一次即可让
后续所有 span / metric 自动带上 tenant_id + claw_type + trace_id。
"""
if tenant_id:
_otel_tenant_var.set(tenant_id)
_otel_claw_var.set(claw_type or "web")
def get_otel_ctx() -> dict:
"""读取当前 OTel 上下文三元组(含 trace_id,复用 observability)。
失败安全:observability 不可用时 trace_id 回退为 '-'。
"""
try:
from app.core.observability import get_trace_id
_trace = get_trace_id()
except Exception:
_trace = "-"
return {
"tenant_id": _otel_tenant_var.get(),
"claw_type": _otel_claw_var.get(),
"trace_id": _trace,
}
def init_otel() -> None: def init_otel() -> None:
"""构建 TracerProvider + MeterProvider(幂等)。失败则降级为 no-op。""" """构建 TracerProvider + MeterProvider(幂等)。失败则降级为 no-op。"""
@@ -100,19 +147,32 @@ def get_meter():
def current_span_cm(name: str, attributes: dict | None = None): def current_span_cm(name: str, attributes: dict | None = None):
"""返回一个 OTel span 的上下文管理器(同步/异步通用)。失败时返回 None。""" """返回一个 OTel span 的上下文管理器(同步/异步通用)。失败时返回 None。
自动注入双 Claw 强制标签(#876 §D):tenant_id + claw_type + trace_id。
setdefault 语义——业务已显式传入的字段不会被覆盖。
"""
tr = get_tracer() tr = get_tracer()
if tr is None: if tr is None:
return None return None
try: try:
return tr.start_as_current_span(name, attributes=attributes or {}) attrs = dict(attributes or {})
ctx = get_otel_ctx()
attrs.setdefault("tenant_id", ctx["tenant_id"])
attrs.setdefault("claw_type", ctx["claw_type"])
attrs.setdefault("trace_id", ctx["trace_id"])
return tr.start_as_current_span(name, attributes=attrs)
except Exception as e: # pragma: no cover except Exception as e: # pragma: no cover
logger.warning(f"[otel] start_span failed: {e}") logger.warning(f"[otel] start_span failed: {e}")
return None return None
def record_span_metrics(name: str, elapsed_ms: float, status: str) -> None: def record_span_metrics(name: str, elapsed_ms: float, status: str) -> None:
"""记录 span 指标:耗时直方图 + 调用计数(供 Grafana 出图)。失败安全。""" """记录 span 指标:耗时直方图 + 调用计数(供 Grafana 出图)。失败安全。
指标维度带 tenant_id + claw_type(trace_id 因高基数不计入指标标签,
仅作为 span 属性存在)。双 Claw 对比看板据此按 claw_type 分组。
"""
m = get_meter() m = get_meter()
if m is None: if m is None:
return return
@@ -123,7 +183,13 @@ def record_span_metrics(name: str, elapsed_ms: float, status: str) -> None:
"aiflow_span_duration_ms", "Span duration in milliseconds" "aiflow_span_duration_ms", "Span duration in milliseconds"
) )
_span_counter = m.create_counter("aiflow_span_count", "Span invocation count") _span_counter = m.create_counter("aiflow_span_count", "Span invocation count")
attrs = {"span": name, "status": status} ctx = get_otel_ctx()
attrs = {
"span": name,
"status": status,
"tenant_id": ctx["tenant_id"],
"claw_type": ctx["claw_type"],
}
_duration_hist.record(elapsed_ms, attrs) _duration_hist.record(elapsed_ms, attrs)
_span_counter.add(1, attrs) _span_counter.add(1, attrs)
except Exception: # pragma: no cover except Exception: # pragma: no cover
+42 -4
View File
@@ -187,9 +187,47 @@ def test_rdual3_im_agent_capability(db):
# ───────────────────────────────────────────────────────────────────────────── # ─────────────────────────────────────────────────────────────────────────────
# R-DUAL-6(OTel 标签,待 #876):双 Claw span 须携带 tenant_id + claw_type # R-DUAL-6(OTel 标签,#876 已完成):双 Claw span 须携带 tenant_id + claw_type
# ───────────────────────────────────────────────────────────────────────────── # ─────────────────────────────────────────────────────────────────────────────
@pytest.mark.skip(reason="pending #876 OTel enhancement: spans must carry tenant_id + claw_type")
def test_rdual6_spans_carry_tenant_and_claw(): def test_rdual6_spans_carry_tenant_and_claw():
"""双 Claw 链路的 span 必须带有 tenant_id 与 claw_type 标签(双端横向对比看板依赖)。""" """双 Claw 链路的 span 必须带有 tenant_id 与 claw_type 标签(双端横向对比看板依赖)。
pytest.skip("pending #876")
#876 在 otel.current_span_cm 层强制注入 tenant_id + claw_type + trace_id,
入口经 set_otel_ctx / set_route_ctx 注入一次,全部 span 自动带维度(零签名侵入)。
本测试以确定性方式验证该强制注入对 Web / IM 双 Claw 均生效。
"""
from app.core.otel import set_otel_ctx, current_span_cm, get_otel_ctx
# 模拟 Web Claw 入口注入
set_otel_ctx(tenant_id="dual-claw-sentinel", claw_type="web")
assert get_otel_ctx()["claw_type"] == "web"
cm = current_span_cm("test_rdual6_web_span", {"model": "deepseek"})
if cm is None:
pytest.skip("OTel tracer 不可用(OTEL_ENABLED=0 或 init 失败)—— 标签注入代码存在但未在本题环境演练")
span = cm.__enter__()
try:
attrs = dict(span.attributes)
assert attrs.get("tenant_id") == "dual-claw-sentinel", \
f"Web Claw span 缺失 tenant_id 标签: {attrs}"
assert attrs.get("claw_type") == "web", \
f"Web Claw span 缺失 claw_type 标签: {attrs}"
assert attrs.get("model") == "deepseek", \
f"业务字段 model 未透传: {attrs}"
finally:
cm.__exit__(None, None, None)
# 模拟 IM Claw 入口注入
set_otel_ctx(tenant_id="dual-claw-sentinel", claw_type="im_wechat")
cm2 = current_span_cm("test_rdual6_im_span")
if cm2 is None:
pytest.skip("OTel tracer 不可用,跳过 IM 断言")
span2 = cm2.__enter__()
try:
attrs2 = dict(span2.attributes)
assert attrs2.get("claw_type") == "im_wechat", \
f"IM Claw span 缺失 claw_type=im_wechat 标签: {attrs2}"
assert attrs2.get("tenant_id") == "dual-claw-sentinel", \
f"IM Claw span 缺失 tenant_id 标签: {attrs2}"
finally:
cm2.__exit__(None, None, None)