cristiano.hoshikawa/agent_platform_oci
1
2
Fork 0
mirror of https://github.com/hoshikawa2/agent_platform_oci.git synced 2026-07-09 14:04:19 +00:00
Code Issues Packages Projects Releases Wiki Activity

bugfix Alex 2026-06-24

Browse Source
This commit is contained in:
Cristiano Hoshikawa
2026-06-24 14:15:44 -03:00
parent 853c286313
commit dc618bfe47
11 changed files with 337 additions and 100 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
.idea/workspace.xml generated
View File

@@ -6,9 +6,11 @@
<component name="ChangeListManager">
<list default="true" id="30a0e1d8-9d7d-469b-b241-f300911cee8a" name="Changes" comment="Ajustes na documentação e remanejamento dos folders">
<change beforePath="$PROJECT_DIR$/.idea/workspace.xml" beforeDir="false" afterPath="$PROJECT_DIR$/.idea/workspace.xml" afterDir="false" />
<change beforePath="$PROJECT_DIR$/README.md" beforeDir="false" afterPath="$PROJECT_DIR$/README.md" afterDir="false" />
<change beforePath="$PROJECT_DIR$/README_en.md" beforeDir="false" afterPath="$PROJECT_DIR$/README_en.md" afterDir="false" />
<change beforePath="$PROJECT_DIR$/specs/SPEC-003-AI-Gateway.md" beforeDir="false" />
<change beforePath="$PROJECT_DIR$/libs/agent_framework/src/agent_framework/config/settings.py" beforeDir="false" afterPath="$PROJECT_DIR$/libs/agent_framework/src/agent_framework/config/settings.py" afterDir="false" />
<change beforePath="$PROJECT_DIR$/libs/agent_framework/src/agent_framework/observability/context.py" beforeDir="false" afterPath="$PROJECT_DIR$/libs/agent_framework/src/agent_framework/observability/context.py" afterDir="false" />
<change beforePath="$PROJECT_DIR$/libs/agent_framework/src/agent_framework/observability/langgraph_telemetry.py" beforeDir="false" afterPath="$PROJECT_DIR$/libs/agent_framework/src/agent_framework/observability/langgraph_telemetry.py" afterDir="false" />
<change beforePath="$PROJECT_DIR$/libs/agent_framework/src/agent_framework/observability/telemetry.py" beforeDir="false" afterPath="$PROJECT_DIR$/libs/agent_framework/src/agent_framework/observability/telemetry.py" afterDir="false" />
<change beforePath="$PROJECT_DIR$/templates/agent_template_backend/.env" beforeDir="false" afterPath="$PROJECT_DIR$/templates/agent_template_backend/.env" afterDir="false" />
</list>
<option name="SHOW_DIALOG" value="false" />
<option name="HIGHLIGHT_CONFLICTS" value="true" />
@@ -86,7 +88,7 @@
<workItem from="1782038604623" duration="6486000" />
<workItem from="1782047166074" duration="27000" />
<workItem from="1782047194363" duration="961000" />
<workItem from="1782048494672" duration="230000" />
<workItem from="1782048494672" duration="2474000" />
</task>
<task id="LOCAL-00001" summary="Ajustes na documentação e remanejamento dos folders">
<option name="closed" value="true" />
@@ -112,7 +114,15 @@
<option name="project" value="LOCAL" />
<updated>1782048304579</updated>
</task>
<option name="localTasksCounter" value="4" />
<task id="LOCAL-00004" summary="Ajustes na documentação e remanejamento dos folders">
<option name="closed" value="true" />
<created>1782048747421</created>
<option name="number" value="00004" />
<option name="presentableId" value="LOCAL-00004" />
<option name="project" value="LOCAL" />
<updated>1782048747421</updated>
</task>
<option name="localTasksCounter" value="5" />
<servers />
</component>
<component name="TypeScriptGeneratedFilesManager">

1
libs/agent_framework/src/agent_framework/config/settings.py
View File

@@ -99,6 +99,7 @@ class Settings(BaseSettings):
OCI_EMBEDDING_MODEL: str = 'cohere.embed-multilingual-v3.0'
ENABLE_LANGFUSE: bool = False
LANGFUSE_TRACE_MODE: Literal['verbose','compact'] = 'verbose'
LANGFUSE_PUBLIC_KEY: str | None = None
LANGFUSE_SECRET_KEY: str | None = None
LANGFUSE_HOST: str = 'https://cloud.langfuse.com'

28
libs/agent_framework/src/agent_framework/observability/context.py
View File

@@ -21,6 +21,7 @@ _workflow_id: ContextVar[str | None] = ContextVar("workflow_id", default=None)
_message_id: ContextVar[str | None] = ContextVar("message_id", default=None)
_trace_id: ContextVar[str | None] = ContextVar("trace_id", default=None)
_current_observation_id: ContextVar[str | None] = ContextVar("current_observation_id", default=None)
_current_span_events: ContextVar[list[dict[str, Any]] | None] = ContextVar("current_span_events", default=None)
@dataclass(slots=True)
class ObservabilityContext:
@@ -66,6 +67,31 @@ def reset_current_observation_id(token) -> None:
_current_observation_id.set(None)
def get_current_span_events() -> list[dict[str, Any]] | None:
"""Return the mutable aggregate-event bucket for the active macro span."""
return _current_span_events.get()
def set_current_span_events(events: list[dict[str, Any]] | None):
"""Set aggregate-event bucket and return ContextVar token."""
return _current_span_events.set(events)
def reset_current_span_events(token) -> None:
"""Restore previous aggregate-event bucket."""
try:
_current_span_events.reset(token)
except Exception:
_current_span_events.set(None)
def record_current_span_event(event: dict[str, Any]) -> None:
"""Append an event summary to the active macro span, if one exists."""
events = _current_span_events.get()
if events is not None:
events.append(event)
def set_observability_context(**kwargs: Any) -> ObservabilityContext:
if not kwargs.get("request_id") and not _request_id.get():
kwargs["request_id"] = str(uuid4())
@@ -82,7 +108,7 @@ def set_observability_context(**kwargs: Any) -> ObservabilityContext:
def clear_observability_context() -> None:
for var in (_request_id, _session_id, _user_id, _tenant_id, _agent_id, _channel, _ura_call_id, _workflow_id, _message_id, _trace_id, _current_observation_id):
for var in (_request_id, _session_id, _user_id, _tenant_id, _agent_id, _channel, _ura_call_id, _workflow_id, _message_id, _trace_id, _current_observation_id, _current_span_events):
var.set(None)

41
libs/agent_framework/src/agent_framework/observability/langgraph_telemetry.py
View File

@@ -3,6 +3,36 @@ import time
from contextlib import asynccontextmanager
from typing import Any
_LANGGRAPH_STEP_ORDER = {
"__start__": 0,
"input_guardrails": 1,
"routing_decision": 2,
"billing_agent": 3,
"product_agent": 3,
"orders_agent": 3,
"support_agent": 3,
"handoff": 3,
"supervisor_agent": 3,
"output_supervisor": 4,
"output_guardrails": 5,
"judge": 6,
"supervisor_review": 7,
"persist": 8,
"__end__": 9,
}
def _langgraph_step(name: str, state: dict[str, Any]) -> int:
explicit_steps = state.get("langgraph_steps")
if isinstance(explicit_steps, dict) and name in explicit_steps:
try:
return int(explicit_steps[name])
except (TypeError, ValueError):
pass
return _LANGGRAPH_STEP_ORDER.get(name, 50)
class LangGraphDeepTelemetry:
"""Eventos profundos do LangGraph no padrão FIRST.
@@ -16,7 +46,16 @@ class LangGraphDeepTelemetry:
async def node(self, name: str, state: dict[str, Any] | None = None):
state=state or {}
session_id=state.get('conversation_key') or state.get('session_id')
payload={'node': name, 'session_id': session_id, 'agent_id': state.get('agent_id'), 'tenant_id': state.get('tenant_id'), 'input_size': len(str(state.get('user_text') or state.get('sanitized_input') or ''))}
payload={
'node': name,
'langgraph_node': name,
'langgraph_step': _langgraph_step(name, state),
'framework': 'langgraph',
'session_id': session_id,
'agent_id': state.get('agent_id'),
'tenant_id': state.get('tenant_id'),
'input_size': len(str(state.get('user_text') or state.get('sanitized_input') or '')),
}
start=time.time()
await self.telemetry.event('langgraph.node.started', payload, kind='langgraph')
async with self.telemetry.span(f'langgraph.node.{name}', **payload):

84
libs/agent_framework/src/agent_framework/observability/telemetry.py
View File

@@ -21,9 +21,13 @@ from uuid import uuid4
from .context import (
context_metadata,
get_current_observation_id,
get_current_span_events,
get_observability_context,
record_current_span_event,
reset_current_observation_id,
reset_current_span_events,
set_current_observation_id,
set_current_span_events,
set_observability_context,
)
from .event_bus import TelemetryEventBus
@@ -42,6 +46,18 @@ def _langfuse_type(kind: str | None) -> str:
_LANGFUSE_TRACE_ID_RE = re.compile(r"^[0-9a-f]{32}$")
_COMPACT_SUPPRESSED_SPAN_PREFIXES = (
"llm.chat_completion",
"workflow.agent.",
"workflow.handoff",
"workflow.input_guardrails",
"workflow.judge",
"workflow.output_guardrails",
"workflow.output_supervisor",
"workflow.persist",
"workflow.routing_decision",
"workflow.supervisor_review",
)
def _raw_correlation_id(attrs: dict[str, Any] | None = None) -> str | None:
@@ -100,6 +116,7 @@ def _inject_langfuse_trace_context(kwargs: dict[str, Any], attrs: dict[str, Any]
which grouped everything in one trace while flattening the tree.
"""
attrs = attrs or kwargs.get("metadata") or {}
ignore_current_parent = bool(attrs.get("_ignore_current_parent") or kwargs.get("_ignore_current_parent"))
raw_id = _raw_correlation_id(attrs)
trace_id = _langfuse_trace_id(raw_id)
parent_id = (
@@ -107,7 +124,7 @@ def _inject_langfuse_trace_context(kwargs: dict[str, Any], attrs: dict[str, Any]
or attrs.get("parent_span_id")
or kwargs.get("parent_observation_id")
or kwargs.get("parent_span_id")
or get_current_observation_id()
or (None if ignore_current_parent else get_current_observation_id())
)
if trace_id:
trace_context = dict(kwargs.get("trace_context") or {})
@@ -121,6 +138,8 @@ def _inject_langfuse_trace_context(kwargs: dict[str, Any], attrs: dict[str, Any]
metadata.setdefault("langfuse_trace_id", trace_id)
if parent_id:
metadata.setdefault("parent_observation_id", str(parent_id))
metadata.pop("_ignore_current_parent", None)
kwargs.pop("_ignore_current_parent", None)
return kwargs
@@ -188,6 +207,15 @@ class Telemetry:
def is_enabled(self) -> bool:
return bool(self.enabled and self.langfuse)
def is_compact_mode(self) -> bool:
mode = getattr(self.settings, "LANGFUSE_TRACE_MODE", "verbose") or "verbose"
return str(mode).lower() == "compact"
def _should_emit_langfuse_span(self, name: str) -> bool:
if not self.is_compact_mode():
return True
return not str(name).startswith(_COMPACT_SUPPRESSED_SPAN_PREFIXES)
def bind_context(self, **kwargs: Any):
return set_observability_context(**kwargs)
@@ -199,6 +227,9 @@ class Telemetry:
"""Cria span correlacionado em logs, Langfuse e OpenTelemetry."""
start = time.time()
attrs = context_metadata(attrs)
attrs.setdefault("_span_name", name)
if self.is_compact_mode() and name == "agent.gateway_message" and not attrs.get("parent_observation_id"):
attrs["_ignore_current_parent"] = True
if not attrs.get("request_id"):
attrs["request_id"] = str(uuid4())
if not attrs.get("trace_id"):
@@ -207,19 +238,27 @@ class Telemetry:
observation_cm = None
observation = None
observation_token = None
parent_observation_id = attrs.get("parent_observation_id") or get_current_observation_id()
ignore_current_parent = bool(attrs.get("_ignore_current_parent"))
parent_observation_id = attrs.get("parent_observation_id")
if not parent_observation_id and not ignore_current_parent:
parent_observation_id = get_current_observation_id()
if parent_observation_id:
attrs.setdefault("parent_observation_id", str(parent_observation_id))
logger.info("span.start %s %s", name, _safe(attrs))
otel_cm = self.otel.span(name, attrs)
otel_span = otel_cm.__enter__()
if self.is_enabled():
emit_langfuse_span = self.is_enabled() and self._should_emit_langfuse_span(name)
span_events: list[dict[str, Any]] | None = [] if emit_langfuse_span and self.is_compact_mode() else None
span_events_token = set_current_span_events(span_events) if span_events is not None else None
observation_metadata = {k: v for k, v in attrs.items() if k != "input" and not str(k).startswith("_")}
if emit_langfuse_span:
observation_cm = self._start_observation(
name=name,
as_type="span",
input=attrs.get("input"),
metadata={k: v for k, v in attrs.items() if k != "input"},
metadata=observation_metadata,
_ignore_current_parent=attrs.get("_ignore_current_parent"),
)
try:
if observation_cm is not None:
@@ -228,6 +267,7 @@ class Telemetry:
if observation_id:
observation_token = set_current_observation_id(observation_id)
attrs.setdefault("observation_id", observation_id)
if not attrs.get("parent_observation_id"):
self._update_trace_from_attrs(observation, attrs)
# Publish span.started only after the Langfuse observation is current,
# so secondary analytics/exporters can attach it as a child instead
@@ -236,7 +276,11 @@ class Telemetry:
yield observation
duration_ms = int((time.time() - start) * 1000)
out = {"status": "ok", "duration_ms": duration_ms}
self._update_observation(observation, output=out, metadata={"duration_ms": duration_ms})
metadata = {**observation_metadata, "duration_ms": duration_ms}
if span_events:
metadata["aggregated_event_count"] = len(span_events)
metadata["aggregated_events"] = span_events
self._update_observation(observation, output=out, metadata=metadata)
if otel_span is not None:
otel_span.set_attribute("duration_ms", duration_ms)
await self.event_bus.publish(f"{name}.completed", {**attrs, **out}, kind="span")
@@ -244,7 +288,11 @@ class Telemetry:
except Exception as exc:
duration_ms = int((time.time() - start) * 1000)
out = {"status": "error", "error": str(exc), "duration_ms": duration_ms}
self._update_observation(observation, level="ERROR", status_message=str(exc), output=out, metadata={"duration_ms": duration_ms})
metadata = {**observation_metadata, "duration_ms": duration_ms}
if span_events:
metadata["aggregated_event_count"] = len(span_events)
metadata["aggregated_events"] = span_events
self._update_observation(observation, level="ERROR", status_message=str(exc), output=out, metadata=metadata)
if otel_span is not None:
try:
otel_span.record_exception(exc)
@@ -260,6 +308,8 @@ class Telemetry:
except Exception: logger.exception("Falha ao finalizar span Langfuse %s", name)
if observation_token is not None:
reset_current_observation_id(observation_token)
if span_events_token is not None:
reset_current_span_events(span_events_token)
try: otel_cm.__exit__(None, None, None)
except Exception: logger.debug("Falha ao fechar span OTEL", exc_info=True)
@@ -267,6 +317,14 @@ class Telemetry:
payload = context_metadata(payload or {})
logger.info("event %s %s", name, _safe(payload))
await self.event_bus.publish(name, payload, kind=kind)
if self.is_compact_mode():
if get_current_span_events() is not None:
record_current_span_event({
"name": name,
"kind": kind,
"payload": payload,
})
return
if not self.is_enabled():
return
# IMPORTANT: do not call ``langfuse.event(...)`` directly here. In SDK
@@ -274,7 +332,10 @@ class Telemetry:
# a new trace row for every telemetry event. We create a correlated
# observation instead, using request_id/trace_id as the stable trace id.
try:
cm = self._start_observation(name=name, as_type=_langfuse_type(kind), metadata={**payload, "event_kind": kind})
metadata = {**payload, "event_kind": kind}
if self.is_compact_mode():
metadata["_ignore_current_parent"] = True
cm = self._start_observation(name=name, as_type=_langfuse_type(kind), metadata=metadata)
if cm is not None:
with cm: pass
except Exception:
@@ -305,6 +366,7 @@ class Telemetry:
# trace row per LLM call when no current observation exists.
if hasattr(self.langfuse, "start_as_current_generation"):
clean = {k: v for k, v in kwargs.items() if k != "as_type" and v is not None}
if not self.is_compact_mode():
clean = _inject_langfuse_trace_context(clean, metadata)
try:
with self.langfuse.start_as_current_generation(**clean) as obs:
@@ -367,7 +429,13 @@ class Telemetry:
clean = {k: v for k, v in kwargs.items() if v is not None}
if "as_type" in clean:
clean["as_type"] = _langfuse_type(clean.get("as_type"))
if self.is_compact_mode():
clean.pop("_ignore_current_parent", None)
else:
clean = _inject_langfuse_trace_context(clean, clean.get("metadata") or {})
metadata = clean.get("metadata")
if isinstance(metadata, dict):
clean["metadata"] = {k: v for k, v in metadata.items() if not str(k).startswith("_")}
try:
return self.langfuse.start_as_current_observation(**clean)
except (TypeError, ValueError):
@@ -416,6 +484,8 @@ class Telemetry:
def _update_trace_from_attrs(self, observation, attrs: dict[str, Any]):
if observation is None: return
trace_attrs = {}
if attrs.get("_span_name"):
trace_attrs["name"] = attrs["_span_name"]
for key in ("session_id", "user_id"):
if attrs.get(key): trace_attrs[key] = attrs[key]
if attrs.get("input"): trace_attrs["input"] = attrs["input"]

22
templates/agent_template_backend/.env
View File

@@ -25,8 +25,6 @@ OCI_GENAI_MODEL=openai.gpt-4.1
OCI_GENAI_API_KEY=sk-ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6
OCI_GENAI_PROJECT_OCID=
# OCI_AUTH_MODE=config_file|instance_principal|resource_principal
OCI_AUTH_MODE=config_file
# OCI SDK / signer / profiles
OCI_CONFIG_FILE=~/.oci/config
OCI_PROFILE=DEFAULT
@@ -37,9 +35,10 @@ OCI_REGION=us-chicago-1
# Persistência
###############################################################################
# Opções: memory, autonomous, mongodb
SESSION_REPOSITORY_PROVIDER=autonomous
MEMORY_REPOSITORY_PROVIDER=autonomous
CHECKPOINT_REPOSITORY_PROVIDER=autonomous
SESSION_REPOSITORY_PROVIDER=sqlite
MEMORY_REPOSITORY_PROVIDER=sqlite
CHECKPOINT_REPOSITORY_PROVIDER=sqlite
SQLITE_DB_PATH=./data/agent_framework.db
# Autonomous Database
ADB_USER=admin
@@ -60,8 +59,8 @@ ENABLE_REDIS_CACHE=false
###############################################################################
# RAG / Vector / Graph
###############################################################################
VECTOR_STORE_PROVIDER=memory
GRAPH_STORE_PROVIDER=memory
VECTOR_STORE_PROVIDER=sqlite
GRAPH_STORE_PROVIDER=sqlite
RAG_TOP_K=5
EMBEDDING_PROVIDER=mock
OCI_EMBEDDING_MODEL=cohere.embed-multilingual-v3.0
@@ -71,8 +70,9 @@ RAG_FILE_GLOBS=*.md,*.txt,*.yaml,*.yml,*.json
# Observabilidade
###############################################################################
ENABLE_LANGFUSE=true
LANGFUSE_PUBLIC_KEY=pk-lf-bd9b0c7e-2b8b-4e5b-a382-284a9b4413b3
LANGFUSE_SECRET_KEY=sk-lf-5f5cc18d-0bb5-424e-b5d0-cb3664d58c20
LANGFUSE_TRACE_MODE=verbose # Opcional: verbose, compact
LANGFUSE_PUBLIC_KEY=pk-lf-2f9da109-5b0f-4c78-b61d-9598ed787eba
LANGFUSE_SECRET_KEY=sk-lf-a4cb0cdd-f2ea-4468-9911-cebeb91ba944
LANGFUSE_HOST=http://localhost:3005
ENABLE_OTEL=false
OTEL_EXPORTER_OTLP_ENDPOINT=
@@ -85,7 +85,7 @@ ENABLE_LANGFUSE_OPENAI_AUTO_INSTRUMENTATION=true
# Quando true, AgentObserver publica eventos IC.*, NOC.* e GRL.* nos providers abaixo.
ENABLE_ANALYTICS=false
# Providers aceitos: oci_streaming,pubsub,noop
ANALYTICS_PROVIDERS=oci_streaming
ANALYTICS_PROVIDERS=pubsub
# Compatibilidade FIRST/TIM: pode informar AGENT_PUBSUB_TOPIC diretamente.
AGENT_PUBSUB_TOPIC=
GCP_PUBSUB_TOPIC_PATH=
@@ -150,7 +150,7 @@ MCP_TOOL_TIMEOUT_SECONDS=30
ROUTING_MODE=router
# Usage/cost accounting
USAGE_REPOSITORY_PROVIDER=autonomous
USAGE_REPOSITORY_PROVIDER=sqlite
IDENTITY_CONFIG_PATH=./config/identity.yaml
MCP_PARAMETER_MAPPING_PATH=./config/mcp_parameter_mapping.yaml

3
templates/agent_template_backend/app/main.py
View File

@@ -74,6 +74,7 @@ logger.info("Framework channel input mode: %s", gateway.input_mode)
@app.middleware("http")
async def observability_context_middleware(request: Request, call_next):
clear_observability_context()
request_id = request.headers.get("x-request-id") or str(uuid4())
set_observability_context(
request_id=request_id,
@@ -99,6 +100,8 @@ async def observability_context_middleware(request: Request, call_next):
"duration_ms": int((time.time() - started) * 1000),
}, kind="http")
raise
finally:
clear_observability_context()
class GatewayRequest(BaseModel):

5
templates/agent_template_backend/app/workflows/agent_graph.py
View File

@@ -320,7 +320,6 @@ class AgentWorkflow:
}
async def billing_agent(self, state):
async with self.langgraph_telemetry.node("billing_agent", state):
async with self.telemetry.span(
"workflow.agent.billing",
session_id=state.get("conversation_key") or state.get("session_id"),
@@ -329,7 +328,6 @@ class AgentWorkflow:
return await self.billing.run(state)
async def product_agent(self, state):
async with self.langgraph_telemetry.node("product_agent", state):
async with self.telemetry.span(
"workflow.agent.product",
session_id=state.get("conversation_key") or state.get("session_id"),
@@ -338,7 +336,6 @@ class AgentWorkflow:
return await self.product.run(state)
async def orders_agent(self, state):
async with self.langgraph_telemetry.node("orders_agent", state):
async with self.telemetry.span(
"workflow.agent.orders",
session_id=state.get("conversation_key") or state.get("session_id"),
@@ -346,9 +343,7 @@ class AgentWorkflow:
):
return await self.orders.run(state)
async def support_agent(self, state):
async with self.langgraph_telemetry.node("support_agent", state):
async with self.telemetry.span(
"workflow.agent.support",
session_id=state.get("conversation_key") or state.get("session_id"),

22
templates/agent_template_backend_day_zero/.env
View File

@@ -25,8 +25,6 @@ OCI_GENAI_MODEL=openai.gpt-4.1
OCI_GENAI_API_KEY=sk-ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6
OCI_GENAI_PROJECT_OCID=
# OCI_AUTH_MODE=config_file|instance_principal|resource_principal
OCI_AUTH_MODE=config_file
# OCI SDK / signer / profiles
OCI_CONFIG_FILE=~/.oci/config
OCI_PROFILE=DEFAULT
@@ -37,9 +35,10 @@ OCI_REGION=us-chicago-1
# Persistência
###############################################################################
# Opções: memory, autonomous, mongodb
SESSION_REPOSITORY_PROVIDER=autonomous
MEMORY_REPOSITORY_PROVIDER=autonomous
CHECKPOINT_REPOSITORY_PROVIDER=autonomous
SESSION_REPOSITORY_PROVIDER=sqlite
MEMORY_REPOSITORY_PROVIDER=sqlite
CHECKPOINT_REPOSITORY_PROVIDER=sqlite
SQLITE_DB_PATH=./data/agent_framework.db
# Autonomous Database
ADB_USER=admin
@@ -60,8 +59,8 @@ ENABLE_REDIS_CACHE=false
###############################################################################
# RAG / Vector / Graph
###############################################################################
VECTOR_STORE_PROVIDER=memory
GRAPH_STORE_PROVIDER=memory
VECTOR_STORE_PROVIDER=sqlite
GRAPH_STORE_PROVIDER=sqlite
RAG_TOP_K=5
EMBEDDING_PROVIDER=mock
OCI_EMBEDDING_MODEL=cohere.embed-multilingual-v3.0
@@ -71,8 +70,9 @@ RAG_FILE_GLOBS=*.md,*.txt,*.yaml,*.yml,*.json
# Observabilidade
###############################################################################
ENABLE_LANGFUSE=true
LANGFUSE_PUBLIC_KEY=pk-lf-bd9b0c7e-2b8b-4e5b-a382-284a9b4413b3
LANGFUSE_SECRET_KEY=sk-lf-5f5cc18d-0bb5-424e-b5d0-cb3664d58c20
LANGFUSE_TRACE_MODE=verbose # Opcional: verbose, compact
LANGFUSE_PUBLIC_KEY=pk-lf-2f9da109-5b0f-4c78-b61d-9598ed787eba
LANGFUSE_SECRET_KEY=sk-lf-a4cb0cdd-f2ea-4468-9911-cebeb91ba944
LANGFUSE_HOST=http://localhost:3005
ENABLE_OTEL=false
OTEL_EXPORTER_OTLP_ENDPOINT=
@@ -85,7 +85,7 @@ ENABLE_LANGFUSE_OPENAI_AUTO_INSTRUMENTATION=true
# Quando true, AgentObserver publica eventos IC.*, NOC.* e GRL.* nos providers abaixo.
ENABLE_ANALYTICS=false
# Providers aceitos: oci_streaming,pubsub,noop
ANALYTICS_PROVIDERS=oci_streaming
ANALYTICS_PROVIDERS=pubsub
# Compatibilidade FIRST/TIM: pode informar AGENT_PUBSUB_TOPIC diretamente.
AGENT_PUBSUB_TOPIC=
GCP_PUBSUB_TOPIC_PATH=
@@ -150,7 +150,7 @@ MCP_TOOL_TIMEOUT_SECONDS=30
ROUTING_MODE=router
# Usage/cost accounting
USAGE_REPOSITORY_PROVIDER=autonomous
USAGE_REPOSITORY_PROVIDER=sqlite
IDENTITY_CONFIG_PATH=./config/identity.yaml
MCP_PARAMETER_MAPPING_PATH=./config/mcp_parameter_mapping.yaml

15
templates/agent_template_backend_day_zero/app/main.py
View File

@@ -4,7 +4,7 @@ import logging
from uuid import uuid4
import time
from fastapi import FastAPI, Request
from fastapi import FastAPI, HTTPException, Request
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import StreamingResponse
from pydantic import BaseModel
@@ -52,7 +52,7 @@ summary_memory = create_conversation_summary_memory(settings, message_history=me
sessions = create_session_repository(settings)
checkpoints = create_checkpoint_repository(settings)
cache = create_cache(settings, telemetry=telemetry)
gateway = ChannelGateway()
gateway = ChannelGateway(input_mode=settings.FRAMEWORK_CHANNEL_INPUT_MODE)
analytics = create_analytics_publisher(settings)
observer = TelemetryBackedAgentObserver(telemetry=telemetry)
configure_global_observer({
@@ -70,9 +70,11 @@ logger.info("LLM provider carregado: %s", llm.__class__.__name__)
logger.info("Langfuse habilitado: %s host=%s", telemetry.is_enabled(), settings.LANGFUSE_HOST)
logger.info("Analytics habilitado: %s providers=%s", getattr(settings, "ENABLE_ANALYTICS", False), getattr(settings, "ANALYTICS_PROVIDERS", ""))
logger.info("Agentes disponíveis: %s", [p.agent_id for p in agent_profiles.list_profiles()])
logger.info("Framework channel input mode: %s", gateway.input_mode)
@app.middleware("http")
async def observability_context_middleware(request: Request, call_next):
clear_observability_context()
request_id = request.headers.get("x-request-id") or str(uuid4())
set_observability_context(
request_id=request_id,
@@ -98,6 +100,8 @@ async def observability_context_middleware(request: Request, call_next):
"duration_ms": int((time.time() - started) * 1000),
}, kind="http")
raise
finally:
clear_observability_context()
class GatewayRequest(BaseModel):
@@ -138,7 +142,10 @@ def _resolve_identity(req: GatewayRequest, msg) -> tuple[AgentIdentity, dict, Bu
async def _process_gateway_message(req: GatewayRequest, emit_sse: bool = False) -> dict:
try:
msg = await gateway.normalize(req.channel, req.payload)
except ValueError as exc:
raise HTTPException(status_code=422, detail=str(exc)) from exc
identity, normalized_context, business_context, missing_identity_keys = _resolve_identity(req, msg)
agent_session_id = identity.conversation_key()
message_id = (req.payload or {}).get("message_id") or str(uuid4())
@@ -331,6 +338,8 @@ async def health():
"usage_repository": settings.USAGE_REPOSITORY_PROVIDER,
"identity_config_path": settings.IDENTITY_CONFIG_PATH,
"mcp_parameter_mapping_path": settings.MCP_PARAMETER_MAPPING_PATH,
"framework_channel_input_mode": settings.FRAMEWORK_CHANNEL_INPUT_MODE,
"legacy_channel_gateway_mode": settings.CHANNEL_GATEWAY_MODE,
}
@@ -354,6 +363,8 @@ async def debug_env():
"AGENTS_CONFIG_PATH": settings.AGENTS_CONFIG_PATH,
"ROUTING_CONFIG_PATH": settings.ROUTING_CONFIG_PATH,
"ROUTING_MODE": settings.ROUTING_MODE,
"FRAMEWORK_CHANNEL_INPUT_MODE": settings.FRAMEWORK_CHANNEL_INPUT_MODE,
"CHANNEL_GATEWAY_MODE": settings.CHANNEL_GATEWAY_MODE,
}

92
templates/agent_template_backend_day_zero/app/workflows/agent_graph.py
View File

@@ -187,6 +187,16 @@ class AgentWorkflow:
input=state.get("user_text"),
):
history_texts = [m.get("content", "") for m in state.get("history", [])]
await self.observer.emit_grl(
"001",
{
"session_id": state.get("conversation_key") or state.get("session_id"),
"tenant_id": state.get("tenant_id"),
"agent_id": state.get("agent_id"),
"phase": "input",
},
component="workflow.input_guardrails.start",
)
sanitized, decisions = await self.guardrails.run_input(
state["user_text"],
{
@@ -199,6 +209,19 @@ class AgentWorkflow:
)
for _decision in decisions:
await self.guardrail_telemetry.evaluated("input", _decision)
await self.observer.emit_grl(
"002" if _decision.allowed else "004",
{
"session_id": state.get("conversation_key") or state.get("session_id"),
"tenant_id": state.get("tenant_id"),
"agent_id": state.get("agent_id"),
"phase": "input",
"rail_code": getattr(_decision, "code", None),
"allowed": bool(_decision.allowed),
"reason": getattr(_decision, "reason", None),
},
component="workflow.input_guardrails.decision",
)
if not _decision.allowed:
await self.guardrail_telemetry.blocked("input", _decision)
await self.telemetry.event(
@@ -210,6 +233,18 @@ class AgentWorkflow:
"decisions": [d.model_dump() for d in decisions],
},
)
await self.observer.emit_grl(
"009",
{
"session_id": state.get("conversation_key") or state.get("session_id"),
"tenant_id": state.get("tenant_id"),
"agent_id": state.get("agent_id"),
"phase": "input",
"blocked": any(not d.allowed for d in decisions),
"decision_count": len(decisions),
},
component="workflow.input_guardrails.final",
)
if any(not d.allowed for d in decisions):
return {
"sanitized_input": sanitized,
@@ -285,7 +320,6 @@ class AgentWorkflow:
}
async def billing_agent(self, state):
async with self.langgraph_telemetry.node("billing_agent", state):
async with self.telemetry.span(
"workflow.agent.billing",
session_id=state.get("conversation_key") or state.get("session_id"),
@@ -294,7 +328,6 @@ class AgentWorkflow:
return await self.billing.run(state)
async def product_agent(self, state):
async with self.langgraph_telemetry.node("product_agent", state):
async with self.telemetry.span(
"workflow.agent.product",
session_id=state.get("conversation_key") or state.get("session_id"),
@@ -303,7 +336,6 @@ class AgentWorkflow:
return await self.product.run(state)
async def orders_agent(self, state):
async with self.langgraph_telemetry.node("orders_agent", state):
async with self.telemetry.span(
"workflow.agent.orders",
session_id=state.get("conversation_key") or state.get("session_id"),
@@ -311,9 +343,7 @@ class AgentWorkflow:
):
return await self.orders.run(state)
async def support_agent(self, state):
async with self.langgraph_telemetry.node("support_agent", state):
async with self.telemetry.span(
"workflow.agent.support",
session_id=state.get("conversation_key") or state.get("session_id"),
@@ -419,6 +449,21 @@ class AgentWorkflow:
},
)
await self.observer.emit_ic(
"IC.OUTPUT_SUPERVISOR_COMPLETED",
{
"session_id": context["session_id"],
"tenant_id": state.get("tenant_id"),
"agent_id": state.get("agent_id"),
"route": state.get("route"),
"intent": state.get("intent"),
"action": action,
"approved": decision.approved,
"result_count": len(decision.results),
},
component="workflow.output_supervisor",
)
if decision.action in {RailAction.ALLOW, RailAction.SANITIZE, RailAction.OBSERVE}:
final_answer = decision.candidate
elif decision.action == RailAction.HANDOVER:
@@ -457,11 +502,36 @@ class AgentWorkflow:
session_id=state.get("conversation_key") or state.get("session_id"),
input=state.get("answer"),
):
await self.observer.emit_grl(
"001",
{
"session_id": state.get("conversation_key") or state.get("session_id"),
"tenant_id": state.get("tenant_id"),
"agent_id": state.get("agent_id"),
"phase": "output",
"route": state.get("route"),
"intent": state.get("intent"),
},
component="workflow.output_guardrails.start",
)
final, decisions = await self.guardrails.run_output(
state["answer"], state.get("context", {})
)
for _decision in decisions:
await self.guardrail_telemetry.evaluated("output", _decision)
await self.observer.emit_grl(
"002" if _decision.allowed else "004",
{
"session_id": state.get("conversation_key") or state.get("session_id"),
"tenant_id": state.get("tenant_id"),
"agent_id": state.get("agent_id"),
"phase": "output",
"rail_code": getattr(_decision, "code", None),
"allowed": bool(_decision.allowed),
"reason": getattr(_decision, "reason", None),
},
component="workflow.output_guardrails.decision",
)
if not _decision.allowed:
await self.guardrail_telemetry.blocked("output", _decision)
await self.telemetry.event(
@@ -473,6 +543,18 @@ class AgentWorkflow:
"decisions": [d.model_dump() for d in decisions],
},
)
await self.observer.emit_grl(
"009",
{
"session_id": state.get("conversation_key") or state.get("session_id"),
"tenant_id": state.get("tenant_id"),
"agent_id": state.get("agent_id"),
"phase": "output",
"blocked": any(not d.allowed for d in decisions),
"decision_count": len(decisions),
},
component="workflow.output_guardrails.final",
)
return {
"final_answer": final,
"guardrail_decisions": state.get("guardrail_decisions", [])