fix: Agent survives slow Hub on large syncs (25-product chunks, 120s timeout, retry) and keeps heartbeating during jobs; 1.1.4

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
2026-09-24 09:33:54 -03:00
parent d1104589cc
commit e40e9500a4
19 changed files with 75 additions and 26 deletions
@@ -1,6 +1,6 @@
 [project]
 name = "gti-fiscal-agent"
-version = "1.1.3"
+version = "1.1.4"
 requires-python = ">=3.10"
 dependencies = ["httpx>=0.27,<1"]

@@ -20,6 +20,7 @@ import logging.handlers
 import os
 import signal
 import sys
+import threading
 import time
 from pathlib import Path

@@ -102,7 +103,7 @@ class Config:
        # (A legacy AGENT_JOBS_ENABLED line in an old agent.env is deliberately ignored.)
        self.agent_jobs_enabled = True
        self.poll_interval_seconds = float(env.get("POLL_INTERVAL_SECONDS", "5"))
-        self.agent_version = env.get("AGENT_VERSION", "1.1.3")
+        self.agent_version = env.get("AGENT_VERSION", "1.1.4")
        self.schema_fingerprint = env.get("SCHEMA_FINGERPRINT", REFERENCE_SCHEMA_V1)
        self.state_db = env.get("AGENT_STATE_DB", str(AGENT_ROOT / "agent_state.sqlite"))
        # Pilot write authorization: both must be explicitly set by the installer for
@@ -239,7 +240,7 @@ def execute_job(job, config, executor, state, env):
            "coverage": coverage,
            "operational": operational,
            "fiscal_mode": getattr(config, "fiscal_mode", None),
-            "release_version": env.get("AGENT_VERSION") or "1.1.3",
+            "release_version": env.get("AGENT_VERSION") or "1.1.4",
            "products_total": len(products),
            "products_changed": len(to_process),
            "products_processed": summary["processed"],
@@ -268,6 +269,17 @@ def execute_job(job, config, executor, state, env):
        return None, f"{type(error).__name__}: {error}"


+BUSY_HEARTBEAT_SECONDS = 20
+
+
+def _busy_heartbeats(client, installation_id, config, job_id, done):
+    while not done.wait(BUSY_HEARTBEAT_SECONDS):
+        try:
+            client.heartbeat_metadata({"installation_id": installation_id, "status": "BUSY", "schema_fingerprint": config.schema_fingerprint, "agent_version": config.agent_version, "last_job_id": job_id})
+        except Exception:  # noqa: BLE001 — Hub hiccup mid-job must not affect the job
+            pass
+
+
 def resolve_installation_id(config, env, state, client):
    """Recover and persist the Hub-bound identity; never invent one for an Agent."""
    persisted = state.stored_installation_id()
@@ -362,7 +374,16 @@ def run(config, env, *, max_iterations=None, stopper=None):
        except Exception:
            pass  # best-effort; the job itself still proceeds

-        result_summary, error = execute_job(job, config, executor, state, env)
+        # A full sync can take minutes: keep the heartbeat alive (BUSY) from a side thread so the
+        # panel never shows a working Agent as OFFLINE. The side thread only talks to the Hub
+        # (no SQLite/Firebird), so it cannot interfere with the job.
+        done = threading.Event()
+        beat = threading.Thread(target=_busy_heartbeats, args=(client, installation_id, config, last_job_id, done), daemon=True)
+        beat.start()
+        try:
+            result_summary, error = execute_job(job, config, executor, state, env)
+        finally:
+            done.set()
        try:
            if error is None:
                client.complete_job(job.get("job_id"), result_summary)
@@ -377,7 +398,7 @@ def run(config, env, *, max_iterations=None, stopper=None):
 def enforce_startup_preflight(env):
    """Full read-only preflight at startup (Hub, auth, Firebird, capabilities, company context, guards). Fail closed."""
    from gti_fiscal_agent.preflight import run_preflight
-    report = run_preflight(env, agent_version=env.get("AGENT_VERSION") or "1.1.3")
+    report = run_preflight(env, agent_version=env.get("AGENT_VERSION") or "1.1.4")
    logger.info("startup preflight: %s", json.dumps({k: report[k] for k in ("PREFLIGHT", "blockers", "fiscal_mode", "effective_firebird_writes")}, ensure_ascii=False))
    print(json.dumps({"PREFLIGHT": report["PREFLIGHT"], "blockers": report["blockers"]}, ensure_ascii=False))
    if report["PREFLIGHT"] != "PASS":
@@ -1,13 +1,20 @@
 import httpx

 class HubClient:
-    def __init__(self, url, token, chunk_size=100): self.url, self.token, self.chunk_size = url.rstrip("/"), token, chunk_size
+    def __init__(self, url, token, chunk_size=25): self.url, self.token, self.chunk_size = url.rstrip("/"), token, chunk_size
    def _headers(self): return {"Authorization": f"Bearer {self.token}"}
    def classify_batch(self, products):
        results = []
        for start in range(0, len(products), self.chunk_size):
            chunk = products[start:start + self.chunk_size]
-            response = httpx.post(f"{self.url}/v1/classify/batch", json={"products": chunk}, headers=self._headers(), timeout=60)
+            # Small chunks + generous timeout + one retry: a large first sync must survive a
+            # slow Hub / reverse-proxy read timeout instead of failing the whole job.
+            for attempt in (1, 2):
+                try:
+                    response = httpx.post(f"{self.url}/v1/classify/batch", json={"products": chunk}, headers=self._headers(), timeout=120)
+                    break
+                except httpx.ReadTimeout:
+                    if attempt == 2: raise
            response.raise_for_status(); results.extend(response.json()["results"])
        return results
    def get_json(self, path):
@@ -9,7 +9,7 @@ import os
 import sqlite3
 from pathlib import Path

-AGENT_VERSION_DEFAULT = "1.1.3"
+AGENT_VERSION_DEFAULT = "1.1.4"
 REQUIRED_ENV = ("GTI_HUB_URL", "AGENT_TOKEN", "FIREBIRD_DATABASE_PATH", "FB_USER", "FB_PASSWORD")
 REQUIRED_CAPABILITIES = ("supports_ncm", "supports_ipi", "supports_cest")
 FISCAL_MODES = ("DRY_RUN", "AUTO_APPLY")
@@ -1,6 +1,7 @@
 """Agent side of the zero-touch model: delta sync via fingerprint cache, read-only guarantee,
 blockers/operational summary in the job result, central pause, Firebird credential resolution."""
 import json
+import time

 import pytest

@@ -125,3 +126,23 @@ def test_central_resume_after_pause_makes_agent_poll_again(monkeypatch, tmp_path
    monkeypatch.setattr(service.time, "sleep", lambda s: None)
    service.run(_config(agent_id="2", poll_interval_seconds=0, state_db=str(tmp_path / "a.db")), {"AGENT_ID": "2"}, stopper=stopper)
    assert calls["next_job"] == 1  # first cycle paused (no poll), second cycle resumed (polled)
+
+
+def test_heartbeats_keep_flowing_while_a_long_job_runs(monkeypatch, tmp_path):
+    beats = []
+
+    class Hub:
+        def identity(self): return {"agent_id": 2, "installation_id": "00000000-0000-4000-8000-000000000222"}
+        def heartbeat_metadata(self, payload): beats.append(payload["status"]); return {"processing_enabled": True}
+        def next_job(self): return {"job_id": 5, "kind": "INITIAL_SYNC", "agent_id": 2, "payload": {}}
+        def complete_job(self, job_id, summary): stopper.stop = True
+        def fail_job(self, job_id, error): stopper.stop = True
+
+    class Stopper: stop = False
+    stopper = Stopper()
+    monkeypatch.setattr(service, "BUSY_HEARTBEAT_SECONDS", 0.05)
+    monkeypatch.setattr(service, "HubClient", lambda *a, **k: Hub())
+    monkeypatch.setattr(service, "load_executor", lambda: _FakeExecutor())
+    monkeypatch.setattr(service, "execute_job", lambda *a, **k: (time.sleep(0.4), ({"ok": 1}, None))[1])
+    service.run(_config(agent_id="2", poll_interval_seconds=0, state_db=str(tmp_path / "a.db")), {"AGENT_ID": "2"}, stopper=stopper)
+    assert beats.count("BUSY") >= 3  # heartbeats during the job, not only before it
@@ -4,5 +4,5 @@ kept identical to this) and what an AgentRelease row's `version` column holds
 once published — never a second, independent version number. Fiscal semantics
 are versioned separately (ENGINE_VERSION/DECISION_VERSION in classification.py,
 DECISION_ENGINE_VERSION etc.) and are NOT changed by a release bump."""
-GTI_FISCAL_VERSION = "1.1.3"
+GTI_FISCAL_VERSION = "1.1.4"
 API_COMPAT_MAJOR_MINOR = "1.0"  # an Agent 1.0.x/1.1.x talks to any Hub 1.0.x
@@ -1,6 +1,6 @@
 [project]
 name = "gti-fiscal-hub"
-version = "1.1.3"
+version = "1.1.4"
 requires-python = ">=3.10"
 dependencies = [
  "fastapi>=0.115,<1",
@@ -17,7 +17,7 @@ def test_health_is_unavailable_when_database_is_down():

 def test_version_endpoint_reports_release_and_fiscal_versions():
    body = TestClient(create_app(database_probe=lambda: True)).get("/version").json()
-    assert body["version"] == "1.1.3" and body["api_compat"] == "1.0" and body["engine_version"] and body["decision_version"]
+    assert body["version"] == "1.1.4" and body["api_compat"] == "1.0" and body["engine_version"] and body["decision_version"]


 def test_production_refuses_the_dev_token(monkeypatch):
@@ -19,4 +19,4 @@ FB_PASSWORD=
 FISCAL_MODE=DRY_RUN
 FIREBIRD_WRITES=0
 POLL_INTERVAL_SECONDS=5
-AGENT_VERSION=1.1.3
+AGENT_VERSION=1.1.4
@@ -164,7 +164,7 @@ switch ($Command) {
                "GTI_HUB_URL=$hub", "AGENT_TOKEN=$token", "AGENT_ID=$agentId",
                "COMPANY_CODE=$code", "COMPANY_CNPJ=$cnpj", "COMPANY_CRT_SOURCE=$crtSource",
                "FIREBIRD_DATABASE_PATH=$db", "FB_USER=$fbUser", "FB_PASSWORD=$fbPass",
-                'FISCAL_MODE=DRY_RUN', 'FIREBIRD_WRITES=0', 'POLL_INTERVAL_SECONDS=5', 'AGENT_VERSION=1.1.3'
+                'FISCAL_MODE=DRY_RUN', 'FIREBIRD_WRITES=0', 'POLL_INTERVAL_SECONDS=5', 'AGENT_VERSION=1.1.4'
            )
            Set-Content -Path $EnvFile -Value $lines -Encoding UTF8
        }
@@ -19,4 +19,4 @@ FB_PASSWORD=
 FISCAL_MODE=DRY_RUN
 FIREBIRD_WRITES=0
 POLL_INTERVAL_SECONDS=5
-AGENT_VERSION=1.1.3
+AGENT_VERSION=1.1.4
@@ -19,4 +19,4 @@ FB_PASSWORD=
 FISCAL_MODE=DRY_RUN
 FIREBIRD_WRITES=0
 POLL_INTERVAL_SECONDS=5
-AGENT_VERSION=1.1.3
+AGENT_VERSION=1.1.4
@@ -19,4 +19,4 @@ FB_PASSWORD=
 FISCAL_MODE=DRY_RUN
 FIREBIRD_WRITES=0
 POLL_INTERVAL_SECONDS=5
-AGENT_VERSION=1.1.3
+AGENT_VERSION=1.1.4
@@ -19,4 +19,4 @@ FB_PASSWORD=
 FISCAL_MODE=DRY_RUN
 FIREBIRD_WRITES=0
 POLL_INTERVAL_SECONDS=5
-AGENT_VERSION=1.1.3
+AGENT_VERSION=1.1.4
@@ -1,5 +1,5 @@
 # GTI Fiscal Hub — production environment (copy to .env, never commit .env)
-GTI_FISCAL_VERSION=1.1.3
+GTI_FISCAL_VERSION=1.1.4

 # REQUIRED. URL-safe secret (used inside a connection URL): openssl rand -hex 24
 POSTGRES_PASSWORD=
@@ -36,7 +36,7 @@ services:
    build:
      context: ../..
      dockerfile: central/Dockerfile
-    image: gti-fiscal-hub:${GTI_FISCAL_VERSION:-1.1.3}
+    image: gti-fiscal-hub:${GTI_FISCAL_VERSION:-1.1.4}
    restart: unless-stopped
    depends_on:
      postgres:
@@ -1,14 +1,14 @@
 {
 "artifact": "RELEASE_MANIFEST",
- "GTI_FISCAL_VERSION": "1.1.3",
- "generated_at": "2026-09-24T11:46:32+00:00",
+ "GTI_FISCAL_VERSION": "1.1.4",
+ "generated_at": "2026-09-24T12:33:17+00:00",
 "hub": {
-  "version": "1.1.3",
+  "version": "1.1.4",
  "api_compat": "1.0",
  "engine_version": "0.3.0",
  "decision_version": "mission-04c-ipi-ex-split"
 },
- "agent_version": "1.1.3",
+ "agent_version": "1.1.4",
 "fiscal_engine": {
  "ENGINE_VERSION": "0.3.0",
  "DECISION_VERSION": "mission-04c-ipi-ex-split",
@@ -1 +1 @@
-1.1.3
+1.1.4
@@ -14,11 +14,11 @@ sys.path.insert(0, str(ROOT))
 def test_versions_are_coherent_across_hub_agent_and_packages():
    from central.app.version import GTI_FISCAL_VERSION
    v = (ROOT / "release" / "VERSION").read_text(encoding="utf-8").strip()
-    assert v == GTI_FISCAL_VERSION == "1.1.3"
+    assert v == GTI_FISCAL_VERSION == "1.1.4"
    pep = v.replace("-rc", "rc")
    for pp in ("central/pyproject.toml", "agent/pyproject.toml"):
        assert f'version = "{pep}"' in (ROOT / pp).read_text(encoding="utf-8")
-    assert '"AGENT_VERSION", "1.1.3"' in (ROOT / "agent" / "scripts" / "run_agent_service.py").read_text(encoding="utf-8")
+    assert '"AGENT_VERSION", "1.1.4"' in (ROOT / "agent" / "scripts" / "run_agent_service.py").read_text(encoding="utf-8")
    manifest = ROOT / "release" / "RELEASE_MANIFEST.json"
    if manifest.exists():
        m = json.loads(manifest.read_text(encoding="utf-8"))
@@ -1 +1 @@
 .1.3
 .1.4