fix: Agent retries transient transport errors/proxy 5xx with backoff (classify, complete, fail); hide console behind the GUI wizard; 1.2.2

Found in the clean-Windows recovery test: a single TLS handshake failure mid-job failed a whole sync.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This commit is contained in:
gti
2026-09-25 12:02:11 -03:00
co-authored by Claude Sonnet 5
parent abeb6ad878
commit 2e5598ce9c
21 changed files with 94 additions and 34 deletions
+29 -10
View File
@@ -1,5 +1,29 @@
import time
import httpx
TRANSIENT_STATUS = (502, 503, 504) # reverse proxy / Hub restarting
BACKOFF_SECONDS = (1, 2, 4, 8) # 5 attempts in total
def _resilient(send):
"""Runs one idempotent HTTP call, retrying transport failures (connect/TLS/read/protocol errors) and proxy 502/503/504 with backoff, so a network blip in
the middle of a long job does not throw the whole job (and minutes of classification) away. Non-transient errors (4xx) are returned untouched."""
last = None
for attempt in range(len(BACKOFF_SECONDS) + 1):
try:
response = send()
if response.status_code not in TRANSIENT_STATUS or attempt == len(BACKOFF_SECONDS):
return response
last = None
except httpx.TransportError as error:
last = error
if attempt == len(BACKOFF_SECONDS):
raise
time.sleep(BACKOFF_SECONDS[attempt])
raise last # pragma: no cover
class HubClient:
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}"}
@@ -7,14 +31,9 @@ class HubClient:
results = []
for start in range(0, len(products), self.chunk_size):
chunk = products[start:start + self.chunk_size]
# 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
# Small chunks + generous timeout + retry with backoff on any transport error or proxy 5xx: a large first sync must survive
# a slow Hub, a reverse-proxy read timeout or a TLS/network blip instead of failing the whole job.
response = _resilient(lambda: httpx.post(f"{self.url}/v1/classify/batch", json={"products": chunk}, headers=self._headers(), timeout=120))
response.raise_for_status(); results.extend(response.json()["results"])
return results
def get_json(self, path):
@@ -48,8 +67,8 @@ class HubClient:
response = httpx.get(f"{self.url}/v1/jobs/{job_id}", headers=self._headers(), timeout=15)
response.raise_for_status(); return response.json()
def complete_job(self, job_id, result_summary):
response = httpx.post(f"{self.url}/v1/jobs/{job_id}/complete", json={"result_summary": result_summary}, headers=self._headers(), timeout=15)
response = _resilient(lambda: httpx.post(f"{self.url}/v1/jobs/{job_id}/complete", json={"result_summary": result_summary}, headers=self._headers(), timeout=30))
response.raise_for_status(); return response.json()
def fail_job(self, job_id, error):
response = httpx.post(f"{self.url}/v1/jobs/{job_id}/fail", json={"error": str(error)}, headers=self._headers(), timeout=15)
response = _resilient(lambda: httpx.post(f"{self.url}/v1/jobs/{job_id}/fail", json={"error": str(error)}, headers=self._headers(), timeout=30))
response.raise_for_status(); return response.json()
+1 -1
View File
@@ -9,7 +9,7 @@ import os
import sqlite3
from pathlib import Path
AGENT_VERSION_DEFAULT = "1.2.1"
AGENT_VERSION_DEFAULT = "1.2.2"
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")