Sandbox test: a 26 s outage exhausted the previous 15 s retry budget and failed a REPROCESS_ALL.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Found by uninstall + fresh install on the clean Windows Sandbox: the old agent_state.sqlite identity conflicted with the new enrollment and the service sat idle.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
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>
- resolver (rules RTC-R0..R4/A1..A3/G1..G2) over the official cClassTrib table 2026-06-22 and LC 214/2025 annexes; local vs official catalog comparison
- ONE root-cause RTC blocker per product (missing/unresolved/ambiguous/conflict/official mismatch/catalog outdated/evidence missing)
- auto-apply: C000025.CCLASSTRIB only, three switches (FIREBIRD_WRITES + RTC_AUTO_APPLY_ENABLED + Hub authorization), per-product re-read and re-decision, never overwrites, rowcount 1, audit table rtc_write_audits (migration 0027)
- copy-only write E2E (280/280 applied, only CCLASSTRIB changed, 2nd run 0 writes, stale decision 0 writes); real databases untouched
- docs: cclasstrib_resolution, rtc_auto_apply, unified_fiscal_completeness
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Hub: installations.processing_enabled / sync_interval_seconds / last_sync_* (migration 0026).
The first heartbeat of an installation without coverage queues exactly one INITIAL_SYNC;
afterwards one SYNC_INCREMENTAL per interval (idempotent, backs off on failure). One fiscal
job at a time per Agent (FIFO, stale RUNNING recovery). Job completion now updates the
coverage snapshot, blockers and JOB_STARTED/COMPLETED/FAILED events. Panel: Sincronizar
agora / Reprocessar tudo / Pausar / Retomar + status; pause is enforced Hub-side.
Agent: local AGENT_JOBS_ENABLED is ignored (central control); SYNC_INCREMENTAL re-classifies
only new/changed products via a per-product fingerprint+decision cache and rebuilds coverage
from it; results carry operational summary and blockers; Firebird credentials come from a
DPAPI firebird.bin. Setup: read-only SGV Firebird discovery (auto-select only when
unambiguous, wizard chooser otherwise), starts the service, waits for the first heartbeat,
upgrade-in-place mode. All jobs stay read-only (FIREBIRD_WRITES=0). Version 1.1.3.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
--env-file only feeds the containers; a plain shell script never saw
BACKUP_DIR, so it silently fell back to ./backups while the Hub itself used
the real value from .env. Found in production: postgres's /backups bind
mount was still pinned to a directory orphaned by an earlier git-archive
swap (mv gti-fiscal gti-fiscal_old && mv gti-fiscal_new gti-fiscal — Linux
bind mounts follow the original dentry, not the path string), so every
backup.sh run had been silently landing dumps nobody could find at the
live path. Fixed by pointing BACKUP_DIR at a stable absolute path outside
the deploy tree and recreating postgres once to pick it up; verified with
a real ./backup.sh run in production.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
backup.sh now also streams the fiscal-hub container's release binaries
(dist tarball, never in Postgres) into BACKUP_DIR alongside the pg_dump,
with the same 14-backup retention; restore_releases.sh is the companion
restore. Verified with a real restore smoke test in production: the
restored dump's agent_releases row and the restored tarball's exe both
match the published sha256 exactly.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Tenant gains source (LOCAL|GTI_GERENCIADOR), external_client_id, external_code,
active and synced_at (migration 0024). central/app/client_sync.py fetches from
the panel's new GET /internal/fiscal/clients (dedicated bearer token, never
MySQL credentials) and upserts Tenant rows, never touching a LOCAL row.
/agents/new now offers only active GTI_GERENCIADOR tenants, with a client-side
search box; the POST validates the same rule server-side, not just in the
dropdown. central/scripts/sync_clients_from_gerenciador.py is the operator-run
sync + reconciliation report.
GTI_FISCAL_VERSION/AGENT_VERSION bumped to 1.1.2 (release/VERSION single
source of truth), incorporating the installation_id persistence fix. Rebuilt
the EXE installer trio at 1.1.2 (scripts/build_windows_installer.ps1).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
One universal Windows Agent package per version (agent_releases: version/
platform/architecture/sha256/status DRAFT|PUBLISHED|REVOKED; the binary
lives on a local release-storage volume, never in PostgreSQL). An operator
creates a pending installation (agent_enrollments) from Agents / Servidores
-> + Nova instalação, then either downloads it directly or generates a
one-time HTTPS link (short TTL, single use). Either path streams the
package plus a freshly minted enrollment.json whose token is never written
to storage in plaintext, only its SHA-256 hash. On the client host,
gti-agent.ps1 install-auto verifies the bundle's SHA-256, refuses a non-
HTTPS Hub, enrolls via POST /v1/agents/enroll (deleting the enrollment
token immediately after, success or failure), writes an ACL-restricted
.env with the newly issued permanent Agent token, and registers the
Windows Service without starting it (Firebird stays unconfigured on
purpose). Idempotent, replay-proof, rate-limited per IP, fully audited,
never logs a secret.
Migration 0022_agent_releases_enrollment (verified fresh-DB ==
upgrade-from-0021 by pg_dump diff). GTI_FISCAL_VERSION/AGENT_VERSION
bumped to 1.1.0 as the single source of truth across code, packages, the
release manifest and every published AgentRelease row.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Collaborators logged into GTI Gerenciador open GTI Fiscal without a
second password. The panel signs a 60 s JWT (Ed25519; the Hub holds only
the public key); the Hub checks signature, iss/aud/exp/nbf, maximum
lifetime, nonce against a one-time state cookie set when the flow starts
at GET /login (login-CSRF protection) and a single-use jti
(sso_used_assertions). Collaborators are provisioned in admin_users with
auth_source=GTI_GERENCIADOR and no password; a CHECK constraint makes a
passwordless local account or a federated account with a password
impossible (migration 0021). Sessions last FEDERATED_SESSION_SECONDS
(1 h) and are renewed silently through the panel.
The local admin stays as break-glass at /login/local, audited. Logout ends
only the Hub session. Every SSO success/rejection (stable reason code),
logout and local login is audited with the real client IP, never the
token. verify_password now fails closed on a missing hash instead of
raising.
Visual identity: gti-theme.css is a verbatim copy of the panel's
design system (tokens, sidebar, light/dark theme) plus its tema.js,
sidebar.js and logo; control-panel.css keeps only Hub components, all
colours from the panel tokens. Page structure and features unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
GTI terminates TLS in its Nginx Proxy Manager (10.0.10.100), not on the
Hub host, so the Hub must be reachable on the LAN — by the proxy only.
- docker-compose.prod.yml passes FORWARDED_ALLOW_IPS (uvicorn then trusts
X-Forwarded-For/-Proto from the proxy only: real client IPs in audit and
session records, https scheme) and WEB_COOKIE_SECURE. Defaults keep the
previous behaviour.
- restrict_hub_port.sh + gti-fiscal-hub-port.service: Docker-published
ports bypass UFW, so the source restriction lives in DOCKER-USER,
matched on the conntrack original destination with --ctdir ORIGINAL,
and is re-applied whenever docker.service starts (the chain is not
persisted across reboots).
- *.service forced to LF like *.sh; tests cover both.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Found by running deploy/hub/backup.sh on the production VM:
- With core.autocrlf=true, `git archive` on the Windows release machine
shipped the .sh files with CRLF and `sh` aborted on line 4
("set: Illegal option -"): backup, restore and seed never ran.
.gitattributes now forces *.sh to LF.
- Docker creates the bind-mounted backups/ directory as root, so the
host-side retention `rm` would fail once the 15th dump existed.
Retention now runs inside the postgres container.
tests/test_deploy_scripts.py covers both.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
gtiandClaude Sonnet 5
OpenCode