bot-backend
3a13bce468
All checks were successful
CI / changes (pull_request) Successful in 11s
CI Trade-In / changes (pull_request) Successful in 11s
CI / frontend-tests (pull_request) Has been skipped
CI Trade-In / backend-tests (pull_request) Has been skipped
CI Trade-In / frontend-checks (pull_request) Has been skipped
CI / openapi-codegen-check (pull_request) Successful in 3m1s
CI / backend-tests (pull_request) Successful in 16m4s
One layer's caught DB error left the shared §22 report Session in Postgres's "aborted transaction" state with no rollback, silently emptying every later layer on the same report. Wrap each failure point in db.begin_nested() (SAVEPOINT): orchestrator._safe_call, macro_series _query_key_rate_monthly / _query_inflation_monthly / _query_mortgage_monthly (per indicator), special_indices._run (per index), and sales_series _query_source_a / _query_source_b — so a sibling builder's failure no longer cascades. Also close the RELEASE-SAVEPOINT trap: inner helpers that swallowed a db.execute failure without their own savepoint left the tx aborted, so the outer builder's savepoint failed at RELEASE (illegal in an aborted tx) and poisoning still cascaded. Add savepoints at the inner catch sites — special_indices._query_parcel_centroid and district_resolver._admin_names / resolve_objective_districts. Refs #2464 (cluster A session-poisoning). |
||
|---|---|---|
| .. | ||
| __init__.py | ||
| affordability.py | ||
| confidence_engine.py | ||
| demand_normalization.py | ||
| demand_supply_forecast.py | ||
| macro_coefficient.py | ||
| macro_series.py | ||
| normalize.py | ||
| orchestrator.py | ||
| product_scoring.py | ||
| rate_sensitivity.py | ||
| recommendation.py | ||
| regression.py | ||
| report.py | ||
| report_assembler.py | ||
| sales_series.py | ||
| scenarios.py | ||
| special_indices.py | ||
| what_to_build.py | ||