fix(backend): alembic drift PR-B — F3 14x index=True + F7 33x server_default (Option A) + env.py compare_* flags

[adamatdevops]

Closes the MEDIUM canonicalization findings from the Alembic drift recon at research/db_audit/ALEMBIC_DRIFT_RECON.md v0.1. Second of three PRs in the AB-037 A/B/C sequence — PR-A (#65) shipped the runtime-bug fixes; this PR aligns model↔migration representation; PR-C will land the CI gate. Filed as AB-037B in the backlog.

What this PR does

Aligns the SQLAlchemy model metadata with what the migrations actually build against Postgres, so alembic revision --autogenerate produces a clean empty diff going forward — instead of emitting spurious op.drop_index() for 14 valid indexes + missing 33 server_default drift signals + missing type change signals.

F3 — 14x index=True added

Table	Columns
actions	type, target_id, created_at
anomalies	service_id, severity, is_active
recommendations	team_id, workload_type
refresh_tokens	expires_at
services	team_id, template_id, status
users	role, is_active

SQLAlchemy's default naming convention produces ix_<table>_<col> — matches the migration index names exactly for all 14. Column types preserved.

F7 — 33x server_default= mirrored (Option A)

All 33 server_default= declarations from the migrations are now mirrored in the models (matching values verbatim). Chose Option A over Option B because zero non-Python writers found in the codebase (grep-verified across airflow-worker, flink-jobs, normalizer, webhook-gateway, job-dispatcher) and zero migration seed-data, so explicit `server_default=` is defense-in-depth for any future writer AND enables `compare_server_default=True` to give accurate autogen signal.

Both `default=` (Python-side) and `server_default=` (DB-side) are preserved — Python code paths continue to work identically.

env.py — compare flags enabled

Both `context.configure(...)` calls now pass `compare_server_default=True` + `compare_type=True`. Without these, autogen silently misses the drift classes that led to the 2026-09-01 recon (F5 ENUM, F6 token_hash length, F7 server_defaults).

Verification

Author-side:

• All 8 models import cleanly
• Model server_default coverage: templates(11) + services(7) + recommendations(5) + anomalies(4) + actions(3) + users(3) = 33 ✓
• Model index count: 15 = migration-alive indexes (15) → parity after PR-A dropped 2 redundant

Final `alembic revision --autogenerate should emit empty` verification requires a fresh dev Postgres — deferred to AB-037C's `alembic check` CI gate.

Scope

7 files, ~60 lines of substantive change (plus formatting):

• `src/backend/alembic/env.py` — 2x compare_* flags
• `src/backend/app/db/models/template.py` — 11x server_default
• `src/backend/app/db/models/service.py` — 7x server_default + 3x index=True
• `src/backend/app/db/models/anomaly.py` — 4x server_default + 3x index=True
• `src/backend/app/db/models/recommendation.py` — 5x server_default + 2x index=True
• `src/backend/app/db/models/action.py` — 3x server_default + 3x index=True
• `src/backend/app/db/models/user.py` — 3x server_default + 3x index=True (2 User + 1 RefreshToken)

`src/backend/app/db/models/team.py` — no changes (no F3/F7 columns).

Runtime impact

None. Metadata-only changes at the SQLAlchemy binding layer. No DDL, no data migration, no runtime behavior change.

References

• Recon report (local-only, gitignored): `research/db_audit/ALEMBIC_DRIFT_RECON.md` v0.1
• Codex loop audit trail (local-only, gitignored): `research/feedback_loops/db_audit-ALEMBIC_DRIFT_RECON/20260901T081525Z/`
• PR-A (runtime fixes): fix(backend): alembic drift PR-A — F1 actions rename + F4/F6 index drops + F5 ENUM values_callable #65 merged as `164371a`
• Backlog: AB-037A ✅ (fix(backend): alembic drift PR-A — F1 actions rename + F4/F6 index drops + F5 ENUM values_callable #65), AB-037B (this PR), AB-037C (blocked on this PR)

🤖 Generated with Claude Code

https://claude.ai/code/session_015G6Qu6rwJZYWjMdxU5bMVt

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Team

Run ID: 1fc6347f-5962-4833-b62f-a86682fc6818

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ All tests successful. No failed tests found.

@@            Coverage Diff             @@
##             main      #66      +/-   ##
==========================================
+ Coverage   52.99%   53.02%   +0.03%     
==========================================
  Files         110      110              
  Lines        6337     6337              
  Branches     1058     1058              
==========================================
+ Hits         3358     3360       +2     
+ Misses       2872     2871       -1     
+ Partials      107      106       -1     

Flag	Coverage Δ
backend	63.86% <100.00%> (+0.04%)	⬆️
frontend	28.25% <ø> (ø)

Files with missing lines	Coverage Δ
src/backend/app/db/models/action.py	98.00% <100.00%> (ø)
src/backend/app/db/models/anomaly.py	97.82% <100.00%> (ø)
src/backend/app/db/models/recommendation.py	96.77% <100.00%> (ø)
src/backend/app/db/models/service.py	97.91% <100.00%> (ø)
src/backend/app/db/models/template.py	96.96% <100.00%> (ø)
src/backend/app/db/models/user.py	90.69% <100.00%> (ø)

... and 3 files with indirect coverage changes

---

Continue to review full report in Codecov by Harness.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 164371a...fa87705. Read the comment docs.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.