3 supported public-file passes
Copied public inputs landed inside the documented SQLAlchemy automation subset.
Public proof
SQLAlchemy migration proof on public files
Public SQLAlchemy migration proof for the supported cleanup subset, including validated supported files and fail-closed manual-review examples.
2 blocked manual-review cases
Risky execution and query-shape patterns were reported instead of guessed through.
0 broad migration promises
The proof covers the narrow legacy query and syntax subset, not every SQLAlchemy 2.0 change.
Trust boundary
Do not trust this blindly with your project
- Visible GitHub stars and forks are weak or absent social proof; treat the public repo as a fit signal, not as proof that the tool is safe for your repo.
- The public repo is scanner-first and does not expose the full paid apply engine.
- Public proof is intentionally narrow: a few supported examples plus fail-closed examples. Run on a branch and trust changes only after your own typecheck, build, and tests pass.
Supported examples
Public files that matched the automation subset
Bogdanp/flask_dramatiq_example
Query.get plus declarative import cleanup
supported and validated on copied public input
dunossauro/live-de-python
select([table]) list syntax cleanup
supported and validated on copied public input
nylas/sync-engine
Query.get plus string loader cleanup
supported and validated on copied public input
Fail-closed examples
Public files that stayed in manual review
nylas/sync-engine
engine.execute(...)
manual_review_required
teamclairvoyant/airflow-maintenance-dags
Query.from_self()
manual_review_required
Artifact trail
One public-proof run, shown end to end
Sample repo before
from sqlalchemy.ext.declarative import declarative_base
Base = declarative_base()
job = session.query(Job).get(job_id)
stmt = select([Job.id, Job.name])Scan JSON
{
"status": "preview_only",
"transform_count": 3,
"supported_candidates": {
"declarative_imports": 1,
"query_get": 1,
"select_list_syntax": 1
},
"manual_todos": []
}Preview diff
-from sqlalchemy.ext.declarative import declarative_base
+from sqlalchemy.orm import declarative_base
-job = session.query(Job).get(job_id)
+job = session.get(Job, job_id)
-stmt = select([Job.id, Job.name])
+stmt = select(Job.id, Job.name)Validation result
validation:
python -m compileall copied_public_input ... passed
scanner report generated ... passed
unsupported findings ... 0Apply output
apply output:
files changed: 1
transforms applied: 3
report written: migration-report.json
mode: preview/apply branch workflowFinal manager summary
manager summary:
supported cleanup findings: 3
manual-review findings: 0
next step: run the same workflow on a branch, then validate with repo tests
boundary: no broad SQLAlchemy migration guaranteeFinal report shape
final report:
files scanned: 1
files changed: 1
transforms applied: declarative_imports, query_get, select_list_syntax
confidence: supported subset, validation required before mergeThis is the visible evidence shape buyers should expect: before input, scanner JSON, preview diff, apply output, validation status, and final manager summary. The paid pack still has to be run on a branch and validated in the buyer's repo.
What this proves
- Direct Query.get, select list, declarative import, and simple loader cleanup can be classified on real public input.
- Unsafe execution and query-shape patterns stay visible as manual-review findings.
- The free scanner is the qualification step before any paid cleanup workflow.
What it does not prove
- Full engine.execute(...) automation.
- Full Query API migration to select().
- Package, test, or database integration success for every repo with one supported snippet.