Source: https://frameworkrecipes.dev/flask/request-db-cleanup/

# Keep Flask request connections from leaking on error paths

Flask application-context teardown runs even when a request raises. A request-scoped connection belongs on g and must be removed and closed there. Closing only at the end of the happy-path route misses exceptions and early responses.

Save this requirement in Perch and check it after each agent edit. The rule gives your agent a named check to run, a finding to repair, and the same command to recheck.

## Implement the Flask boundary

Save this illustrative integration module as `database.py`.

```
import sqlite3
from flask import current_app, g

def get_db():
    if "db" not in g:
        g.db = sqlite3.connect(current_app.config["DATABASE"])
        g.db.row_factory = sqlite3.Row
    return g.db

def close_db(error=None):
    connection = g.pop("db", None)
    if connection is not None:
        connection.close()

def init_app(app):
    app.teardown_appcontext(close_db)
```

The order, inventory, or resource-lifetime scenario is adapted from Perch’s public [order-service source at commit 4af67af](https://github.com/lakeday-org/perch/blob/4af67af46d76ac99398d61c7ad0a45441760d4dc/test/fixtures/order-service/storage.py). This Flask implementation is an illustrative adaptation.

## Integrate with your application

Set DATABASE to your application database path and call init_app in the application factory. SQLite suits this focused lifetime test; use the appropriate pooled adapter for a multi-process production database. Do not put a connection or authenticated user in a process-global variable.

## Exercise the boundary and failure paths

| Scenario | Expected outcome |
| --- | --- |
| Two get_db calls in one context | Same connection |
| Request raises after acquisition | Connection closes at teardown |
| No database use | Teardown does nothing |
| New application context | New connection, no previous user state |

Add regression tests for these cases in your existing Flask test project. Use separate connections or deliberately reordered completion where the case involves concurrency; a sequential happy-path test does not exercise that failure. Run the test command after creating the matching test file or suite:

```
python -m pytest tests/test_database_lifecycle.py
```

A focused test to adapt to the paths and test dependencies in your project:

```
from unittest.mock import Mock
from flask import Flask, g
from database import init_app

def test_error_path_closes_connection():
    app = Flask(__name__)
    init_app(app)
    connection = Mock()
    try:
        with app.app_context():
            g.db = connection
            raise RuntimeError("request failed")
    except RuntimeError:
        pass
    connection.close.assert_called_once()
```

## Introduce the specific agent regression

In a throwaway copy, replace this source span:

```
connection.close()
```

with:

```
pass
```

Run the relevant negative test and the Perch command on that broken copy. Restore the corrected source before committing the application. The supplied mutation reproduces the exact regression for the rule check.

## Preserve the requirement as a Perch rule

Save this in `.perch/rules/perch-guide-flask-request-db-cleanup.yaml` at the repository root, keeping your existing `perch.yaml`:

```
- name: request-db-cleanup
  where: "database.py"
  ensure: >-
    close_db must close a database connection removed from g, and init_app must register close_db as teardown_appcontext.
```

Install and authenticate the CLI once, then check the current working file:

```
npm install -g @lakeday/perch
perch login
perch check database.py --rules request-db-cleanup --json
```

Use Node 22 or newer. The framework can use another language; Perch’s CLI remains a Node command. A gated finding exits 3. Exits 1 or 2 mean the check did not complete and must be reported as a blocker.

## Inspect the captured rule check

On 2026-10-14, Perch using jev-latest checked the exact fixed and broken source provided here. The capture records SHA-256 hashes of both sources and the requirement.

| Source control | Exit | Outcome |
| --- | --- | --- |
| Broken | 3 | Rule finding |
| Corrected | 0 | No rule finding |

The rule distinguished these controls in this run. [Raw Perch capture](https://frameworkrecipes.dev/flask/request-db-cleanup/capture.json) · [Exact sources and mutation](https://frameworkrecipes.dev/flask/request-db-cleanup/source.json).

## Install the rule pack and configure your agent

[Download this rule pack ZIP](https://reporules.fyi/flask/request-db-cleanup/rule-pack.zip) · [Download the YAML rules](https://reporules.fyi/flask/request-db-cleanup/rules.yaml) · [Agent task](https://reporules.fyi/flask/request-db-cleanup/agent-task.txt). The pack contains the rule, its example paths and the repair/recheck instructions. Review the `where` selector before using it in an existing project. Keep your root `perch.yaml`; install this rule as a separate file under `.perch/rules/`.

1. Install Perch using Node 22 or newer and sign in once:

```
npm install -g @lakeday/perch
perch login
```

1. Download the pack, inspect it, and extract it from your repository root. The extraction preserves any existing files:

```
curl -fL 'https://frameworkrecipes.dev/flask/request-db-cleanup/rule-pack.zip' -o 'perch-guide-flask-request-db-cleanup.zip'
unzip -n 'perch-guide-flask-request-db-cleanup.zip'
```

The installed rule is `.perch/rules/perch-guide-flask-request-db-cleanup.yaml`. Its selector matches `database.py` in this example; adapt the selector to your application file before checking an existing project.

1. Configure the coding agent you use. Run one of these commands:

Claude Code:

```
perch setup claude-code
```

Codex:

```
perch setup codex
```

Cursor:

```
perch setup cursor
```

Pi:

```
perch setup pi
```

1. Ask the agent to edit the implementation and check the saved requirement:

> Use the installed Perch skill while editing database.py. Preserve request-db-cleanup. Run perch check database.py --rules request-db-cleanup --json after the edit. Read each finding and the affected code, fix the implementation, and run the identical check again after the fix. Keep the rule and regression tests intact. Run the application's relevant regression tests and report the commands and actual results. Exit 3 means a rule finding; exit 1 or 2 means the check did not complete. Report an unavailable or incomplete check as blocked. Do not claim the application tests ran if only the Perch check ran.

1. After repairing a finding, run the same check again and inspect the result:

```
perch check database.py --rules request-db-cleanup --json
```

Commit the downloaded rule and the installed agent configuration with your application. The rule file adds no guarantee of correctness; keep the regression tests from this guide.

With the base reference fetched, use `perch scan --since origin/main --filter type=lint` in CI.

## Framework references and verification scope

- [flask.palletsprojects.com reference](https://flask.palletsprojects.com/en/stable/tutorial/database/)

The dated capture records the source check shown above. Implement the adapters, schema and regression tests described in this guide in your application. Unchanged requests reuse cached answers; fresh model judgments can vary. Model analysis sends relevant source to the configured service.

**Framework Recipes**

Implementation, failure path, check.

Published by Perch · © 2026 Verglas LLC. Framework Recipes is a Perch publication.

Perch’s MIT license covers the CLI; hosted model analysis is a paid service. Relevant source is sent to the configured model service.

[Perch CLI MIT license](https://github.com/lakeday-org/perch/blob/main/LICENSE) · [Article text](https://frameworkrecipes.dev/flask/request-db-cleanup/index.md) · [Resource index](https://frameworkrecipes.dev/llms.txt)