recallfsck 0.2.0rc2__tar.gz

recallfsck-0.2.0rc2/.github/ISSUE_TEMPLATE/adapter.md

@@ -0,0 +1,12 @@
+---
+name: Application or backend adapter
+about: Propose a real integration with observable lifecycle semantics
+---
+
+Application/backend:
+Acknowledgement and consistency guarantees:
+Disposable namespace strategy:
+Retrieval paths:
+Controlled pause/resume hooks:
+Cleanup behavior:
+Reproducible synthetic failure:

recallfsck-0.2.0rc2/.github/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,15 @@
+---
+name: Integrity failure or tool bug
+about: Share reproducible evidence using synthetic data
+---
+
+Backend and version:
+Python and RecallFSCK version:
+Application adapter:
+Command and timing settings (remove credentials):
+Expected state:
+Observed state:
+Synthetic reproduction JSON:
+Does replay reproduce it?
+
+Do not include private source documents, credentials or security exploit details.

recallfsck-0.2.0rc2/.github/ISSUE_TEMPLATE/feedback.yml

@@ -0,0 +1,34 @@
+name: Independent installation feedback
+description: Record your own attempt to install and run RecallFSCK
+title: 'Installation feedback: '
+body:
+  - type: input
+    id: platform
+    attributes:
+      label: Operating system and Python version
+    validations:
+      required: true
+  - type: textarea
+    id: steps
+    attributes:
+      label: Commands you actually ran
+      description: Remove credentials and private documents.
+    validations:
+      required: true
+  - type: dropdown
+    id: outcome
+    attributes:
+      label: Outcome
+      options:
+        - Demo reproduced the expected failures and passes
+        - Installation failed
+        - Demo produced unexpected results
+    validations:
+      required: true
+  - type: textarea
+    id: evidence
+    attributes:
+      label: Observed results and confusing instructions
+      description: Include synthetic evidence or error output, not guesses.
+    validations:
+      required: true

recallfsck-0.2.0rc2/.github/workflows/publish.yml

@@ -0,0 +1,57 @@
+name: Publish Python package
+on:
+  workflow_dispatch:
+    inputs:
+      destination:
+        description: Package index
+        required: true
+        default: testpypi
+        type: choice
+        options: [testpypi, pypi]
+permissions:
+  contents: read
+jobs:
+  checks:
+    uses: ./.github/workflows/tests.yml
+  build:
+    needs: checks
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+      - name: Verify tagged package version
+        run: |
+          python - <<'PY'
+          import os, tomllib
+          version = tomllib.load(open('pyproject.toml', 'rb'))['project']['version']
+          if os.environ['GITHUB_REF'] != 'refs/tags/v' + version:
+              raise SystemExit('Dispatch this workflow from the matching release tag')
+          PY
+      - run: python -m pip install build twine
+      - run: python -m build
+      - run: python -m twine check dist/*
+      - uses: actions/upload-artifact@v4
+        with:
+          name: publish-distributions
+          path: dist/
+  publish:
+    needs: build
+    runs-on: ubuntu-latest
+    environment: ${{ inputs.destination }}
+    permissions:
+      id-token: write
+    steps:
+      - uses: actions/download-artifact@v4
+        with:
+          name: publish-distributions
+          path: dist/
+      - name: Upload to TestPyPI
+        if: inputs.destination == 'testpypi'
+        uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          repository-url: https://test.pypi.org/legacy/
+      - name: Upload to PyPI
+        if: inputs.destination == 'pypi'
+        uses: pypa/gh-action-pypi-publish@release/v1

recallfsck-0.2.0rc2/.github/workflows/release.yml

@@ -0,0 +1,21 @@
+name: Build release candidate
+on:
+  workflow_dispatch:
+permissions:
+  contents: read
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+      - run: python -m pip install -e '.[test]'
+      - run: python -m pytest -q
+      - run: python -m build
+      - run: python -m twine check dist/*
+      - uses: actions/upload-artifact@v4
+        with:
+          name: recallfsck-release-candidate
+          path: dist/

recallfsck-0.2.0rc2/.github/workflows/tests.yml

@@ -0,0 +1,78 @@
+name: tests
+on:
+  push:
+  pull_request:
+  workflow_call:
+permissions:
+  contents: read
+jobs:
+  unit:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python: ['3.11', '3.12', '3.13']
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '${{ matrix.python }}'
+      - run: python -m pip install -e '.[test]'
+      - run: ruff check .
+      - run: ruff format --check recallfsck tests examples
+      - run: python -m pytest -q --junitxml=report/tests.xml
+      - run: recallfsck demo --output report/sqlite
+      - run: recallfsck demo --app recallfsck.demo_http:create_application --output report/http
+      - uses: actions/upload-artifact@v4
+        if: always()
+        with:
+          name: unit-evidence-${{ matrix.python }}
+          path: report/
+  databases:
+    runs-on: ubuntu-latest
+    services:
+      qdrant:
+        image: qdrant/qdrant:v1.13.4
+        ports: ['6333:6333']
+      postgres:
+        image: pgvector/pgvector:pg16
+        env:
+          POSTGRES_PASSWORD: recallfsck-local-only
+          POSTGRES_DB: recallfsck
+        ports: ['5432:5432']
+        options: >-
+          --health-cmd "pg_isready -U postgres -d recallfsck"
+          --health-interval 2s --health-timeout 2s --health-retries 30
+    env:
+      RECALLFSCK_QDRANT_URL: http://localhost:6333
+      RECALLFSCK_PGVECTOR_URL: postgresql://postgres:recallfsck-local-only@localhost:5432/recallfsck
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+      - run: python -m pip install -e '.[all,test]'
+      - name: Prepare disposable pgvector database and await Qdrant
+        run: |
+          python - <<'PY'
+          import os, time, urllib.request
+          import psycopg
+          with psycopg.connect(os.environ['RECALLFSCK_PGVECTOR_URL'], autocommit=True) as conn:
+              conn.execute('CREATE EXTENSION IF NOT EXISTS vector')
+          for attempt in range(30):
+              try:
+                  urllib.request.urlopen(os.environ['RECALLFSCK_QDRANT_URL'] + '/readyz', timeout=2).close()
+                  break
+              except OSError:
+                  if attempt == 29:
+                      raise
+                  time.sleep(1)
+          PY
+      - run: python -m pytest -q --junitxml=report/tests.xml
+      - run: recallfsck demo --backend qdrant --output report/qdrant
+      - run: recallfsck demo --backend chroma --output report/chroma
+      - run: recallfsck demo --backend pgvector --url "$RECALLFSCK_PGVECTOR_URL" --output report/pgvector
+      - uses: actions/upload-artifact@v4
+        if: always()
+        with:
+          name: database-evidence
+          path: report/

recallfsck-0.2.0rc2/CHANGELOG.md

@@ -0,0 +1,21 @@
+# Changelog
+
+## 0.2.0rc2
+
+- Fix hosted CI to use the installed HTTP application module.
+- Verify Python 3.11, 3.12 and 3.13 plus actual database integrations on GitHub Actions.
+- Prepare the first PyPI upload through the registered trusted publisher.
+
+## 0.2.0rc1
+
+- Add Qdrant, Chroma and pgvector document/vector store adapters.
+- Add ten lifecycle scenarios with full-record and filtered retrieval assertions.
+- Add finite convergence deadlines and post-convergence monitoring.
+- Add YAML init/configuration, schema-2 replay and failure minimization.
+- Add Hypothesis search/shrinking, pytest integration and HTTP application demo.
+- Add JSON, HTML and JUnit evidence, service-error reporting and release tooling.
+- Retain schema-1 replay compatibility.
+
+## 0.1.0
+
+- Initial deterministic create/delete/resurrection red/green experiment.

recallfsck-0.2.0rc2/CODE_OF_CONDUCT.md

@@ -0,0 +1,11 @@
+# Code of conduct
+
+Treat contributors with respect. Technical disagreement is welcome; harassment,
+threats, discrimination, doxxing and repeated unwanted contact are not.
+Keep discussion focused on evidence and the project. Respect personal boundaries
+and avoid sharing private information or credentials.
+
+Maintainers may edit or remove harmful content and restrict participation, using
+proportionate action and considering context. Report concerns privately to the
+repository owner. A public contact address must be established before launch;
+do not post sensitive reports publicly while that channel is being set up.

recallfsck-0.2.0rc2/CONTRIBUTING.md

@@ -0,0 +1,20 @@
+# Contributing
+
+Use Python 3.11+ in a virtual environment. Install with
+`python -m pip install -e '.[all,test]'`, then run `ruff check .` and
+`python -m pytest -q`. Start the disposable services with
+`docker compose up -d --wait` for integration coverage:
+
+```sh
+RECALLFSCK_QDRANT_URL=http://127.0.0.1:16333 \
+RECALLFSCK_PGVECTOR_URL=postgresql://postgres:recallfsck-local-only@127.0.0.1:15432/recallfsck \
+  python -m pytest -q
+```
+
+Run affected adapter demos and include replayable synthetic failure evidence.
+Keep changes scoped, preserve old artifact replay, and document the exact scope
+of every invariant. Add a deliberately faulty fixture when adding an invariant;
+checking only a passing implementation cannot show the checker detects a defect.
+Do not expose private documents, tokens or connection credentials in reports.
+
+See CODE_OF_CONDUCT.md, SECURITY.md and docs/adapters.md before contributing.

recallfsck-0.2.0rc2/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 RecallFSCK contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

recallfsck-0.2.0rc2/MANIFEST.in

@@ -0,0 +1,7 @@
+include README.md LICENSE CHANGELOG.md CONTRIBUTING.md CODE_OF_CONDUCT.md SECURITY.md VALIDATION.md compose.yaml
+include recallfsck.yaml
+recursive-include docs *.md
+recursive-include examples *.py *.sql
+recursive-include examples *.json
+recursive-include tests *.py
+recursive-include .github *.yml *.md

recallfsck-0.2.0rc2/PKG-INFO

@@ -0,0 +1,178 @@
+Metadata-Version: 2.4
+Name: recallfsck
+Version: 0.2.0rc2
+Summary: Deterministic lifecycle integrity experiments for AI knowledge indexes
+License-Expression: MIT
+Project-URL: Repository, https://github.com/sanketsinghai/recallfsck
+Project-URL: Issues, https://github.com/sanketsinghai/recallfsck/issues
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: PyYAML<7,>=6
+Provides-Extra: chroma
+Requires-Dist: chromadb<2,>=1; extra == "chroma"
+Provides-Extra: langchain
+Requires-Dist: langchain-chroma<2,>=1; extra == "langchain"
+Provides-Extra: pgvector
+Requires-Dist: psycopg[binary]<4,>=3.2; extra == "pgvector"
+Provides-Extra: test
+Requires-Dist: pytest<10,>=8; extra == "test"
+Requires-Dist: hypothesis<7,>=6; extra == "test"
+Requires-Dist: build<2,>=1; extra == "test"
+Requires-Dist: twine<7,>=6; extra == "test"
+Requires-Dist: ruff<1,>=0.11; extra == "test"
+Provides-Extra: all
+Requires-Dist: chromadb<2,>=1; extra == "all"
+Requires-Dist: langchain-chroma<2,>=1; extra == "all"
+Requires-Dist: psycopg[binary]<4,>=3.2; extra == "all"
+Requires-Dist: pytest<10,>=8; extra == "all"
+Requires-Dist: hypothesis<7,>=6; extra == "all"
+Dynamic: license-file
+
+# RecallFSCK
+
+A delete returns success. An older ingestion worker finishes afterward and puts
+that document back into search. RecallFSCK runs that schedule, compares actual
+retrieval with an independent reference model, and saves a replayable failure.
+
+## One-command demo
+
+From this repository, Python 3.11+:
+
+```sh
+python -m pip install -e '.[test]'
+recallfsck demo
+```
+
+The demo runs ten scenarios against two applications. The deliberately buggy
+worker fails three races: delete/index, replacement/index, and migration/index.
+The corrected worker passes all ten. Demo exits 0 only when that exact pattern is
+observed. The evidence still labels the buggy cases FAIL.
+
+Open `report/index.html`; each case also has JSON evidence. `report/junit.xml`
+records failures for CI. These are executed results, not illustrative screenshots.
+
+```sh
+recallfsck init --adapter sqlite
+recallfsck test --config recallfsck.yaml
+recallfsck replay report/buggy-delete-during-index.json
+recallfsck minimize report/buggy-delete-during-index.json --output report/minimal
+recallfsck explore --worker buggy --examples 100 --output report/discovery
+```
+
+Replay, minimize and explore exit 1 when they reproduce an integrity failure.
+Exit 2 means an input, service, dependency or execution error. A missing database
+never becomes PASS. `explore` uses Hypothesis to search actual executions and
+shrink failures; `minimize` produces a 1-minimal valid deletion subsequence.
+
+## What is checked
+
+| Invariant | Observed evidence |
+| --- | --- |
+| Create | Acknowledged source appears in application queries and index scan |
+| Replacement | Only the expected revision/content remains |
+| Metadata | Stored and retrieved revision matches expected metadata |
+| Delete | Document remains absent throughout the configured monitor window |
+| Idempotency | Repeat upsert/delete does not introduce duplicate records |
+| Isolation | Alpha, beta, and outsider queries expose only authorized records |
+| Concurrency | Paused stale workers cannot overwrite deletion, revision or model |
+| Model migration | All live records move to the synthetic v2 vector configuration and remain retrievable |
+
+The generation check and final commit share the deletion lock in the corrected
+fixture. The scheduler coordinates with events rather than hoping sleeps cause a
+race. Passing results cover finite observation windows, not indefinite absence.
+
+## Integrations
+
+| Adapter | Installation | Runtime |
+| --- | --- | --- |
+| SQLite | Base package | Disposable in-memory SQL; no vector-search claim |
+| Qdrant | Base package | Disposable Qdrant REST collection |
+| Chroma | `pip install -e '.[chroma]'` | Embedded ephemeral Chroma vector collection |
+| pgvector | `pip install -e '.[pgvector]'` | Disposable table on a PostgreSQL server with vector enabled |
+
+```sh
+# Start isolated databases (Docker Compose):
+docker compose up -d --wait
+recallfsck demo --backend qdrant --url http://127.0.0.1:16333 --output report/qdrant
+recallfsck demo --backend chroma --output report/chroma
+RECALLFSCK_URL=postgresql://postgres:recallfsck-local-only@127.0.0.1:15432/recallfsck \
+  recallfsck demo --backend pgvector --output report/pgvector
+docker compose down -v
+```
+
+For source integration, implement the [application adapter contract](docs/adapters.md).
+A runnable [HTTP application example](examples/http_application.py) goes through
+real local HTTP calls while using the runner's disposable index:
+
+```sh
+recallfsck demo --app recallfsck.demo_http:create_application --output report/http
+```
+
+A real LangChain/Chroma integration is available below. Dify and other applications
+need their own adapters. Backend conformance and bundled fixtures do not prove an
+unrelated production application's correctness.
+
+## pytest
+
+```python
+def test_delete_integrity(recallfsck_runner):
+    result = recallfsck_runner("delete-during-index", worker="fixed")
+    assert result["status"] == "PASS", result["first_divergence"]
+```
+
+Install the `test` extra for pytest and Hypothesis. Importing the plugin does not
+create a database or automatically add tests to your project.
+
+## Development and release
+
+```sh
+python -m pip install -e '.[all,test]'
+python -m pytest -q
+ruff check .
+python -m build
+python -m twine check dist/*
+```
+
+See [usage](docs/usage.md), [architecture](docs/architecture.md),
+[roadmap status](docs/roadmap.md), [validation](VALIDATION.md), and
+[release instructions](docs/releasing.md).
+
+## Limits and safety
+
+Run mutating tests only on disposable services and synthetic data. Each adapter
+creates and cleans up its own randomly named collection/table. Interrupted runs
+may leave owned objects behind; existing namespaces cannot be selected.
+
+The vector tests use known two-dimensional synthetic embeddings. Migration checks
+cover same-dimensional configuration changes, not a real model's semantic quality
+or cross-dimension rebuild. Queries are top-100 with full index scans as a separate
+check; generated schedules are capped at 200 operations and the bundled corpus
+fits below the query limit. Unknown applications need explicit integration and
+scheduling hooks. Network calls have adapter timeouts; consistency settings are
+not a hard wall-clock cancellation limit for blocking drivers.
+
+Logical source deletion, retrieval absence, and physical erasure are distinct.
+This tool does not prove disk/backup erasure, untested query paths, distributed
+linearizability, production authorization policy or arbitrary future absence.
+There is no production scanning mode. Saved reports contain test document text;
+use synthetic inputs. No hosted service or API keys are required for the demos.
+
+Repository: https://github.com/sanketsinghai/recallfsck . The PyPI name returned
+HTTP 404 when checked on October 26, 2026; this does not reserve the name or
+establish trademark clearance.
+MIT license; optional dependencies keep their respective licenses.
+
+## Real LangChain integration
+
+```sh
+python -m pip install -e '.[langchain,test]'
+recallfsck demo --backend chroma \
+  --app recallfsck.integrations.langchain_chroma:create_application \
+  --output report/langchain
+```
+
+In this integration, writes, deletion, embedding and retrieval call the actual
+LangChain Chroma vector-store APIs. The controlled ingestion scheduling and the
+injected bugs belong to RecallFSCK's demonstration application. No upstream
+LangChain defect or production application correctness is claimed.

recallfsck-0.2.0rc2/README.md

@@ -0,0 +1,147 @@
+# RecallFSCK
+
+A delete returns success. An older ingestion worker finishes afterward and puts
+that document back into search. RecallFSCK runs that schedule, compares actual
+retrieval with an independent reference model, and saves a replayable failure.
+
+## One-command demo
+
+From this repository, Python 3.11+:
+
+```sh
+python -m pip install -e '.[test]'
+recallfsck demo
+```
+
+The demo runs ten scenarios against two applications. The deliberately buggy
+worker fails three races: delete/index, replacement/index, and migration/index.
+The corrected worker passes all ten. Demo exits 0 only when that exact pattern is
+observed. The evidence still labels the buggy cases FAIL.
+
+Open `report/index.html`; each case also has JSON evidence. `report/junit.xml`
+records failures for CI. These are executed results, not illustrative screenshots.
+
+```sh
+recallfsck init --adapter sqlite
+recallfsck test --config recallfsck.yaml
+recallfsck replay report/buggy-delete-during-index.json
+recallfsck minimize report/buggy-delete-during-index.json --output report/minimal
+recallfsck explore --worker buggy --examples 100 --output report/discovery
+```
+
+Replay, minimize and explore exit 1 when they reproduce an integrity failure.
+Exit 2 means an input, service, dependency or execution error. A missing database
+never becomes PASS. `explore` uses Hypothesis to search actual executions and
+shrink failures; `minimize` produces a 1-minimal valid deletion subsequence.
+
+## What is checked
+
+| Invariant | Observed evidence |
+| --- | --- |
+| Create | Acknowledged source appears in application queries and index scan |
+| Replacement | Only the expected revision/content remains |
+| Metadata | Stored and retrieved revision matches expected metadata |
+| Delete | Document remains absent throughout the configured monitor window |
+| Idempotency | Repeat upsert/delete does not introduce duplicate records |
+| Isolation | Alpha, beta, and outsider queries expose only authorized records |
+| Concurrency | Paused stale workers cannot overwrite deletion, revision or model |
+| Model migration | All live records move to the synthetic v2 vector configuration and remain retrievable |
+
+The generation check and final commit share the deletion lock in the corrected
+fixture. The scheduler coordinates with events rather than hoping sleeps cause a
+race. Passing results cover finite observation windows, not indefinite absence.
+
+## Integrations
+
+| Adapter | Installation | Runtime |
+| --- | --- | --- |
+| SQLite | Base package | Disposable in-memory SQL; no vector-search claim |
+| Qdrant | Base package | Disposable Qdrant REST collection |
+| Chroma | `pip install -e '.[chroma]'` | Embedded ephemeral Chroma vector collection |
+| pgvector | `pip install -e '.[pgvector]'` | Disposable table on a PostgreSQL server with vector enabled |
+
+```sh
+# Start isolated databases (Docker Compose):
+docker compose up -d --wait
+recallfsck demo --backend qdrant --url http://127.0.0.1:16333 --output report/qdrant
+recallfsck demo --backend chroma --output report/chroma
+RECALLFSCK_URL=postgresql://postgres:recallfsck-local-only@127.0.0.1:15432/recallfsck \
+  recallfsck demo --backend pgvector --output report/pgvector
+docker compose down -v
+```
+
+For source integration, implement the [application adapter contract](docs/adapters.md).
+A runnable [HTTP application example](examples/http_application.py) goes through
+real local HTTP calls while using the runner's disposable index:
+
+```sh
+recallfsck demo --app recallfsck.demo_http:create_application --output report/http
+```
+
+A real LangChain/Chroma integration is available below. Dify and other applications
+need their own adapters. Backend conformance and bundled fixtures do not prove an
+unrelated production application's correctness.
+
+## pytest
+
+```python
+def test_delete_integrity(recallfsck_runner):
+    result = recallfsck_runner("delete-during-index", worker="fixed")
+    assert result["status"] == "PASS", result["first_divergence"]
+```
+
+Install the `test` extra for pytest and Hypothesis. Importing the plugin does not
+create a database or automatically add tests to your project.
+
+## Development and release
+
+```sh
+python -m pip install -e '.[all,test]'
+python -m pytest -q
+ruff check .
+python -m build
+python -m twine check dist/*
+```
+
+See [usage](docs/usage.md), [architecture](docs/architecture.md),
+[roadmap status](docs/roadmap.md), [validation](VALIDATION.md), and
+[release instructions](docs/releasing.md).
+
+## Limits and safety
+
+Run mutating tests only on disposable services and synthetic data. Each adapter
+creates and cleans up its own randomly named collection/table. Interrupted runs
+may leave owned objects behind; existing namespaces cannot be selected.
+
+The vector tests use known two-dimensional synthetic embeddings. Migration checks
+cover same-dimensional configuration changes, not a real model's semantic quality
+or cross-dimension rebuild. Queries are top-100 with full index scans as a separate
+check; generated schedules are capped at 200 operations and the bundled corpus
+fits below the query limit. Unknown applications need explicit integration and
+scheduling hooks. Network calls have adapter timeouts; consistency settings are
+not a hard wall-clock cancellation limit for blocking drivers.
+
+Logical source deletion, retrieval absence, and physical erasure are distinct.
+This tool does not prove disk/backup erasure, untested query paths, distributed
+linearizability, production authorization policy or arbitrary future absence.
+There is no production scanning mode. Saved reports contain test document text;
+use synthetic inputs. No hosted service or API keys are required for the demos.
+
+Repository: https://github.com/sanketsinghai/recallfsck . The PyPI name returned
+HTTP 404 when checked on October 26, 2026; this does not reserve the name or
+establish trademark clearance.
+MIT license; optional dependencies keep their respective licenses.
+
+## Real LangChain integration
+
+```sh
+python -m pip install -e '.[langchain,test]'
+recallfsck demo --backend chroma \
+  --app recallfsck.integrations.langchain_chroma:create_application \
+  --output report/langchain
+```
+
+In this integration, writes, deletion, embedding and retrieval call the actual
+LangChain Chroma vector-store APIs. The controlled ingestion scheduling and the
+injected bugs belong to RecallFSCK's demonstration application. No upstream
+LangChain defect or production application correctness is claimed.

recallfsck-0.2.0rc2/SECURITY.md

@@ -0,0 +1,15 @@
+# Security
+
+Report vulnerabilities privately through
+https://github.com/sanketsinghai/recallfsck/security/advisories/new .
+Private vulnerability reporting is enabled. Do not disclose credentials, customer
+source documents or exploit details in public issues.
+
+This release candidate supports disposable synthetic tests. Never run mutating
+experiments against production servers or real customer data. Reports contain
+synthetic document text and operation histories. Application adapters must avoid
+persisting credentials and private source content in evidence.
+
+The latest release candidate is the supported version during initial development.
+Security reports are reviewed by the repository owner; no response-time service
+level is promised.