Skip to content

CLAUDE.md

Latest commit

 

History

History
167 lines (109 loc) · 7.55 KB

CLAUDE.md

File metadata and controls

167 lines (109 loc) · 7.55 KB
Raw

SONAR Claude guide

Overview

SONAR (Swiss Open Access Repository) is the Python/Flask backend for an archive of scholarly publications from Swiss public research institutions. Some frontend elements live in this project as HTML/Jinja templates; the rest is a separate Angular project (sonar-ui) based on ng-core.

Stack: Python 3.14, Flask (Invenio), PostgreSQL, Elasticsearch 7, Celery, RabbitMQ, Redis Package manager: uv with poethepoet for task running

Commands

During development, all commands are run through uv's virtual env with uv run.

Linting and formatting

IMPORTANT: After editing files, make sure that there are no errors in the formatting and linting.

uv run poe lint     # ruff check sonar tests
uv run poe format   # ruff format .

Setup (done by humans)

Human developers will run the required containers, the app setup and the servers on their own terms.

Architecture

Module Structure

Most business logic lives in sonar/modules/. Each module follows a consistent pattern:

sonar/modules/<module_name>/
├── api.py            # Record class + Search class + Indexer (core business logic)
├── models.py         # SQLAlchemy model + Identifier + Metadata
├── views.py          # Flask blueprint (UI routes)
├── rest.py           # Flask blueprint (REST API routes)
├── tasks.py          # Celery async tasks
├── receivers.py      # Signal handlers (enrich data before indexing, file events)
├── permissions.py    # Access control rules
├── minters.py        # PID minting
├── jsonschemas/      # JSON Schema for validation
├── mappings/v7/      # Elasticsearch index mappings
├── serializers/      # REST response serializers
├── dumpers.py        # Data dumpers for ES indexing
├── jsonresolvers.py  # JSON $ref resolver
└── marshmallow/      # Marshmallow schemas (loaders/dumpers for REST)

A newer sonar/resources/ tree hosts modules built on invenio-records-resources (e.g. projects) with the service/resource pattern; new resources should generally follow that style. Organisation-specific customizations live under sonar/dedicated/ (e.g. hepvs).

Base Classes

  • SonarRecord (sonar/modules/api.py): extends invenio_records_files.api.Record with a FilesMixin. All domain records (Document, Deposit, Organisation, User, Collection, etc.) inherit from this. Provides PID management, ref-link helpers, file handling, and reindexing.
  • SonarSearch (sonar/modules/api.py): extends invenio_search.api.RecordsSearch. Each module defines its own search class with a specific ES index.
  • SonarIndexer (sonar/modules/api.py): extends invenio_indexer.api.RecordIndexer and flushes the ES index after each operation.

Signal/Event Flow

The sonar/ext.py file wires up signal listeners. Before a record is indexed in Elasticsearch, receivers.py in each module (and dumpers) can enrich the data (e.g., adding computed fields, resolving references). This is the primary mechanism for denormalizing data into ES. File upload/delete signals are also bridged through sonar/modules/receivers.py.

API Entry Points

REST endpoints are registered in pyproject.toml under [project.entry-points."invenio_base.api_blueprints"]. Each module's rest.py exports an api_blueprint. UI blueprints are registered under [project.entry-points."invenio_base.blueprints"] and exported from views.py.

Permissions

Each module has a permissions.py using invenio-records-permissions. Access is typically scoped by organisation membership (multi-tenancy); some records are further scoped by subdivision.

Code Style

  • Be clear and concise in the docstrings and do not over-comment the code.
  • Do not use Python type annotations (no -> str, : str, etc. in signatures).
  • Ruff is configured with line-length = 120 and pep257 docstring convention; see [tool.ruff] in pyproject.toml for the enabled rule sets.
  • Commit messages follow Conventional Commits

Translations

Translations are only added manually before a release. During standard development, only make sure that any strings that must be displayed to the end-user are marked for translations in the code, but do not run the extractor or edit any files in sonar/translations.

Testing Notes

  • Tests use function-based style (no class-based tests).
  • Tests are split into tests/api/, tests/ui/, tests/unit/
  • The project follows a test-driven development methodology. Each commit must be accompanied by tests that ensure that the functionality works as intended. Tests must follow DRY principles and should only test specific app behaviour and not the behaviour of external modules (e.g. invenio dependencies).
  • Test fixtures (shared data) are in tests/conftest.py and the per-layer conftest.py files (tests/api/conftest.py, tests/ui/conftest.py).
  • Sample data is in tests/data/ and data/
  • pytest is configured in pyproject.toml ([tool.pytest.ini_options]) with --ruff, --doctest-modules and coverage on sonar.

Running the tests (done by humans)

Human developers will run the needed tests from their consoles because they need to make sure the tests run only when their testing container runs.

Behavioral guidelines

Tradeoff: These guidelines bias toward caution over speed. For trivial tasks, use judgment.

1. Think Before Coding

Don't assume. Don't hide confusion. Surface tradeoffs.

Before implementing:

  • State your assumptions explicitly. If uncertain, ask.
  • If multiple interpretations exist, present them - don't pick silently.
  • If a simpler approach exists, say so. Push back when warranted.
  • If something is unclear, stop. Name what's confusing. Ask.

2. Simplicity First

Minimum code that solves the problem. Nothing speculative.

  • No features beyond what was asked.
  • No abstractions for single-use code.
  • No "flexibility" or "configurability" that wasn't requested.
  • No error handling for impossible scenarios.
  • If you write 200 lines and it could be 50, rewrite it.

Ask yourself: "Would a senior engineer say this is overcomplicated?" If yes, simplify.

3. Surgical Changes

Touch only what you must. Clean up only your own mess.

When editing existing code:

  • Don't "improve" adjacent code, comments, or formatting.
  • Don't refactor things that aren't broken.
  • Match existing style, even if you'd do it differently.
  • If you notice unrelated dead code, mention it - don't delete it.

When your changes create orphans:

  • Remove imports/variables/functions that YOUR changes made unused.
  • Don't remove pre-existing dead code unless asked.

The test: Every changed line should trace directly to the user's request.

4. Goal-Driven Execution

Define success criteria. Loop until verified.

Transform tasks into verifiable goals:

  • "Add validation" → "Write tests for invalid inputs, then make them pass"
  • "Fix the bug" → "Write a test that reproduces it, then make it pass"
  • "Refactor X" → "Ensure tests pass before and after"

For multi-step tasks, state a brief plan:

1. [Step] → verify: [check]
2. [Step] → verify: [check]
3. [Step] → verify: [check]

Strong success criteria let you loop independently. Weak criteria ("make it work") require constant clarification.

These guidelines are working if: fewer unnecessary changes in diffs, fewer rewrites due to overcomplication, and clarifying questions come before implementation rather than after mistakes.