docs: add CLAUDE.md file for repository guidance [ignore]

CLAUDE.md

@@ -0,0 +1,133 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## Build and development commands
+
+```bash
+# Install all dependencies (run from repo root)
+make install-dev
+
+# Linting and type checking
+make lint              # Run ruff linter
+make type-check        # Run mypy
+make check-code        # Run both lint and type-check
+make format            # Auto-fix lint issues and format code
+
+# Testing
+make test-unit                    # Run unit tests only
+make test-integration             # Run integration tests (requires databases via docker-compose)
+make test                         # Run all tests
+
+# Run a single test
+poetry run -C code pytest code/tests/test_utils.py::test_name -v
+
+# Generate pydantic models from input schemas
+make pydantic-model
+
+# Start local databases for integration testing
+docker compose up -d
+```
+
+## Running actors locally
+
+```bash
+export ACTOR_PATH_IN_DOCKER_CONTEXT=actors/pinecone  # or chroma, qdrant, etc.
+apify run -p
+```
+
+## Git workflow and commit conventions
+
+### Branching strategy
+
+- `master` - Production branch, all PRs target this branch
+- Feature branches should be created from `master`
+
+### Commit messages and PR titles
+
+All commits and PR titles must follow **[Conventional Commits](https://www.conventionalcommits.org/)** format:
+
+```
+<type>(<scope>): <description>
+```
+
+**Required elements:**
+- **type**: `feat`, `fix`, `chore`, `refactor`, `docs`, `test`, etc.
+- **scope**: Component or area affected (e.g., `chroma`, `pinecone`, `tests`, `ci`)
+- **description**: Brief summary in imperative mood
+
+**Breaking changes:** Append `!` after scope (e.g., `feat(api)!: ...`)
+
+**Changelog classification** (append to end):
+- _(none)_: user-facing change (default)
+- `[admin]`: admin-only change
+- `[internal]`: internal change (migrations, refactoring)
+- `[ignore]`: non-important change (dependency updates, DX improvements)
+
+**Examples:**
+```
+feat(chroma): add remote database support
+fix(pinecone): handle rate limit errors
+refactor(tests): extract common fixtures [internal]
+chore(deps): update langchain version [ignore]
+```
+
+### Naming conventions
+
+- **Functions & Variables**: `camelCase`
+- **Classes, Types, Components**: `PascalCase`
+- **Files & Folders**: `snake_case`
+- **Constants** (module-level, immutable): `UPPER_SNAKE_CASE`
+- **Booleans**: Prefix with `is`, `has`, or `should` (e.g., `isValid`, `hasFinished`)
+- **Units**: Suffix with unit (e.g., `timeoutSeconds`, `maxRetries`)
+- **Date/Time**: Suffix with `At` (e.g., `lastSeenAt`, `createdAt`)
+
+## Architecture overview
+
+This is a monorepo containing multiple Apify Actors for vector database integrations. All Actors share a common codebase in `code/src/` with database-specific implementations.
+
+### Directory structure
+
+- `actors/` - Individual Actor definitions (one per database: chroma, milvus, opensearch, pgvector, pinecone, qdrant, weaviate)
+  - Each contains `.actor/actor.json` (Actor definition) and `.actor/input_schema.json` (input schema)
+- `code/src/` - Shared source code for all Actors
+- `code/tests/` - Test suite
+
+### Core flow
+
+1. `entrypoint.py` - Entry point, determines which database Actor to run based on `ACTOR_PATH_IN_DOCKER_CONTEXT`
+2. `main.py:run_actor()` - Main orchestration: load dataset → compute embeddings → chunk text → update vector store
+3. `vcs.py` - Vector store operations: delta updates, upserts, expired object deletion
+
+### Key components
+
+**Input Models** (`code/src/models/`): Auto-generated pydantic models from `actors/*/.actor/input_schema.json`. After modifying an input schema, run `make pydantic-model` to regenerate.
+
+**Vector Store Implementations** (`code/src/vector_stores/`): Each database has its own module implementing `VectorDbBase` from `base.py`. Required methods: `get_by_item_id`, `update_last_seen_at`, `delete_by_item_id`, `delete_expired`, `delete_all`, `is_connected`, `search_by_vector`.
+
+**Type Definitions** (`code/src/_types.py`): `ActorInputsDb` (union of all input models) and `VectorDb` (union of all database classes).
+
+**Embeddings** (`code/src/emb.py`): Supports OpenAI and Cohere embedding providers.
+
+### Data update strategies
+
+- `deltaUpdates` - Only update changed data (compares checksums)
+- `add` - Add all documents without checking existing
+- `upsert` - Delete by item_id then add all documents
+
+### Adding a new database integration
+
+See README.md for step-by-step instructions. Key steps:
+1. Add database to `docker-compose.yaml`
+2. Add poetry group with dependencies in `code/pyproject.toml`
+3. Create Actor in `actors/<name>/` with `.actor/actor.json` and `input_schema.json`
+4. Generate pydantic model with `make pydantic-model`
+5. Implement database class in `code/src/vector_stores/<name>.py` extending `VectorDbBase`
+6. Register in `constants.py`, `entrypoint.py`, `_types.py`, and `vcs.py`
+7. Add test fixture in `code/tests/conftest.py` and add to `DATABASE_FIXTURES` list
+
+### Testing pattern
+
+Integration tests use pytest fixtures for each database (e.g., `db_chroma`, `db_pinecone`). Tests are parameterized over `DATABASE_FIXTURES` list. Some databases (Pinecone, OpenSearch) have eventual consistency delays handled via `unit_test_wait_for_index`.
+
+Environment variables for tests are loaded from `.env` file (e.g., `OPENAI_API_KEY`, database connection strings).