fix: address staff engineer feedback

[zvsdev]

Summary

• Consolidate duplicate _get_prefix logic between uplid.py and sqlalchemy.py
• Add SQLAlchemy string validation on write (catches prefix mismatches early)
• Add concurrency tests to validate thread-safety claim
• Soften "compile-time" language in docs to accurately reflect Python's type checking
• Redesign README with ASCII logo and Stripe tagline
• DRY up test files with shared conftest fixtures
• Require 100% test coverage in CI
• Performance optimization: skip redundant prefix validation in factory()

Changes

1. Consolidate _get_prefix

Export _get_prefix from the uplid module. sqlalchemy._extract_prefix now wraps it, converting UPLIDError to TypeError for SQLAlchemy context.

2. SQLAlchemy write validation

process_bind_param now validates:

• UPLID objects have the correct prefix
• Strings are valid UPLIDs with the correct prefix

This catches errors at write time rather than read time.

3. Concurrency tests

• test_concurrent_generation_produces_unique_ids - 10 threads × 1000 IDs
• test_concurrent_generation_with_different_prefixes - 6 threads with different prefixes
• test_concurrent_generation_maintains_ordering - ordering within threads
• test_no_duplicate_uuids_under_contention - 50 threads with barrier sync

4. Documentation accuracy

• "at compile time" → "caught by type checkers like mypy/pyright/ty"
• "compile-time errors" → "Type error"
• Fix misleading "zero dependencies" claim (Pydantic is a dependency)

5. README redesign

• ASCII block art UPLID logo
• "Stripe-style IDs for Python" tagline
• Before/after comparison at top
• Benefits-focused "Why UPLID?" section
• Condensed integration examples

6. Test DRY refactoring

• Move common type aliases (UserId, OrgId, ApiKeyId) to conftest.py
• Move common factories to conftest.py
• Add shared Hypothesis strategies
• Add pytest fixtures for SQLAlchemy/SQLModel engines

7. Performance optimization

factory() now validates the prefix once at creation time, then uses an internal _generate_unchecked() method. This removes ~15% overhead from ID generation when using factories.

Test plan

• All 162 tests pass
• 100% coverage (required)
• Type checker passes
• Ruff passes

🤖 Generated with Claude Code