Consolidate all test and development configuration into a single source
of truth (.env.dev) for cleaner, more maintainable setup.

Changes:
- Remove obsolete .env.test and .env.test.example files
- Update Makefile to load .env.dev variables automatically via include
- Simplify test commands (no more bash subshells or complex sourcing)
- Update integration tests to read config from environment variables
- Rewrite TESTING_SUMMARY.md with current unified approach
- Update LOCAL_DEVELOPMENT.md to reference single config file

Benefits:
- Single source of truth for all configuration
- Simpler test execution: just `make test`
- Isolated test DB (port 5434) separate from dev (port 5433)
- Better documentation and developer experience

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

**Simplified Configuration:**
- Test database credentials now in .env.dev (single source of truth)
- docker-compose.dev.yml uses env vars for test DB (POSTGRES_TEST_*)
- Makefile sources .env.dev for all test commands
- No need for separate .env.test file

**Removed:**
- run-tests.sh - Redundant, use `make test` instead

**Benefits:**
- All local dev config in one place (.env.dev)
- Less mental overhead (dev + test in same file)
- Consistent variable usage throughout stack
- Simpler developer onboarding

Usage:
- `make test` - Run all tests (handles DB automatically)
- All test config in .env.dev under "Test Database Configuration"

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

Major cleanup now that PDS handles all repository operations:

**Removed:**
- internal/core/repository/ - Repository domain logic (PDS handles this)
- internal/db/postgres/repository_repo.go - Repository database operations
- internal/api/handlers/repository_handler.go - Repository API handlers
- internal/api/routes/repository.go - Repository routes
- tests/integration/repository_test.go - Repository integration tests
- Migrations 002, 003, 004 - Repository/CAR storage tables
- internal/db/local_dev_db_compose/ - Separate dev database setup
- internal/db/test_db_compose/ - Separate test database setup

**Unified:**
- docker-compose.dev.yml now includes PostgreSQL + PDS + optional test DB
- All database management moved to Makefile commands
- Consistent use of .env.dev variables throughout

**Updated:**
- cmd/server/main.go - Simplified to only use user service
- Makefile - All-in-one commands (dev-up starts both PostgreSQL + PDS)
- Added db-migrate, db-reset, test commands using Docker profiles

**Architecture:**
- PDS: Self-contained with SQLite + CAR files (port 3001)
- PostgreSQL: Only for Coves AppView indexing (port 5433)
- Test DB: Available via --profile test (port 5434)
- Single source of truth: docker-compose.dev.yml + .env.dev

Commands:
- `make dev-up` - Start PostgreSQL + PDS
- `make test` - Start test DB + run tests
- `make db-migrate` - Run migrations
- `make db-shell` - Open psql shell

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

- Add docker-compose.dev.yml with Bluesky PDS (port 3001)
- Add .env.dev with development configuration
- Add Makefile with convenient dev commands (help, dev-up, dev-down, etc.)
- Add comprehensive docs/LOCAL_DEVELOPMENT.md guide
- Update CLAUDE.md and ATPROTO_GUIDE.md with correct architecture
- Remove custom carstore implementation (PDS handles this)
- Remove internal/atproto/repo wrapper (not needed)
- Add feed lexicon schemas (getAll, getCommunity, getTimeline)
- Update post lexicons to remove getFeed (replaced by feed queries)
- Update PROJECT_STRUCTURE.md to reflect new architecture

Architecture:
- PDS is self-contained with internal SQLite + CAR storage
- PostgreSQL database only used by Coves AppView for indexing
- AppView subscribes directly to PDS firehose (no relay needed for local dev)
- PDS runs on port 3001 to avoid conflicts with production PDS on 3000

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

feat: Implement union types for moderation configuration

- Replace single moderationConfig with union type supporting both moderator and sortition variants
- Add $type discriminator field following atProto patterns
- Separate tribunal-specific fields (tribunalThreshold, jurySize) to sortition variant only
- Update test files to use new union structure with proper $type fields
- Ensure type safety: only sortition communities can configure tribunal settings

This change improves the lexicon design by:
- Following atProto best practices for discriminated unions
- Providing clear separation between moderation types
- Enabling future extensibility for new moderation approaches
- Maintaining backwards compatibility through the union pattern

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

feat: Add comprehensive lexicon test data and validation fixes

- Add 35 test data files covering all lexicon record types
- Actor records: block, membership, preferences, profile, saved, subscription
- Community records: moderator, profile, rules, wiki
- Interaction records: comment, share, tag
- Moderation records: ruleProposal, tribunalVote, vote
- Post records: post validation

- Fix lexicon schema issues:
- Add 'handle' format to actor profile
- Fix typo in community profile moderation type
- Add required $type field to interaction comments
- Add minItems constraint to tag arrays
- Fix enum values in moderation schemas

- Improve validate-lexicon tool:
- Better number handling to prevent float64 conversion issues
- Add json.Number support for accurate integer validation

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

feat: Implement comprehensive lexicon validation system

- Add internal/validation/lexicon.go with ValidateLexiconData function
- Create validate-lexicon CLI tool for testing lexicon schemas
- Add comprehensive test suite with valid and invalid test cases
- Include test data for actors, communities, posts, interactions, and moderation
- Replace shell-based validation script with Go implementation
- Support for complex validation including enums, unions, and references

This provides a robust foundation for validating atProto lexicon data
in the Coves platform, ensuring data integrity and type safety.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

feat: Add domain knowledge documentation and update lexicon schemas

- Add comprehensive DOMAIN_KNOWLEDGE.md documenting Coves platform concepts
- Update CLAUDE.md with improved development guidelines
- Enhance actor preferences and subscription lexicon schemas
- Improve post retrieval and feed functionality with better tag support
- Update create-pr command documentation

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

feat: Unify post lexicon architecture and improve tag handling

Major architectural improvements to the post lexicon system:

- Unified 5 separate post types into single record with postType discriminator
- Moved tags from author-created content to community interactions (following Bluesky pattern)
- Added tagCounts to post stats and viewer tags to viewerState
- Cleaned up legacy fields (removed nsfw boolean, use contentLabels)
- Simplified embedType enum (image-gallery → image)
- Updated federation schema to use specific platform names
- Clarified title field as optional for certain post types
- Support for up to 8 images in image embeds

This change enables:
- Simpler codebase with single post record type
- Filtering by post type(s) at query level
- Community-driven tagging system
- Consistent embed handling across all post types
- Better federation support with originalAuthor field

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

This reverts commit 32f6836b6ebb8af1a475b06c4045a545f06cdcff.

Major architectural improvements to the post lexicon system:

- Unified 5 separate post types into single record with postType discriminator
- Moved tags from author-created content to community interactions (following Bluesky pattern)
- Added tagCounts to post stats and viewer tags to viewerState
- Cleaned up legacy fields (removed nsfw boolean, use contentLabels)
- Simplified embedType enum (image-gallery → image)
- Updated federation schema to use specific platform names
- Clarified title field as optional for certain post types
- Support for up to 8 images in image embeds

This change enables:
- Simpler codebase with single post record type
- Filtering by post type(s) at query level
- Community-driven tagging system
- Consistent embed handling across all post types
- Better federation support with originalAuthor field

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

feat: Reimplement richtext using Bluesky-inspired facet system

- Fixed typo "filesred" → "files" in validate-lexicon/main.go
- Added validate-lexicon binary to .gitignore and removed from git
- Updated facet schema to include required $type fields for AT Protocol compatibility
- Removed duplicate mentions field from microblog.json (use facets instead)
- Added comprehensive facet tests covering UTF-8 byte counting and all feature types
- Fixed lexicon validation test to reference correct schema names
- Added detailed facet documentation with UTF-8 byte counting examples

The facet implementation now follows AT Protocol standards with proper $type
fields for each feature (mention, link, bold, italic, strikethrough, spoiler).
All byte indices use UTF-8 encoding for cross-platform consistency.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

- Replace separate markup/mention/link lexicons with unified facet approach
- Use byte-indexed positioning (byteStart/byteEnd) for text annotations
- Support overlapping features: bold, italic, strikethrough, spoiler, mention, link
- Add native support for community mentions with \! prefix
- Update all schemas to use *Facets fields instead of *Markup
- Align with AT Protocol standards for better federation compatibility

BREAKING CHANGE: All richtext markup fields renamed to facets

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

Implement query and procedure lexicon definitions

This commit adds a complete set of lexicon definitions required for the
core functionality of the Coves social platform, following AT Protocol
best practices.

Added 29 new lexicon files across 4 namespaces:

Actor namespace (7 files):
- blockUser.json: Block another user
- getProfile.json: Retrieve user profile with stats
- preferences.json: User preferences record type
- saveItem.json: Save posts/comments for later
- unblockUser.json: Unblock a previously blocked user
- unsaveItem.json: Remove saved items
- updateProfile.json: Update user profile information

Community namespace (10 files):
- create.json: Create a new community
- get.json: Retrieve community details
- getMembers.json: List community members
- getSubscribers.json: List community subscribers
- list.json: List communities with filtering
- moderator.json: Moderator assignment record type
- search.json: Search communities by text
- subscribe.json: Subscribe to a community
- unsubscribe.json: Unsubscribe from a community
- update.json: Update community settings

Post namespace (6 files):
- create.json: Create posts (text, image, video, article, microblog)
- delete.json: Delete a post
- get.json: Retrieve a single post
- getFeed.json: Get post feeds with sorting options
- search.json: Search posts by text
- update.json: Update post content

Interaction namespace (4 files):
- createComment.json: Comment on posts/comments
- deleteComment.json: Delete comments
- createVote.json: Vote on posts/comments
- deleteVote.json: Remove votes

Bug fixes:
- Fixed duplicate 'createdAt' in membership.json required array
- Corrected getSaved.json type from 'procedure' to 'query'

All lexicons follow the established patterns:
- Queries for read operations
- Procedures for write operations
- Proper error definitions
- Consistent parameter naming
- Support for pagination where appropriate

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

Refactor: ATProto Repository Storage with Indigo Carstore Integration

- Remove blob storage from repository interface and implementation
- Add GORM integration for better database management
- Integrate Indigo's carstore for proper CAR file storage
- Update repository wrapper to use external blockstore
- Add comprehensive test infrastructure with test database setup
- Add test environment configuration example
- Update dependencies to include required Indigo and IPFS packages
- Fix repository record operations to work without inline value storage
- Add proper error handling and migrations for carstore integration

This refactoring aligns the repository storage with ATProto standards by:
1. Using Indigo's carstore for efficient CAR file management
2. Removing custom blob storage in favor of standard IPFS blockstore
3. Improving test infrastructure for better integration testing

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

Feature/lexicon implementation