haiku.rag/CHANGELOG.md

# Changelog
## [Unreleased]

### Changed

- **Download Models Progress**: `haiku-rag download-models` now shows real-time progress with Rich progress bars for Ollama model downloads
- **Refactored Download Models**: Moved core download logic to `HaikuRAG.download_models()` async generator that yields `DownloadProgress` events, separating business logic from UI

## [0.20.0] - 2025-11-28

### Added

- **DoclingDocument Storage**: Full DoclingDocument JSON is now stored with each document, enabling rich context and visual grounding
  - Documents store the complete DoclingDocument structure (JSON) and schema version
  - Chunks store metadata with JSON pointer references (`doc_item_refs`), semantic labels, section headings, and page numbers
  - New `ChunkMetadata` model for structured chunk provenance: `doc_item_refs`, `headings`, `labels`, `page_numbers`
  - `Document.get_docling_document()` method to parse stored DoclingDocument
  - `ChunkMetadata.resolve_doc_items()` to resolve JSON pointer refs to actual DocItem objects
  - `ChunkMetadata.resolve_bounding_boxes()` for visual grounding with page coordinates
  - LRU cache (100 documents) for parsed DoclingDocument objects to avoid repeated JSON parsing
- **Enhanced Search Results**: `search()` and `expand_context()` now return full provenance information
  - `SearchResult` includes `page_numbers`, `headings`, `labels`, and `doc_item_refs`
  - QA and research agents use provenance for better citations (page numbers, section headings)
- **Inspector Visual Grounding**: New visual grounding modal in the database inspector
  - View page images with highlighted bounding boxes for chunks
  - Keyboard navigation between pages (←/→ arrows)
  - Access from both main detail view and search results
  - Requires `textual-image` dependency
- **Visual Grounding CLI**: New `haiku-rag visualize <chunk_id>` command
  - Displays page images with highlighted bounding boxes for a chunk
  - Requires terminal with image support (iTerm2, Kitty, etc.)
- **New `import_document()` Method**: Import pre-processed documents with custom chunks
  - Use when document conversion, chunking, and embedding were done externally
  - `chunks` required; `content` optional if `docling_document_json` is provided
  - When `docling_document_json` is provided without `content`, content is extracted from the DoclingDocument
  - Validates docling JSON parses correctly if provided
- **`update_document_fields()` DoclingDocument Support**: Added `docling_document_json` and `docling_version` parameters
  - When `docling_document_json` is provided without `chunks`, content is extracted and document is rechunked
  - When `docling_document_json` is provided with `chunks`, both are stored (chunks used as-is)
  - `content` and `docling_document_json` are mutually exclusive to avoid ambiguity
- **New `convert()` Method**: Convert files, URLs, or text to DoclingDocument
  - `client.convert(Path(...))` - convert local file
  - `client.convert("https://...")` - download and convert URL
  - `client.convert("text content")` - convert plain text
  - Supports `file://` URIs
- **New `chunk()` Method**: Chunk a DoclingDocument into Chunk objects
  - `client.chunk(docling_doc)` - returns `list[Chunk]` without embeddings
- **New `contextualize()` and `embed_chunks()` Utilities**: Standalone embedding utilities in `haiku.rag.embeddings`
  - `contextualize(chunks)` - prepend section headings to chunk content for better semantic search
  - `embed_chunks(chunks)` - generate embeddings for chunks, returns new Chunk objects with embeddings set

### Changed

- **BREAKING: `create_document()` API**: Removed `chunks` parameter
  - `create_document()` now always processes content (converts, chunks, embeds)
  - Use new `import_document()` for pre-processed documents with custom chunks
- **BREAKING: Chunker Interface**: `DocumentChunker.chunk()` now returns `list[Chunk]` instead of `list[str]`
  - Chunks include structured metadata (doc_item_refs, labels, headings, page_numbers) in the `metadata` dict
  - All chunker implementations updated: `DoclingLocalChunker`, `DoclingServeChunker`
- **Page Image Generation**: `generate_page_images=True` is now always enabled for local docling converter
  - Required for visual grounding features
  - Removed `generate_page_images` config option (docling-serve already generates page images by default)
- **Chunk Text Storage**: Chunks now store raw text without heading contextualization
  - Section headings are prepended only at embedding time for better semantic search
  - Stored chunk content stays clean without duplicate heading prefixes
  - Headings remain available in `ChunkMetadata` for display and citations
  - Local and serve chunkers now produce identical output
- **QA Prompts**: Updated to use page numbers and section headings in citations when available
- **Citation Models**: Introduced `RawSearchAnswer` for LLM output, `SearchAnswer` extends it with resolved citations
  - Cleaner separation: LLM outputs chunk IDs, citations resolved programmatically
  - `Citation` fields are now required (no defaults) for type safety

### Removed

- **BREAKING: `markdown_preprocessor` Config Option**: Removed the `processing.markdown_preprocessor` configuration option
  - Use `convert()`, `chunk()`, and `embed_chunks()` primitives for custom processing pipelines
  - Transform content at any stage before calling `import_document()`

### Migration

This release requires a database rebuild to populate the new DoclingDocument fields:

```bash
haiku-rag rebuild
```

Existing documents without DoclingDocument data will work but won't have provenance information. The `rebuild` command re-processes all documents to populate the new fields.

## [0.19.6] - 2025-12-03

### Changed

- **BREAKING: Explicit Database Creation**: Databases must now be explicitly created before use
  - New `haiku-rag init` command creates a new empty database
  - Python API: `HaikuRAG(path, create=True)` to create database programmatically
  - Operations on non-existent databases raise `FileNotFoundError`
- **BREAKING: Embeddings Configuration**: Restructured to nested `EmbeddingModelConfig`
  - Config path changed from `embeddings.{provider, model, vector_dim}` to `embeddings.model.{provider, name, vector_dim}`
  - Automatic migration upgrades existing databases to new format
- **Database Migrations**: Always run when opening an existing database

## [0.19.5] - 2025-12-01

### Changed

- **Rebuild Performance**: Optimized `rebuild --embed-only` to use batch updates via LanceDB's `merge_insert` instead of individual chunk updates, and skip chunks with unchanged embeddings

## [0.19.4] - 2025-11-28

### Added

- **Rebuild Modes**: New options for `rebuild` command to control what gets rebuilt
  - `--embed-only`: Only regenerate embeddings, keeping existing chunks (fastest option when changing embedding model)
  - `--rechunk`: Re-chunk from existing document content without accessing source files
  - Default (no flag): Full rebuild with source file re-conversion
  - Python API: `rebuild_database(mode=RebuildMode.EMBED_ONLY | RECHUNK | FULL)`

## [0.19.3] - 2025-11-27

### Changed

- **Async Chunker**: `DoclingServeChunker` now uses `httpx.AsyncClient` instead of sync `requests`

### Fixed

- **OCR Options**: Fixed `DoclingLocalConverter` using base `OcrOptions` class which docling's OCR factory doesn't recognize. Now uses `OcrAutoOptions` for automatic OCR engine selection.
- **Dependencies**: Added `opencv-python-headless` to the `docling` optional dependency for table structure detection.

## [0.19.2] - 2025-11-27

### Changed

- **Async Converters**: Made document converters fully async
  - `BaseConverter.convert_file()` and `convert_text()` are now async methods
  - `DoclingLocalConverter` wraps blocking Docling operations with `asyncio.to_thread()`
  - `DoclingServeConverter` now uses `httpx.AsyncClient` instead of sync `requests`
- **Async Model Prefetch**: `prefetch_models()` is now async
  - Uses `httpx.AsyncClient` for Ollama model pulls
  - Wraps blocking Docling and HuggingFace downloads with `asyncio.to_thread()`

## [0.19.1] - 2025-11-26

### Added

- **LM Studio Provider**: Added support for LM Studio as a provider for embeddings and QA/research models
  - Configure with `provider: lm_studio` in embeddings, QA, or research model settings
  - Supports thinking control for reasoning models (gpt-oss, etc.)
  - Default base URL: `http://localhost:1234`

### Fixed

- **Configuration**: Fixed `init-config` command generating invalid configuration files (#165)
  - Refactored `generate_default_config()` to use Pydantic model serialization instead of manual dict construction
  - Updated `qa`, `research`, and `reranking` sections to use new `ModelConfig` structure

## [0.19.0] - 2025-11-25

### Added

- **Model Customization**: Added support for per-model configuration settings
  - New `enable_thinking` parameter to control reasoning behavior (true/false/None)
  - Support for `temperature` and `max_tokens` settings on QA and research models
  - All settings apply to any provider that supports them
- **Database Inspector**: New `inspect` CLI command launches interactive TUI for browsing documents and chunks & searching
- **Evaluations**: Added `evaluations` CLI script for running benchmarks (replaces `python -m evaluations.benchmark`)
- **Evaluations**: Added `--db` option to override evaluation database path
  - Default database location moved to haiku.rag data directory:
    - macOS: `~/Library/Application Support/haiku.rag/evaluations/dbs/`
    - Linux: `~/.local/share/haiku.rag/evaluations/dbs/`
    - Windows: `C:/Users/<USER>/AppData/Roaming/haiku.rag/evaluations/dbs/`
  - Previously stored in `evaluations/data/` within the repository
- **Evaluations**: Added comprehensive experiment metadata tracking for better reproducibility
  - Records dataset name, test case count, and all model configurations
  - Tracks embedder settings: provider, model, and vector dimensions
  - Tracks QA model: provider and model name
  - Tracks judge model: provider and model name for LLM evaluation
  - Tracks processing parameters: `chunk_size` and `context_chunk_radius`
  - Tracks retrieval configuration: `retrieval_limit` for number of chunks retrieved
  - Tracks reranking configuration: `rerank_provider` and `rerank_model`
  - Enables comparison of evaluation runs with different configurations in Logfire
- **Evaluations**: Refactored retrieval evaluation to use pydantic-ai experiment framework
  - New `evaluators` module with `MRREvaluator` (Mean Reciprocal Rank) and `MAPEvaluator` (Mean Average Precision)
  - Retrieval benchmarks now use `Dataset.evaluate()` with full Logfire experiment tracking
  - Dataset specifications now declare their retrieval evaluator (MRR for RepliQA, MAP for Wix)
  - Replaced Recall@K and Success@K with industry-standard MRR and MAP metrics
  - Unified evaluation framework for both retrieval and QA benchmarks
- **AG-UI Events**: Enhanced ActivitySnapshot events with richer structured data
  - Added `stepName` field to identify which graph node emitted each activity
  - Added structured fields to activity content while preserving backward-compatible `message` field:
    - **Planning**: `sub_questions` - list of sub-question strings
    - **Searching**: `query` - the search query, `confidence` - answer confidence (on success), `error` - error message (on failure)
    - **Analyzing** (research): `insights` - list of insight objects, `gaps` - list of gap objects, `resolved_gaps` - list of resolved gap strings
    - **Evaluating** (research): `confidence` - confidence score, `is_sufficient` - sufficiency flag
    - **Evaluating** (deep QA): `is_sufficient` - sufficiency flag, `iterations` - iteration count

### Changed

- **Evaluations**: Renamed `--qa-limit` CLI parameter to `--limit`, now applies to both retrieval and QA benchmarks
- **Evaluations**: Retrieval evaluator selection moved from runtime logic to dataset configuration

## [0.18.0] - 2025-11-21

### Added

- **Manual Vector Indexing**: New `create-index` CLI command for explicit vector index creation
  - Creates IVF_PQ indexes
  - Requires minimum 256 chunks (LanceDB training data requirement)
  - New `search.vector_index_metric` config option: `cosine` (default), `l2`, or `dot`
  - New `search.vector_refine_factor` config option (default: 30) for accuracy/speed tradeoff
  - Indexes not created automatically during ingestion to avoid performance degradation
  - Manual rebuilding required after adding significant new data
- **Enhanced Info Command**: `haiku-rag info` now shows storage sizes and vector index statistics
  - Displays storage size for documents and chunks tables in human-readable format
  - Shows vector index status (exists/not created)
  - Shows indexed and unindexed chunk counts for monitoring index staleness

### Changed

- **BREAKING: Default Embedding Model**: Changed default embedding model from `qwen3-embedding` to `qwen3-embedding:4b` with vector dimension 2560 (previously 4096)
  - New installations will use the smaller, more efficient 4B parameter model by default
  - **Action required**: Existing databases created with the old default will be incompatible. Users must either:
    - Explicitly set `embeddings.model: "qwen3-embedding"` and `embeddings.vector_dim: 4096` in their config to maintain compatibility with existing databases
    - Or run `haiku-rag rebuild` to re-embed all documents with the new default
  - This change provides better performance for most use cases while reducing resource requirements
- **Evaluations**: Improved evaluation dataset naming and simplified evaluator configuration
  - `EvalDataset` now accepts dataset name for better organization in Logfire
  - Added `--name` CLI parameter to override evaluation run names
  - Removed `IsInstance` evaluator, using only `LLMJudge` for QA evaluation
- **Search Accuracy**: Applied `refine_factor` to vector and hybrid searches for improved accuracy
  - Retrieves `refine_factor * limit` candidates and re-ranks in memory
  - Higher values increase accuracy but slow down queries

### Fixed

- **AG-UI Activity Events**: Activity events now correctly use structured dict content instead of strings
- **Graph Configuration**: Graph builder functions now properly accept and use non-global config (#149)
  - `build_research_graph()` and `build_deep_qa_graph()` now pass config to all agents and model creation
  - `get_model()` utility function accepts `config` parameter (defaults to global Config)
  - Allows creating multiple graphs with different configurations in the same application


## [0.17.2] - 2025-11-19

### Added

- **Document Update API**: New `update_document_fields()` method for partial document updates
  - Update individual fields (content, metadata, title, chunks) without fetching full document
  - Support for custom chunks or auto-generation from content

### Changed

- **Chunk Creation**: `ChunkRepository.create()` now accepts both single chunks and lists for batch insertion
  - Batch insertion reduces LanceDB version creation when adding multiple chunks with custom chunks
  - Batch embedding generation for improved performance with multiple chunks
- Updated core dependencies

## [0.17.1] - 2025-11-18

### Added

- **Conversion Options**: Fine-grained control over document conversion for both local and remote converters
  - New `conversion_options` config section in `ProcessingConfig`
  - OCR settings: `do_ocr`, `force_ocr`, `ocr_lang` for controlling OCR behavior
  - Table extraction: `do_table_structure`, `table_mode` (fast/accurate), `table_cell_matching`
  - Image settings: `images_scale` to control image resolution
  - Options work identically with both `docling-local` and `docling-serve` converters

### Changed

- Increase reranking candidate retrieval multiplier from 3x to 10x for improved result quality
- **Docker Images**: Main `haiku.rag` image no longer automatically built and published
- **Conversion Options**: Removed the legacy `pdf_backend` setting; docling now chooses the optimal backend automatically

## [0.17.0] - 2025-11-17

### Added

- **Remote Processing**: Support for docling-serve as remote document processing and chunking service
  - New `converter` config option: `docling-local` (default) or `docling-serve`
  - New `chunker` config option: `docling-local` (default) or `docling-serve`
  - New `providers.docling_serve` config section with `base_url`, `api_key`, and `timeout`
  - Comprehensive error handling for connection, timeout, and authentication issues
- **Chunking Strategies**: Support for both hybrid and hierarchical chunking
  - New `chunker_type` config option: `hybrid` (default) or `hierarchical`
  - Hybrid chunking: Structure-aware splitting that respects document boundaries
  - Hierarchical chunking: Preserves document hierarchy for nested documents
- **Table Serialization Control**: Configurable table representation in chunks
  - New `chunking_use_markdown_tables` config option (default: `false`)
  - `false`: Tables serialized as narrative text ("Value A, Column 2 = Value B")
  - `true`: Tables preserved as markdown format with structure
- **Chunking Configuration**: Additional chunking control options
  - New `chunking_merge_peers` config option (default: `true`) to merge undersized successive chunks
- **Docker Images**: Two Docker images for different deployment scenarios
  - `haiku.rag`: Full image with all dependencies for self-contained deployments
  - `haiku.rag-slim`: Minimal image designed for use with external docling-serve
  - Multi-platform support (linux/amd64, linux/arm64)
  - Docker Compose examples with docling-serve integration
  - Automated CI/CD workflows for both images
  - Build script (`scripts/build-docker-images.sh`) for local multi-platform builds

### Changed

- **BREAKING: Chunking Tokenizer**: Switched from tiktoken to HuggingFace tokenizers for consistency with docling-serve
  - Default tokenizer changed from tiktoken "gpt-4o" to "Qwen/Qwen3-Embedding-0.6B"
  - New `chunking_tokenizer` config option in `ProcessingConfig` for customization
  - `download-models` CLI command now also downloads the configured HuggingFace tokenizer
- **Docker Examples**: Updated examples to demonstrate remote processing
  - `examples/docker` now uses slim image with docling-serve
  - `examples/ag-ui-research` backend uses slim image with docling-serve
  - Configuration examples include remote processing setup

## [0.16.1] - 2025-11-14

### Changed

- **Evaluations**: Refactored QA benchmark to run entire dataset as single evaluation for better Logfire experiment tracking
- **Evaluations**: Added `.env` file loading support via `python-dotenv` dependency

## [0.16.0] - 2025-11-13

### Added

- **AG-UI Protocol Support**: Full AG-UI (Agent-UI) protocol implementation for graph execution with event streaming
  - New `AGUIEmitter` class for emitting AG-UI events from graphs
  - Support for all AG-UI event types: lifecycle events (`RUN_STARTED`, `RUN_FINISHED`, `RUN_ERROR`), step events (`STEP_STARTED`, `STEP_FINISHED`), state updates (`STATE_SNAPSHOT`, `STATE_DELTA`), activity narration (`ACTIVITY_SNAPSHOT`), and text messages (`TEXT_MESSAGE_CHUNK`)
  - `AGUIConsoleRenderer` for rendering AG-UI event streams to terminal with Rich formatting
  - `stream_graph()` utility function for executing graphs with AG-UI event emission
  - State diff computation for efficient state synchronization
  - **Delta State Updates**: AG-UI emitter now supports incremental state updates via JSON Patch operations (`STATE_DELTA` events) to reduce bandwidth, configurable via `use_deltas` parameter (enabled by default)
- **AG-UI Server**: Starlette-based HTTP server for serving graphs via AG-UI protocol
  - Server-Sent Events (SSE) streaming endpoint at `/v1/agent/stream`
  - Health check endpoint at `/health`
  - Full CORS support configurable via `agui` config section
  - `create_agui_server()` function for programmatic server creation
- **Deep QA AG-UI Support**: Deep QA graph now fully supports AG-UI event streaming
  - Integration with `AGUIEmitter` for progress tracking
  - Step-by-step execution visibility via AG-UI events
- **CLI AG-UI Flag**: New `--agui` flag for `serve` command to start AG-UI server
- **Graph Module**: New unified `haiku.rag.graph` module containing all graph-related functionality
- **Common Graph Nodes**: New factory functions (`create_plan_node`, `create_search_node`) in `haiku.rag.graph.common.nodes` for reusable graph components
- **AG-UI Research Example**: New full-stack example (`examples/ag-ui-research`) demonstrating agent+graph architecture with CopilotKit frontend
  - Pydantic AI agent with research tool that invokes the research graph
  - Custom AG-UI streaming endpoint with anyio memory streams
  - React/Next.js frontend with split-pane UI showing live research state
  - Real-time progress tracking of questions, answers, insights, and gaps
  - Docker Compose setup for easy local development

### Changed

- **Vacuum Retention**: Default `vacuum_retention_seconds` increased from 60 seconds to 86400 seconds (1 day) for better version retention in typical workflows
- **BREAKING**: Major refactoring of graph-related code into unified `haiku.rag.graph` module structure:
  - `haiku.rag.research` → `haiku.rag.graph.research`
  - `haiku.rag.qa.deep` → `haiku.rag.graph.deep_qa`
  - `haiku.rag.agui` → `haiku.rag.graph.agui`
  - `haiku.rag.graph_common` → `haiku.rag.graph.common`
- **BREAKING**: Research and Deep QA graphs now use AG-UI event protocol instead of direct console logging
  - Removed `console` and `stream` parameters from graph dependencies
  - All progress updates now emit through `AGUIEmitter`
- **BREAKING**: `ResearchState` converted from dataclass to Pydantic `BaseModel` for JSON serialization and AG-UI compatibility
- Research and Deep QA graphs now emit detailed execution events for better observability
- CLI research command now uses AG-UI event rendering for `--verbose` output
- Improved graph execution visibility with step-by-step progress tracking
- Updated all documentation to reflect new import paths and AG-UI usage
- Updated examples (ag-ui-research, a2a-server) to use new import paths

### Fixed

- **Document Creation**: Optimized `create_document` to skip unnecessary DoclingDocument conversion when chunks are pre-provided
- **FileReader**: Error messages now include both original exception details and file path for easier debugging
- **Database Auto-creation**: Read operations (search, list, get, ask, research) no longer auto-create empty databases. Write operations (add, add-src, delete, rebuild) still create the database as needed. This prevents the confusing scenario where a search query creates an empty database. Fixes issue #137.

### Removed

- **BREAKING**: Removed `disable_autocreate` config option - the behavior is now automatic based on operation type
- **BREAKING**: Removed legacy `ResearchStream` and `ResearchStreamEvent` classes (replaced by AG-UI event protocol)

## [0.15.0] - 2025-11-07

### Added

- **File Monitor**: Orphan deletion feature - automatically removes documents from database when source files are deleted (enabled via `monitor.delete_orphans` config option, default: false)

### Changed

- **Configuration**: All CLI commands now properly support `--config` parameter for specifying custom configuration files
- Configuration loading consolidated across CLI, app, and client with consistent resolution order
- `HaikuRAGApp` and MCP server now accept `config` parameter for programmatic configuration
- Updated CLI documentation to clarify global vs per-command options
- **BREAKING**: Standardized configuration filename to `haiku.rag.yaml` in user directories (was incorrectly using `config.yaml`). Users with existing `config.yaml` in their user directory will need to rename it to `haiku.rag.yaml`

### Fixed

- **File Monitor**: Fixed incorrect "Updated document" logging for unchanged files - monitor now properly skips files when MD5 hash hasn't changed

### Removed

- **BREAKING**: A2A (Agent-to-Agent) protocol support has been moved to a separate self-contained package in `examples/a2a-server/`. The A2A server is no longer part of the main haiku.rag package. Users who need A2A functionality can install and run it from the examples directory with `cd examples/a2a-server && uv sync`.
- **BREAKING**: Removed deprecated `.env`-based configuration system. The `haiku-rag init-config --from-env` command and `load_config_from_env()` function have been removed. All configuration must now be done via YAML files. Environment variables for API keys (e.g., `OPENAI_API_KEY`, `ANTHROPIC_API_KEY`) and service URLs (e.g., `OLLAMA_BASE_URL`) are still supported and can be set via `.env` files.

## [0.14.1] - 2025-11-06

### Added

- Migrated research and deep QA agents to use Pydantic Graph beta API for better graph execution
- Automatic semaphore-based concurrency control for parallel sub-question processing
- `max_concurrency` parameter for controlling parallel execution in research and deep QA (default: 1)

### Changed

- **BREAKING**: Research and Deep QA graphs now use `pydantic_graph.beta` instead of the class-based graph implementation
- Refactored graph common patterns into `graph_common` module
- Sub-questions now process using `.map()` for true parallel execution
- Improved graph structure with cleaner node definitions and flow control
- Pinned critical dependencies: `docling-core`, `lancedb`, `docling`

## [0.14.0] - 2024-11-05

### Added

- New `haiku.rag-slim` package with minimal dependencies for users who want to install only what they need
- Evaluations package (`haiku.rag-evals`) for internal benchmarking and testing
- Improved search filtering performance by using pandas DataFrames for joins instead of SQL WHERE IN clauses

### Changed

- **BREAKING**: Restructured project into UV workspace with three packages:
  - `haiku.rag-slim` - Core package with minimal dependencies
  - `haiku.rag` - Full package with all extras (recommended for most users)
  - `haiku.rag-evals` - Internal benchmarking and evaluation tools
- Migrated from `pydantic-ai` to `pydantic-ai-slim` with extras system
- Docling is now an optional dependency (install with `haiku.rag-slim[docling]`)
- Package metadata checks now use `haiku.rag-slim` (always present) instead of `haiku.rag`
- Docker image optimized: removed evaluations package, reducing installed packages from 307 to 259
- Improved vector search performance through optimized score normalization

### Fixed

- ImportError now properly raised when optional docling dependency is missing

## [0.13.3] - 2024-11-04

### Added

- Support for Zero Entropy reranker
- Filter parameter to `search()` for filtering documents before search
- Filter parameter to CLI `search` command
- Filter parameter to CLI `list` command for filtering document listings
- Config option to pass custom configuration files to evaluation commands
- Document filtering now respects configured include/exclude patterns when using `add-src` with directories
- Max retries to insight_agent when producing structured output

### Fixed

- CLI now loads `.env` files at startup
- Info command no longer attempts to use deprecated `.env` settings
- Documentation typos

## [0.13.2] - 2024-11-04

### Added

- Gitignore-style pattern filtering for file monitoring using pathspec
- Include/exclude pattern documentation for FileMonitor

### Changed

- Moved monitor configuration to its own section in config
- Improved configuration documentation
- Updated dependencies

## [0.13.1] - 2024-11-03

### Added

- Initial version tracking

[Unreleased]: https://github.com/ggozad/haiku.rag/compare/0.14.0...HEAD
[0.14.0]: https://github.com/ggozad/haiku.rag/compare/0.13.3...0.14.0
[0.13.3]: https://github.com/ggozad/haiku.rag/compare/0.13.2...0.13.3
[0.13.2]: https://github.com/ggozad/haiku.rag/compare/0.13.1...0.13.2
[0.13.1]: https://github.com/ggozad/haiku.rag/releases/tag/0.13.1