Skip to content

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.0.43] - 2026-04-30

Added

  • Structured Streaming API (invoke_model_streaming) (AA-3251): New method on BaseLLMService and AWSBedrockService providing a typed streaming interface
  • Yields structured dicts instead of raw strings: {"type": "text", "text": "..."} for response content and {"type": "thinking", "thinking": "..."} for reasoning traces

  • Enables callers to separately handle visible response text and chain-of-thought reasoning

  • Extended Thinking Support (AA-3251): Full end-to-end support for Bedrock extended thinking (chain-of-thought reasoning before the final answer)

  • extended_thinking: bool and thinking_budget_tokens: int parameters accepted at the LangGraphAgent.query() call site
  • Parameters propagated through the full LangGraph stack: LangGraphAgent → ToolUseWorkflow → ResponseGenerationNode → AWSBedrockService
  • Automatically enforces Bedrock requirements: sets temperature=1, removes top_p, applies a minimum budget_tokens of 1,024

  • Reasoning blocks logged at DEBUG / INFO level as they arrive

  • Extended Thinking Examples: Four new runnable examples demonstrating extended thinking workflows

  • extended_thinking_tdd.py: Talk-to-document agent with extended thinking enabled
  • extended_thinking_tdd_execute.py: Script executor for the above
  • extended_thinking_objectives.py: Objective-based agent with extended thinking enabled

  • extended_thinking_objectives_execute.py: Script executor for the above

  • langgraph_with_simple_tool.py: New example demonstrating LangGraph workflow with a simple tool alongside a structured streaming consumer

Changed

  • invoke_model_stream() enhanced: Now handles Bedrock thinking content blocks in the event stream
  • _yield_structured=True flag switches from raw string yield to typed dict yield (used internally by invoke_model_streaming)

  • Backwards-compatible: existing callers that use invoke_model_stream() without the flag receive unchanged string output

  • Streaming completion log: Now reports both response character count and thinking character count for observability

  • Before: Streaming completed. Total length: N

  • After: Streaming completed. response=N chars, thinking=M chars, execution_time=X.XXs

  • Example cleanup: Removed nine outdated example scripts that no longer reflected current SDK patterns

  • draft_mail.py, lang_tool.py, langgraph_ttd.py, langgraph_with_simple_tool copy.py, objctive_agent_troubled_doc.py, objctive_agent_with_tools.py, objective_no_lang.py, objectives.py, objectives_execute.py

Technical Details

  • AWSBedrockService.invoke_model_stream(): Extended thinking path sets body["thinking"] = {"type": "enabled", "budget_tokens": budget} before the Bedrock invoke_model_with_response_stream call
  • AWSBedrockService.invoke_model_streaming(): Thin wrapper that sets _yield_structured=True and delegates to invoke_model_stream()
  • WorkflowConfig metadata keys added: extended_thinking (default False), thinking_budget_tokens (default 10000)
  • BaseLLMService.invoke_model_streaming() declared as an abstract method — all concrete subclasses must implement it

[0.0.38] - 2026-03-04

Changed

  • Removed all in-memory prompt and agent caching from SDK core and examples
  • get_system_prompt() now always fetches fresh from DynamoDB
  • DynamoDBPromptConfig.get_prompt_template() no longer caches results
  • Removed module-level caches from generator_agent.py and langgraph_ttd.py
  • Fixes stale configuration in Lambda container reuse scenarios

Improved

  • Consolidated test suite from 1,147 tests to 238 while maintaining 82% coverage
  • Adopted Given/When/Then docstring convention for test methods
  • Fixed all E501 flake8 violations in test files

[0.0.34] - 2026-03-18

Added

  • Module Embeddings (AS-1593): New akordi_agents/embeddings/ module with pluggable provider architecture
  • EmbeddingProviderInterface ABC with embed(), embed_single(), get_dimensions() methods
  • BedrockTitanEmbeddingProvider using Amazon Titan Text Embeddings V2 with concurrent batching
  • EmbeddingService with provider registry and runtime switching
  • Pydantic models: EmbeddingConfig, EmbeddingRequest, EmbeddingResult
  • API docs (docs/api/embeddings.md) and usage examples (examples/tools/embedding_example.py)
  • 37 unit + integration tests covering models, interfaces, provider, service, and agent integration

[0.0.33] - 2026-03-05

Changed

  • SearchResponse.answer type (AS-1692): Updated answer field from str to Union[str, List[Dict[str, Any]]] to support structured data responses

[0.0.31] - 2026-02-25

Fixed

  • get_system_prompt() parameter naming (AE-1511): Renamed misleading agent_code parameter to prompt_code in get_system_prompt() for clarity
  • The function fetches prompt templates, not agent configs — the parameter name now reflects this
  • Updated all call sites in examples/talk_to_document.py to use prompt_code=

  • Fixed 6 failing tests in TestGetSystemPrompt that used the old keyword argument

  • Pricing service bug (AS-1511): Fixed pricing service bug in model pricing lookup

  • Resolved incorrect pricing cache behavior
  • Updated unit tests to cover the fix

Removed

  • Dead code cleanup: Removed unused classes and functions from examples/talk_to_document.py
  • Removed ChatAgentLLMService class (was defined but never called)
  • Removed create_agent_from_params() function (was defined but never called)
  • The example now exclusively uses get_agent() to build agents from DynamoDB config

Enhanced

  • Documentation: Optimised docs for vectorisation (AS-1511)

  • Improved documentation structure for better searchability and indexing

  • Architecture Presentation: Added 6 new detailed slides to the SDK presentation (preso/)

  • Slide 22: Architectural Patterns Overview — 12 design patterns across 4 categories (Creational, Structural, Behavioral, Architectural)
  • Slide 23: Layered Architecture Deep Dive — 6 layers with directories, responsibilities, and key classes
  • Slide 24: Complete Module & File Map — every module and file with specific class names
  • Slide 25: Services & Infrastructure Detail — all 6 service domains with feature breakdown
  • Slide 26: Design Patterns in Code — Builder, Factory, Strategy, DI with real SDK code examples

  • Slide 27: Data Flow & Module Dependencies — request flow diagram and dependency rules

  • Security policy (FS-1155): Added security policy tests

Technical Details

  • akordi_agents/utils/agent.py: get_system_prompt(prompt_code=, user_persona=) — renamed parameter
  • Total: 1114 tests passing, 81.03% overall coverage
  • Presentation expanded from 22 to 28 slides

[0.0.30] - 2026-02-18

Added

  • Agent Utilities Module: New akordi_agents/utils/agent.py with reusable functions
  • get_system_prompt(): Fetch system prompt templates from DynamoDB with caching
  • get_agent_by_code(): Retrieve active agent configurations from AKORDI_AGENT_TABLE
  • get_agent(): Build CustomAgent from DynamoDB agent records

  • _create_default_llm_service(): Helper for creating AWSBedrockService-backed LLM services

  • LangGraphAgent Token Tracking: Added token usage tracking to LangGraphAgent

  • _get_token_service(): Lazy initialization of TokenUsageService
  • _track_token_usage(): Track token usage from workflow responses
  • _skip_token_tracking flag propagation through workflow metadata
  • Prevents double-tracking when AWSBedrockService is used internally

Enhanced

  • Test Coverage: Improved from 77.72% to 81.03%
  • Added 46 new tests across multiple modules
  • New test file: tests/test_agent_utils.py with 22 tests
  • Added 12 tests to tests/test_agent_builder.py

  • Added 12 tests to tests/test_model_service.py for pricing methods

  • Model Service: Added tests for pricing cache and model pricing lookup

  • _load_pricing_cache(): Load and cache model pricing from DynamoDB
  • get_model_pricing(): Look up input/output token pricing by model ARN
  • _extract_pricing(): Extract pricing from DynamoDB model records

Fixed

  • Module Import Performance: Fixed module-level AWS client initialization in llm_service.py
  • Changed _llm_service = LLMServiceFactory.create_service() to lazy initialization
  • Added _get_default_llm_service() function for on-demand service creation

  • Prevents AWS client creation at import time

  • Test Execution: Resolved test collection errors caused by AWS credential requirements

  • Tests no longer require valid AWS credentials during import

Technical Details

  • akordi_agents/utils/agent.py: 88 statements, 100% test coverage
  • akordi_agents/services/model_service.py: 98.45% test coverage (up from 70.47%)
  • akordi_agents/core/agent_builder.py: 48.72% test coverage (up from 45.09%)
  • Total: 1114 tests passing, 81.03% overall coverage

[0.0.28] - 2026-02-16

Added

  • Configurable Context Results Limit (AS-1639): New context_results_limit parameter in ClientParams
  • Controls how many search results are included in the LLM context
  • Default value of 50 (previously hardcoded to 3)
  • Allows fine-tuning the balance between context comprehensiveness and token usage
  • Separate from max_results which controls knowledge base retrieval count

Enhanced

  • Search Handler: Updated prepare_context() method to accept configurable limit
  • Method signature: prepare_context(search_results, context_results_limit=50)

  • Improved logging to show actual number of results used in context

  • Documentation: Comprehensive updates across multiple documentation files

  • Added context_results_limit parameter to docs/getting-started/configuration.md
  • Updated RAG examples in docs/examples/rag-document-qa.md
  • Enhanced akordi_agents/handlers/README.md with ClientParams configuration section
  • Added explanation of max_results vs context_results_limit distinction

Fixed

  • WebSocketLogger Recursion Error: Fixed infinite recursion in demo backend logging
  • Added recursion guard (_emitting flag) to prevent re-entry during log emission
  • Implemented message filtering to skip self-referential log messages
  • Fixed asyncio.create_task() usage in synchronous context
  • Replaced error logging with silent error handling to prevent recursion cascade

Technical Details

  • ClientParams.context_results_limit: Optional[int] = 50
  • The parameter is passed through query_vector_store() to prepare_context()
  • Test coverage added for new parameter in test_search_handler.py and test_general.py

[0.0.21] - 2026-01-14

Fixed

  • LLMResponse Validation Error: Fixed Pydantic validation error where usage.cache_creation expected an integer but received a dictionary

  • Changed usage: Dict[str, int] to usage: Dict[str, Any] in LLMResponse model to accommodate nested cache token structures

  • Knowledge Base Search Results: Fixed issue where KB search was not returning document content

  • Corrected S3 URI construction in search_handler.py - Bedrock stores URIs without URL encoding
  • Fixed {context} placeholder replacement in system prompts within llm_service.py
  • Added {messages} placeholder handling for chat history injection

  • Added clear header [DOCUMENT CONTENT RETRIEVED: X relevant excerpts] to injected KB content

  • Empty search_results in Response: Fixed propagation of search_results through the response chain

  • Added search_results: List[Any] = [] field to LLMResponse model
  • Updated invoke_model to include formatted knowledge_base_results in response

  • Fixed search_results extraction in agent_builder.py and example scripts

  • Stale Chat History: Fixed issue where scripts were using old chat history from DynamoDB

  • Changed default chat_id and user_id to empty strings to prevent fetching stale history

  • Duplicate enable_prompt_caching Argument: Fixed multiple values error in several example scripts

  • Added enable_prompt_caching to excluded keys in talk_to_document.py, simple_tool.py, lang_tool.py, and agent_to_agent_flow.py

  • Linting Errors: Fixed F541 and E501 errors across example scripts

  • Removed unnecessary f-string prefixes where no placeholders were used

  • Split long S3 path strings into multi-line format using implicit concatenation

  • Claude 3.7+ Model Support (AS-1261): Fixed exclusive parameter selection for Claude 3.7+ models

  • Enhanced parameter handling in llm_service.py to correctly manage model-specific features

  • Improved compatibility with latest Claude model versions

  • Prompt Caching Support (AS-1261): Fixed system prompt caching for models that support it

  • Implemented proper caching logic for compatible Claude models
  • Optimized token usage through intelligent prompt caching

Enhanced

  • Non-LangGraph Agent Examples: Refactored example scripts to demonstrate direct agent building without LangGraph
  • talk_to_document.py: Uses CustomAgent with direct KB search via ChatAgentLLMService
  • simple_tool.py: Implements manual tool orchestration loop with SimpleToolLLMService

  • Added accumulated token tracking across multiple LLM calls in tool loops

  • Performance Optimizations: Improved execution time for document interaction scripts

  • Implemented caching for system prompts, LLM service instances, and agent objects
  • Reduced DynamoDB calls by limiting chat history fetch to last 10 messages

  • Added skip_save flag to conditionally skip DynamoDB save operations

  • Response Metadata: Enhanced response structure with comprehensive metadata

  • Added metadata object containing query, knowledge_base details, config, and workflow info

  • Included workflow.type ("direct" or "langgraph") and agent_type in responses

  • Model ID Propagation (AS-1261): Enhanced model_id passing through the SDK

  • Updated search_handler.py, types.py, and llm_service.py for consistent model tracking
  • Improved model identification across the agent workflow

Changed

  • Test Configuration (AS-1261): Updated pytest configuration for async test support
  • Added pytest-asyncio>=0.21.0 dependency
  • Configured asyncio marker in pytest settings
  • Updated pyproject.toml and poetry.lock for async test compatibility

Added

  • LangGraph Talk-to-Document Example: New langgraph_ttd.py example demonstrating LangGraph-based document interaction
  • Full LangGraph workflow with validation, search, and response generation nodes

  • Configurable workflow options for tools, tracing, and validation

  • LLM Service Test Coverage (AS-1261): Added comprehensive test suite for Claude 3.7+ features

  • 17 new tests covering exclusive sampling parameters and prompt caching logic
  • Validates Claude 3.7+ exclusive temperature/top_p handling
  • Tests prompt caching for models 3.5+ with intelligent enabling/disabling
  • Improved llm_service.py coverage from 59% to 64%
  • Automatic Token Usage Tracking: Comprehensive token consumption monitoring for all agents
  • Agent-level aggregated statistics (total tokens, LLM calls, model, status)
  • Per-query token usage tracking with detailed metadata
  • Automatic tracking without code changes - just configure environment
  • DynamoDB storage with atomic updates for statistics
  • New TokenUsageService for querying and managing token data
  • Data models: AgentTokenUsage and QueryTokenUsage
  • Integration in both CustomAgent and LangGraphAgent
  • Cost tracking field (set to $0.00 as per requirements)

  • Query metadata: user_id, chat_id, query_text, tools_used, success/error status

  • Token Usage Tools and Examples:

  • create_token_usage_table.py: DynamoDB table creation script
  • token_usage_tracking_example.py: Complete working example
  • TOKEN_USAGE_TRACKING.md: Comprehensive documentation with use cases
  • Updated README.md with token tracking quick start guide

Enhanced

  • Agent Builder: Added _track_token_usage() method to both agent classes
  • Automatic initialization of agent usage statistics
  • Per-query recording with aggregated updates
  • Non-blocking error handling - tracking failures don't affect agent responses
  • Graceful degradation when tracking is disabled

[0.0.20] - 2025-11-19

Added

  • Comprehensive Test Suite Overhaul: Complete conversion from unittest to pytest framework
  • Migrated all 58 tests to use pytest with proper fixtures and assertions
  • Enhanced test mocking with pytest-mock for better isolation

  • Improved test discoverability and execution speed

  • Bedrock Guardrail Test Coverage: Extensive unit test coverage for AWS Bedrock Guardrail functionality

  • 20 comprehensive tests covering all guardrail creation, versioning, and configuration scenarios
  • Regional PII configuration testing (US, UK, Australia, Canada)
  • Parameter validation and error handling tests

  • Function call verification and return type validation

  • Interactive Web Demo: Complete web interface for agent-to-agent orchestration visualization

  • FastAPI backend server with WebSocket real-time streaming
  • React frontend with TypeScript for live agent execution monitoring
  • Dynamic agent selection visualization based on LLM analysis
  • Real-time progress tracking and status updates
  • Automated startup script for both backend and frontend servers

Enhanced

  • Test Framework: Upgraded from unittest to pytest with modern testing patterns
  • Better fixture management and test organization
  • Improved error reporting and debugging capabilities

  • Enhanced test parallelization support

  • Test Mocking Strategy: Implemented proper dependency isolation

  • Module-level mocking for external dependencies (boto3, pydantic, etc.)
  • Clean separation between unit tests and integration tests

  • Reduced test flakiness and improved reliability

  • CI/CD Integration: Enhanced test execution in Poetry environment

  • All tests now run via poetry run pytest
  • Improved test configuration for automated pipelines
  • Better test output formatting and reporting

Fixed

  • Test Execution Issues: Resolved all 58 test failures across three test modules
  • Fixed boto3 client mocking conflicts in bedrock guardrail tests
  • Corrected AWS configuration parameter assertions in provider config tests
  • Updated exception handling expectations in service tests

  • Resolved import path issues and module isolation problems

  • Mocking Conflicts: Eliminated complex module-level mocking that caused test instability

  • Implemented clean test isolation using pytest fixtures
  • Fixed logger assertion issues with multiple call scenarios

  • Corrected AWS client parameter matching in integration tests

  • CI/CD Pipeline Dependency Issue: Fixed pytest-mock fixture not available in GitLab pipelines

  • Moved pytest-mock from test group to main dependencies to ensure availability in all environments
  • Resolved "fixture 'mocker' not found" errors in CI pipelines while maintaining local functionality

Technical Details

  • Test suite now uses pytest fixtures instead of unittest setUp/tearDown methods
  • Enhanced mocking with pytest-mock for better control over external dependencies
  • All tests pass in isolated environments with proper dependency injection
  • Improved test execution time from ~30s to ~0.17s for 58 tests
  • Comprehensive coverage of bedrock guardrail API interactions and error scenarios

Changed

  • Test Structure: Complete migration from unittest.TestCase to pytest classes
  • Replaced self.assertEqual() with assert statements
  • Updated exception testing from assertRaises to pytest.raises
  • Modernized test organization with descriptive test method names

[0.0.18] - 2025-10-30

Added

  • LangGraph Integration: Complete LangGraph workflow orchestration system
  • Advanced state management with typed AgentState
  • Conditional routing and dynamic workflow execution
  • Multi-agent coordination capabilities
  • Streaming response support foundation

  • Enhanced observability with LangSmith tracing

  • LangGraph Core Components:

  • AgentState: Comprehensive state management for complex workflows
  • WorkflowConfig: Configurable workflow behavior and parameters
  • BaseAgentWorkflow: Extensible base class for custom workflows
  • ToolUseWorkflow: Intelligent tool selection and execution

  • MultiAgentWorkflow: Multi-agent coordination system

  • LangGraph Nodes: Modular workflow components

  • ValidationNode: Input validation with error handling
  • SearchNode: Knowledge base search integration
  • ToolDecisionNode: AI-powered tool usage decisions
  • ToolExecutionNode: Safe tool execution with error recovery
  • ResponseGenerationNode: Context-aware response generation
  • CoordinatorNode: Multi-agent coordination logic

  • AggregationNode: Result aggregation from multiple agents

  • LangGraph-Enhanced Examples:

  • langgraph_enhanced_agent.py: Comprehensive demonstration of LangGraph capabilities
  • Multi-step reasoning workflows with tool integration
  • Conditional branching based on validation and tool decisions
  • Real-time weather data integration with risk assessment

Enhanced

  • Dependencies: Added langgraph>=0.2.0 and langchain-core>=0.3.0
  • Documentation: Updated README.md and examples/README.md with LangGraph usage
  • Architecture: Enhanced agent builder to support LangGraph workflows
  • Tool Framework: Improved integration with LangGraph state management

Technical Details

  • LangGraph workflows support complex agent interactions with state persistence
  • Conditional edges enable dynamic routing based on validation results and tool decisions
  • Tool execution includes comprehensive error handling and recovery
  • Multi-agent coordination allows specialized agents to collaborate on complex tasks
  • Enhanced logging and tracing for workflow observability
  • Streaming support foundation for real-time response generation

[0.0.17] - 2025-10-27

Added

  • WeatherTool Integration: New WeatherTool class in akordi_agents.tools that integrates with WeatherAPI.com
  • Real-time weather data fetching for any city worldwide
  • Defaults to Auckland, New Zealand when no city specified
  • Comprehensive weather information including temperature, humidity, wind, pressure, UV index
  • Intelligent query detection in RiskDescribeService for conditional tool usage
  • Weather queries automatically use WeatherTool, risk queries use LLM
  • Proper error handling and user-friendly formatted responses

  • Example implementation in examples/tool_simple.py with integrated agent

  • HTTP Client Dependency: Added requests>=2.25.0 dependency for external API integrations

Enhanced

  • Tool Framework: Improved Tool abstract base class with better documentation
  • Example Documentation: Updated examples/README.md with WeatherAPI setup instructions
  • Environment Configuration: Added WEATHER_API_KEY requirement documentation

Technical Details

  • WeatherTool uses WeatherAPI.com v1 current weather endpoint
  • Supports both Celsius and Fahrenheit temperature display
  • Includes wind speed in both km/h and mph
  • Provides formatted markdown responses for better readability
  • 10-second timeout for API requests with comprehensive error handling

Removed

  • Unused Abstract Methods: Removed get_supported_models() and is_model_supported() abstract methods from BaseLLMService interface
  • These methods were raising NotImplementedError in AWSBedrockService and were not being used
  • Removed corresponding _supported_models attribute initialization
  • Cleaned up model support logic that was not implemented

Fixed

  • Test Suite Improvements: Fixed pytest execution issues
  • Removed complex module-level mocking in conftest.py that was causing test failures
  • Simplified test configuration for better maintainability
  • All 133 tests now pass successfully
  • Fixed import path for ConfigManager in test fixtures

Changed

  • Test Coverage: Updated test suite to reflect removed methods
  • Removed tests for get_supported_models() and is_model_supported()
  • Updated test_model_support_completeness() to validate service initialization instead
  • Removed assertions checking for _supported_models attribute

Technical Details

  • Simplified AWSBedrockService.__init__() by removing unused model list initialization
  • Cleaned up abstract base class to only include actively used methods
  • Improved test configuration maintainability by removing mock complexity

[0.0.15] - 2025-09-19

Added

  • Initial Release: Complete Akordi Agents SDK framework
  • Core Agent Framework: Comprehensive agent building capabilities with modular architecture
  • LLM Integration: Support for multiple LLM providers with configurable services
  • Chat History Management: Full chat history persistence using DynamoDB
  • Knowledge Base Search: Advanced search handlers for document and content retrieval
  • Command-Line Interface (CLI): Interactive chat interface with multiple modes
  • Interactive chat with persistent history
  • Single query mode for quick questions
  • Knowledge base enhanced responses
  • Configuration management interface
  • Configuration Management: Flexible prompt and LLM provider configuration
  • DynamoDB-based prompt configuration
  • Environment-based provider settings
  • Langsmith integration for monitoring
  • Type Safety: Comprehensive Pydantic models for all data structures
  • Validation Utilities: Robust input validation and error handling
  • Logging System: Configurable logging with multiple output formats
  • Response Processing: Advanced response formatting and streaming capabilities
  • Development Tools: Complete development environment setup
  • Poetry-based dependency management
  • Comprehensive test suite with pytest
  • Code quality tools (black, isort, flake8, mypy)
  • Pre-commit hooks and CI/CD support

Features

  • Agent Builder: Modular agent construction with pluggable components
  • Multi-Provider LLM Support: AWS Bedrock, OpenAI, and custom provider integration
  • Persistent Chat Sessions: DynamoDB-backed conversation history
  • Document Search Integration: Knowledge base queries with relevance scoring
  • Interactive CLI: Rich command-line interface with auto-completion
  • Configuration Flexibility: Environment-based and file-based configuration
  • Error Handling: Comprehensive error handling and recovery mechanisms
  • Async Support: Asynchronous operations for improved performance
  • Streaming Responses: Real-time response streaming for better user experience

Development

  • Testing Framework: Extensive unit and integration test coverage
  • Documentation: Comprehensive README with installation and usage guides
  • Example Scripts: Working examples for common use cases
  • CI/CD Pipeline: Automated testing and deployment workflows
  • Code Quality: Linting, formatting, and type checking integration

Dependencies

  • Core Dependencies:
  • boto3>=1.34.0 - AWS SDK for DynamoDB and Bedrock integration
  • python-dotenv>=1.0.0 - Environment variable management
  • langsmith>=0.1.0 - LLM monitoring and tracing
  • pydantic>=2.0.0 - Data validation and type safety
  • Development Dependencies:
  • pytest>=7.0.0 - Testing framework
  • black>=23.0.0 - Code formatting
  • isort>=5.12.0 - Import sorting
  • flake8>=6.0.0 - Linting
  • mypy>=1.0.0 - Type checking

Installation

  • PyPI Package: pip install akordi-agents
  • Development Mode: pip install -e .
  • Poetry Support: Full Poetry integration with dependency management
  • CLI Installation: Automatic CLI script installation (akordi-cli)

Documentation

  • Complete installation guide for multiple environments
  • CLI usage documentation with examples
  • API reference and integration examples
  • Troubleshooting guides for common issues
  • Development workflow documentation

Types of Changes

  • Added for new features
  • Changed for changes in existing functionality
  • Deprecated for soon-to-be removed features
  • Removed for now removed features
  • Fixed for any bug fixes
  • Security in case of vulnerabilities

Contributing

This project follows semantic versioning. For major changes, please open an issue first to discuss what you would like to change.

For older changes, please refer to the git commit history.