Changelog

Note

Release History and Updates

This changelog documents all notable changes to the HoneyHive Python SDK. For the complete, up-to-date changelog, see the CHANGELOG.md file in the repository.

Important

Format: This project follows Keep a Changelog format and adheres to Semantic Versioning.

Latest Release Notes

For the complete and always up-to-date changelog, see: CHANGELOG.md

Current Version Highlights

🐛 FIXED: Metrics Table Printing Empty Values (Unreleased)

  • Root Cause: SDK expected metrics as dynamic top-level keys but backend returns them in a details array format per OpenAPI spec

  • New Typed Models: Added MetricDetail, DatapointResult, DatapointMetric, MetricDatapoints Pydantic models with proper type hints

  • Updated AggregatedMetrics: Now uses details: List[MetricDetail] field to match backend response format

  • Updated print_table(): Iterates over details array and extracts aggregate values from MetricDetail objects

  • Backward Compatible: get_metric(), list_metrics(), get_all_metrics() support both new details array format and legacy model_extra format

  • Test Coverage: Added 19 unit tests for typed models and integration test against real API

  • Files: src/honeyhive/experiments/models.py, src/honeyhive/experiments/results.py

🐛 FIXED: enrich_session() Metadata Requirement (Unreleased)

  • Flexible Enrichment: enrich_session() now works correctly even when no parameters (metadata, inputs, outputs, etc.) are explicitly provided

  • Bug Fix: Previously, calling enrich_session() without parameters would silently fail due to empty update_params check

  • Improved Behavior: Session enrichment now proceeds when a valid session_id is available, regardless of whether enrichment data is provided

  • Files: src/honeyhive/tracer/core/context.py

🐛 FIXED: Session ID Initialization (Dec 2025)

  • Backend Sync: Sessions are now always initialized in backend, even when session_id is explicitly provided

  • Bug Prevention: Prevents backend bug where inputs/outputs would be auto-populated from the first event if session didn’t exist

  • UUID Validation: Added automatic UUID format validation with fallback to new UUID if invalid

  • Session Preservation: Properly preserves explicitly provided session IDs through the session creation process

✨ NEW: Enhanced API Schemas & Filtering (Dec 2, 2025)

  • Metric Schema Updates: Updated to match backend with new enum values (PYTHON/LLM/HUMAN/COMPOSITE), added sampling_percentage, scale, categories, filters, timestamps

  • Datapoints Filtering: Added dataset_id and dataset_name parameters, removed limit parameter (not in API spec), auto-detects ID vs name for legacy dataset parameter

  • EventFilter Convenience: list_events() now accepts single EventFilter or List[EventFilter] with automatic conversion

  • UUIDType Display: Added __str__ and __repr__ methods for proper string representation in logs

  • Metric Deletion: Now raises AuthenticationError - users must use webapp to delete metrics

  • Test Fixes: Fixed all unit test failures after model regeneration (enum name changes, UUID validation, AsyncMock issues)

  • Integration Test Fixes: Updated backend_verification.py to use event_filters parameter, removed limit from list_datapoints calls, fixed invalid UUID strings

  • Files Modified: 17 files (+492/-503 lines) with comprehensive test coverage

🛡️ NEW: Configurable Span Limits & Core Attribute Preservation (Nov 18, 2025)

  • Lazy-Activated Preservation: Automatically preserves critical attributes (session_id, event_type, event_name, source) to prevent data loss when spans exceed attribute limits

  • Performance Optimized: Only triggers for large spans (95%+ of limit), <0.001ms overhead for normal spans, ~0.5ms for large spans

  • Configurable Limits: New span limit controls - max_attributes (1024, up from OTel’s 128), max_events (1024), max_links (128), max_span_size (10MB)

  • Zero Configuration: Works out of the box with sane defaults, fully configurable via TracerConfig or environment variables

  • Data Safety: Prevents span rejection by backend when critical attributes are evicted by OpenTelemetry’s FIFO policy

🚀 INFRA: praxis OS Migration & Bug Fixes (Nov 14, 2025)

  • ✨ NEW: Pretty Table Output for Evaluations: Added beautiful terminal table display for evaluate() results with color, emojis, and formatted metrics

  • AI Development Framework: Migrated from Agent OS to praxis OS with MCP (Model Context Protocol) integration

  • Enhanced Tooling: Added multi-repo code intelligence, advanced RAG search, and phase-gated workflows

  • Bug Fix: Completed praxis OS pre-commit migration - fixed all hooks to use new .praxis-os/ paths (10 files, 43 references updated)

  • Bug Fix: Fixed enrich_session inputs parameter causing 400 errors - now maps unsupported fields to metadata

  • Bug Fix: Fixed OpenInference event type detection - ensures correct classification of instrumented spans (CHAIN, LLM, TOOL, etc.)

  • Bug Fix: Enhanced error logging for 400 errors in experiment runs for better debugging

  • Bug Fix: Corrected user_properties and metrics handling in enrich_span/enrich_session methods

  • Testing: Added Google ADK instrumentation exercise script with rate limiting, callbacks, and comprehensive test scenarios

  • Breaking Change (Dev Only): AI development workflows now require praxis OS installation

✨ NEW: DatasetsAPI Filtering - Find Datasets Efficiently (Nov 10, 2025)

  • Server-Side Filtering: Find datasets by name, type, or ID without fetching all datasets

  • Performance: Much faster for large projects with 100+ datasets

  • New Parameters: name, dataset_type, dataset_id, include_datapoints

  • Backward Compatible: All parameters optional, existing code works unchanged

  • Customer Request: Addresses scalability concerns as projects grow

📚 IMPROVED: Strands Integration - Best Practices Pattern (Nov 6, 2025)

  • Instance Method Pattern: All examples now use tracer.enrich_span() instead of global enrich_span()

  • Multi-Instance Safety: Explicit tracer references work reliably in all environments

  • Future-Proof: Avoids global function that will be deprecated in v2.0

  • Best Practices: Documentation showcases recommended v1.0+ patterns

  • Explicit Context: All @trace decorators include explicit tracer=tracer parameter

🔧 NEW: Manual PyPI Publishing for Release Candidates (Nov 6, 2025)

  • Manual Trigger: Added workflow_dispatch to PyPI publishing workflow

  • RC Testing: Can now publish release candidates (e.g., 1.0.0-rc3) from any branch

  • Pre-Merge Testing: Enables user testing of RCs before merging to main

  • Automated: Still performs all validation, integrity checks, and creates GitHub releases

  • Fixed: Version extraction now uses sed to avoid Python import errors

📚 UPDATED: AWS Strands Documentation with Current Model IDs (Nov 6, 2025)

  • Version Bump: Updated to 1.0.0-rc3 to reflect stable API

  • Model Access: Clarified that AWS Bedrock models are now automatically available (no manual request)

  • Current Models: Replaced deprecated Claude 3 models with Claude 4.5 series (Haiku 4.5, Sonnet 4.5)

  • EULA Info: Added documentation about Anthropic EULA acceptance on first invocation

  • Verification: All updates verified against official AWS Bedrock documentation

✨ NEW: Automatic Span Capture for Evaluation Functions (Nov 3, 2025)

  • Auto-Decoration: User functions in evaluate() are now automatically wrapped with @trace decorator

  • Zero-Config Observability: Automatic span capture with inputs/outputs without manual decorator application

  • Event Type: Functions traced as “chain” type events for proper categorization

  • Transparent: Works seamlessly with both functions that accept tracer parameter and those that don’t

✨ NEW: v1.0 Evaluation Enhancements (Oct 31, 2025)

  • Smart Session Naming: Experiments now use experiment name as default session name

  • Tracer Injection: Auto-inject tracer parameter into evaluation functions for enrich_session() support

  • Ground Truth Tracking: Automatic ground truth capture in session feedback

  • Auto-Input Tracking: @trace decorator automatically captures function inputs (no manual enrichment needed)

  • Session Linking: Propagate run_id through OpenTelemetry baggage for correct span association

  • Backward Compatible: Functions without tracer parameter continue to work

  • New Tutorial: “Run Your First Experiment” with evaluators and result comparison

  • Test Coverage: 14 new tests with end-to-end backend verification

🐛 CRITICAL FIX: Config Priority Bug (Oct 30, 2025)

  • Issue: SessionConfig and EvaluationConfig values not promoted to root, hidden in nested configs

  • Root Cause: create_unified_config() didn’t implement field promotion logic

  • Solution: Added priority-aware promotion: individual params > SessionConfig > EvaluationConfig > TracerConfig

  • Impact: 15 colliding fields now work correctly (session_id, project, api_key, server_url, etc.)

  • Tests: Added 19 unit tests, 35 API integration tests, 10 backend verification tests

🐛 CRITICAL FIX: Evaluation Metadata Propagation to Child Spans (Nov 3, 2025)

  • Issue: Evaluation context (run_id, dataset_id, datapoint_id) not propagating from evaluate() to child spans created by @trace decorators

  • Root Cause: HoneyHiveSpanProcessor wasn’t reading evaluation-specific baggage keys

  • Solution: Added _get_evaluation_attributes_from_baggage() method to extract and apply evaluation metadata

  • Impact: All spans created during evaluate() datapoint processing now inherit evaluation context

  • Tests: Added 3 unit tests (all baggage scenarios) + 1 integration test for end-to-end validation

🚨 BREAKING CHANGE: Ground Truth Field Name Migration (Nov 3, 2025)

  • Breaking Change: Migrated from ground_truths (plural) to ground_truth (singular) throughout SDK

  • Critical Bug Fixed: Ground truth data was inaccessible to metrics, UI, and LLM evaluators

  • Root Cause: SDK sent feedback: {“ground_truths”: {…}} but backend expects feedback: {“ground_truth”: {…}}

  • Impact Before Fix: Metrics with needs_ground_truth=true failed, UI couldn’t display ground truth, LLM evaluators couldn’t access data

  • Migration Required:

    • Dataset format: Change “ground_truths” → “ground_truth” in all datasets

    • Evaluator signatures: Change ground_truths parameter → ground_truth parameter

  • Before: dataset = [{“inputs”: {…}, “ground_truths”: {…}}]

  • After: dataset = [{“inputs”: {…}, “ground_truth”: {…}}]

  • Migration Time: 15 minutes to 2 hours (simple find-replace operation)

  • Benefits: Fixes broken metrics, enables UI display, enables LLM evaluator access, aligns with backend API and industry standards

  • Files Updated: 15 files (1 source, 4 tests, 9 docs, 1 example) with 322 total line changes

✨ NEW: Instance Method Pattern for Span/Session Enrichment (v1.0)

  • Primary API: tracer.enrich_span() and tracer.enrich_session() instance methods

  • Backward Compatible: Free functions still work but deprecated

  • Multi-Instance Safe: Proper tracer discovery via baggage propagation

  • Comprehensive Examples: Updated all examples with new patterns

🐛 CRITICAL FIX: Multi-Instance Context Isolation (Oct 29, 2025)

  • Issue: project and source leaked between tracer instances via global baggage

  • Root Cause: project/source were in SAFE_PROPAGATION_KEYS, causing context pollution

  • Solution: Removed from safe keys, prioritize tracer instance values in span processor

  • Result: Each tracer instance maintains isolated context in multi-instance scenarios

🐛 CRITICAL FIX: enrich_span() Immediate Execution (Oct 29, 2025)

  • Issue: enrich_span(metadata={…}) returned lazy object instead of executing

  • Root Cause: UnifiedEnrichSpan.__call__() deferred execution

  • Solution: Modified to immediately execute enrich_span_unified()

  • Result: Direct calls now work without context manager or boolean evaluation

🐛 FIX: Decorator API Parameter Handling (Oct 29, 2025)

  • Issue: @trace decorator passed span object to enrich_span_unified(), polluting spans

  • Solution: Removed erroneous span parameter from decorator enrichment calls

  • Result: Spans no longer contain honeyhive_metadata: “Span(…)” pollution

🐛 FIX: None Value Defense-in-Depth Filtering (Oct 29, 2025)

  • Issue: None values serialized to “null” strings in span attributes

  • Solution: Two-layer filtering at decorator and attribute-setting levels

  • Result: Spans no longer polluted with “null” string values

🐛 CRITICAL FIX: evaluate() + enrich_span() Pattern

  • Issue: Span enrichment failed in evaluation workflows

  • Root Cause: Baggage propagation was disabled to avoid session conflicts

  • Solution: Selective baggage with safe keys (updated Oct 29: removed project/source)

  • Result: Tracer discovery works while preventing multi-instance conflicts

🧪 ADDED: Nested enrich_span() Backend Validation

  • Comprehensive Test: Validates nested function calls with enrich_span() in evaluate() workflows

  • Backend Verification: Confirms enriched properties (metadata, metrics, config, feedback) persist

  • Pattern Coverage: Parent function → nested helper function enrichment

  • Real Fixtures: Uses real_project and integration_client for accurate validation

  • Zero False Positives: CRITICAL assertions fail if enrichment not found in backend

📚 ADDED: Strands Multi-Agent Integration Examples

  • Swarm Collaboration: Comprehensive example with researcher → coder → reviewer flow

  • Graph Workflow: Parallel processing pattern with research → analysis/fact_check → report

  • Advanced Patterns: Entry points, max handoffs/iterations, execution timeouts, node timeouts

  • Tracing Support: Expected spans, agent collaboration flow, and agent-level metrics documented

📋 ADDED: Integration Examples Requirements File

  • Comprehensive Dependencies: Added requirements.txt with all packages for integration examples

  • Organized by Category: Core, LLM providers, OpenInference instrumentors, Traceloop instrumentors, and agent frameworks

  • Installation Commands: Per-integration pip install commands for easy setup

  • Environment Variables: Documentation of required credentials for each provider

📚 ADDED: New Example Files

  • Evaluation Example: Simple demonstration of the evaluate() function with dataset evaluation and span enrichment

  • Legacy SDK Example: Reference example showing basic tracer initialization and OpenAI integration

🔧 FIXED: Session Enrichment in evaluate() Function

  • Always Enriches Sessions: Fixed bug where sessions weren’t enriched when no evaluators were provided

  • Output Persistence: Ensures outputs are always saved to backend regardless of evaluator presence

  • Better Logging: Upgraded log level from debug to info for session enrichment visibility

🔧 IMPROVED: Tracer Internal Cleanup

  • Code Simplification: Removed redundant experiment baggage code path

  • No User Impact: Experiment tracking continues to work exactly as before

  • Performance: Simplified baggage discovery logic

🔧 FIXED: enrich_session() Backwards Compatibility Restored

  • Legacy Parameters: Restored session_id as optional positional parameter and user_properties support

  • Automatic Conversion: User properties automatically merged into metadata with user_properties. prefix

  • Comprehensive Documentation: Added 685-line documentation guide with 15+ examples and 5 common patterns

  • API Reference: Complete function signature documentation with backwards compatibility examples

  • Regression Tests: Added tests for legacy positional arguments and user_properties handling

🔧 FIXED: enrich_span() Dynamic Tracer Discovery

  • Automatic Resolution: Added tracer discovery when not explicitly provided via tracer_instance

  • Priority-Based: Explicit parameter → baggage context → global default tracer

  • Multi-Instance Safe: Ensures correct tracer in multi-tracer applications

  • Regression Tests: Added tests for auto-discovery, explicit tracer priority, and graceful degradation

🔧 FIXED: Integration Examples Bug Fixes

  • Google ADK: Fixed LoopAgent parameter name (sub_agent → agent), disabled parallel workflow test

  • Strands: Removed redundant global TracerProvider setting

  • Documentation: Enhanced README with expanded links to all integration guides organized by category

🔧 FIXED: enrich_span() Backwards Compatibility Restored

  • Original Interface Restored: Fixed enrich_span() to support main branch’s reserved namespaces (metadata, metrics, feedback, inputs, outputs, config, error, event_id)

  • New Patterns Added: Simple dictionary (routes to metadata), arbitrary kwargs (routes to metadata), and context manager support

  • Circular Import Resolved: Extracted _set_span_attributes() to new span_utils.py module

  • 100% Test Coverage: Added 48 unit tests + 3 integration tests with backend verification

  • Documentation Updated: Comprehensive updates to tutorials, how-to guides, and API reference with new examples

🧪 NEW: Span Capture and Test Case Generation

  • Span Recording: Capture OpenTelemetry spans during integration runs

  • Test Generation: Convert captured spans to unit test cases

  • Provider Coverage: Generate tests for AutoGen, Google ADK, Semantic Kernel

  • Environment Flag: Enable via CAPTURE_SPANS=true

  • Automated Workflow: Complete guide for test case generation

📚 NEW: AutoGen Integration Example

  • Two-Agent Conversations: User proxy and assistant agent collaboration

  • Group Chat: Multiple specialized agents (writer, critic, planner)

  • Sequential Chat: State-based transitions between agents

  • Nested Chat: Complex task decomposition with agent hierarchies

  • Code Execution: Automatic Docker-based code execution

  • Tool Registration: Function calling with custom tools

📚 NEW: DSPy Integration Example

  • Signatures: Declarative task definitions with input/output specifications

  • Chain of Thought: CoT reasoning with assertions and validation

  • ReAct Pattern: Agent-based reasoning with tool use

  • Optimization: BootstrapFewShot for program optimization

  • Multi-Hop Reasoning: Retrieve-then-read patterns for complex queries

📚 NEW: AWS Bedrock Direct Integration Example

  • Multi-Model Support: Amazon Nova, Titan Text, and Anthropic Claude models

  • Converse API: Unified interface for all Bedrock models

  • Streaming: ConverseStream API for real-time responses

  • Document Understanding: PDF, TXT, and DOC format support

  • Flexible Auth: Multiple authentication methods (keys, session tokens, IAM roles)

📚 NEW: Pydantic AI Integration Example

  • Type-Safe Agents: Complete Pydantic AI integration with structured outputs

  • Agent Tools: Demonstrates @agent.tool decorator for function calling

  • Dynamic Prompts: System prompt generation with @agent.system_prompt

  • Dependency Injection: RunContext for passing dependencies to agents

  • Streaming Support: Async iteration for streaming responses

📚 NEW: LangGraph Integration Example

  • State Graph Workflows: Complete LangGraph integration with sequential node execution

  • Conditional Routing: Demonstrates dynamic routing based on graph state

  • Multi-Step Agents: Agent graphs with state management across nodes

  • Node Tracing: Node-level tracing with @trace decorator integration

  • Automatic Instrumentation: LangChain call tracing via OpenInference

🔍 NEW: Raw Span Data Dumping for Debugging

  • Comprehensive Span Extraction: New _dump_raw_span_data() method captures all OpenTelemetry span properties

  • Full Context Capture: Includes trace_id, span_id, parent spans, status, attributes, events, links

  • Resource Information: Captures resource attributes and instrumentation info for complete observability

  • JSON Formatting: Outputs pretty-printed JSON for easy debugging and troubleshooting

🔧 CHANGED: Enhanced evaluate() Environment Variable Support

  • Optional API Key: api_key parameter now optional, reads from environment variables

  • Server URL Support: Added server_url parameter with env var support

  • Dual Prefix Support: Accepts both HONEYHIVE_* and HH_* environment variable prefixes

  • Better UX: More flexible configuration without hardcoding credentials

🔄 CHANGED: Updated Google ADK Integration with Async Support

  • Modern API: Updated to newer Google ADK API with LlmAgent, Runner, and InMemorySessionService

  • Async/Await: Added full async support to all test functions for better performance

  • Simplified Auth: Migrated from GOOGLE_ADK_API_KEY to standard GOOGLE_API_KEY environment variable

  • Session Management: Improved session handling with explicit session service

🔄 CHANGED: Refactored Strands Integration Example

  • TracerProvider Pattern: Updated AWS Strands integration to use recommended tracing pattern

  • 6 Focused Test Cases: Replaced complex workflow with targeted tests (basic invocation, tools, streaming, etc.)

  • AWS Bedrock Integration: Switched from OpenAI to AWS Bedrock model implementation

  • Comprehensive Documentation: Added detailed tracing expectations and GenAI semantic conventions

🔧 NEW: MCP Server Upgrade (v0.1.0rc3)

  • Agent OS Enhanced Architecture: Upgraded from prototype to modular product architecture (+5,823 lines)

  • Workflow Engine: Phase gating with evidence validation for controlled AI development

  • File Watcher: Automatic incremental RAG index updates on content changes

  • Framework Generator: Create new AI-assisted workflows programmatically

  • FastMCP Integration: Modern server factory with automatic tool registration

📦 Version Refactoring: Single Source of Truth (v0.1.0rc3)

  • Consolidated Version Management: Reduced from 5 hardcoded locations to 1

  • Dynamic Imports: Late binding pattern following Agent OS standards

  • 80% Less Maintenance: Version updates now require editing only 1 file

  • MyPy Compliance: Fixed circular import errors with proper import strategy

📚 NEW: Restructured Evaluation Documentation

  • Modular How-To Guides: Created 9 focused problem-oriented guides following Divio Documentation System

  • Simplified Tutorial: Redesigned 04-evaluation-basics.rst as a true 15-minute introductory guide

  • Question-Based Format: All sections use questions as titles for better scannability (e.g., “How do I run experiments?”)

  • Clear Navigation: Updated index with toctree and quick links to common use cases

  • API Focus: All guides prioritize evaluate() function over decorator-based approach

🤖 NEW: Agent OS MCP/RAG Server (Dogfooding)

  • Model Context Protocol Integration: Complete MCP server implementation with 5 tools for AI-assisted development

  • 90% Context Reduction: RAG engine with LanceDB achieving semantic search over standards (50KB → 5KB)

  • Phase-Gated Workflows: Workflow engine enforcing controlled AI development with checkpoint validation

  • HoneyHive Tracing: Complete instrumentation with @trace decorators on all tools for observability dogfooding

  • Import Verification Standard: New “2-Minute Rule” preventing import path hallucination in AI-generated code

  • Quality Excellence: 28 unit tests with 10.0/10 Pylint score, full type annotations, and independent dependency management

Development Tools

  • Improved pre-commit checks for Agent OS spec proposals

v0.1.0+ (Development) - Major Architectural Refactor

🏗️ NEW: Modular Tracer Architecture

  • Mixin-Based Design: Complete rewrite with 6 core modules for better maintainability

  • Enhanced Multi-Instance: True isolation between tracer instances with independent configurations

  • OpenTelemetry Compliance: Full OTel standard adherence with enhanced provider strategies

  • 35 New Files: Comprehensive modular architecture across core, infra, instrumentation, integration, lifecycle, processing, and utils modules

🔧 NEW: Hybrid Configuration System

  • Type-Safe Config Objects: New Pydantic models (TracerConfig, SessionConfig, APIClientConfig, etc.)

  • Three Initialization Patterns: Traditional .init() (recommended), modern config objects, environment variables

  • 100% Backwards Compatible: All existing .init() usage continues to work unchanged

  • Dynamic Environment Mapping: Flexible environment variable configuration with AliasChoices

📚 NEW: Comprehensive Documentation

  • Complete Migration Guide: Zero-breaking-change upgrade paths with detailed examples

  • Architecture Reference: Mixin composition patterns and multi-instance scenarios

  • Enhanced Tutorials: Configuration patterns and best practices

  • API Reference Expansion: Full documentation for all new Pydantic models

🔧 QUALITY: Perfect Test Suite

  • 2,904 Total Tests: 2,735 unit + 169 integration tests with 100% pass rate

  • 94.13% Coverage: Significantly exceeds 80% requirement

  • 10.0/10 Pylint Score: Perfect code quality with 0 MyPy errors

  • Enhanced Performance Testing: Dynamic thresholds for parallel vs isolation execution

v0.1.0rc2 (Development) - Full Backwards Compatibility and Environment Variable Fixes

🔄 NEW: Complete Backwards Compatibility Implementation

  • All 16 Original Parameters: Complete parameter compatibility with main branch HoneyHiveTracer

  • Context Association Properties: Multi-tracer coordination support for complex deployments

  • Session ID Validation: UUID validation with proper error handling for session linking

  • Server URL Override: Custom deployment support with runtime URL configuration

  • Verbose Debug Control: Granular output control throughout tracer initialization

  • Evaluation Workflows: Full evaluation baggage support (run_id, dataset_id, datapoint_id)

  • Batch Processing Control: disable_batch parameter controls SimpleSpanProcessor vs BatchSpanProcessor

  • Git Metadata Collection: Automatic git information collection for session metadata

  • Context Propagation: Link/unlink/inject methods for carrier-based context propagation

  • Session Enhancement: Inputs and metadata support for enriched session creation

🔧 FIXED: Runtime Environment Variable Support

  • HH_API_URL Override: Environment variables now properly picked up when set at runtime

  • Boolean Variables: Fixed HH_VERIFY_SSL and HH_FOLLOW_REDIRECTS precedence logic

  • Fresh Config Loading: API client and tracer use fresh config instances

  • API Key Precedence: Fixed HH_API_KEY environment variable precedence over constructor parameters

  • HTTP Tracing Configuration: Fixed disable_http_tracing environment variable handling for multi-instance support

  • Comprehensive Testing: Added 17 backwards compatibility integration tests covering runtime behavior

⚡ BREAKING: Structured Logging Infrastructure

  • Production Ready: Replaced all print statements with structured HoneyHive logging

  • Better Observability: Structured logging with honeyhive_data for context

  • Proper Log Levels: Debug, info, warning, and error levels for appropriate output

  • Maintained Compatibility: Docstring examples still use print statements per Python conventions

🚀 NEW: Pre-commit Test Suite Execution

  • Zero Failing Tests Policy: Automated test execution in pre-commit hooks

  • Unit Test Enforcement: All unit tests must pass before commit

  • Basic Integration Tests: Fast subset of integration tests with credential validation

  • Quality Gate Enhancement: Comprehensive pre-commit validation pipeline

🔧 FIXES: GitHub Actions Integration

  • Workflow Environment Variables: Fixed missing HH_PROJECT in GitHub Actions workflows

  • Tox Environment Configuration: Fixed missing HH_PROJECT in local tox test environments

  • Integration Test Reliability: Resolved authentication failures in both CI/CD and local testing

  • Lambda Test Compatibility: Added proper environment configuration for AWS Lambda tests

v0.1.0rc1 (2025-09-11) - Release Candidate with Performance Improvements

🚀 NEW: Performance Optimization Framework

  • OTLP Performance Tuning: Configurable batch sizes and flush intervals for production optimization

  • Environment Variables: HH_BATCH_SIZE and HH_FLUSH_INTERVAL for fine-tuned performance control

  • Enhanced Span Processing: Improved batching performance with configurable parameters

  • API Client Improvements: Better error handling and configuration management

  • Documentation Navigation: Comprehensive validation framework with 0 broken links across 69 URLs

  • Integration Testing: Consolidated two-tier testing strategy with real API validation

  • RST Hierarchy: Fixed documentation structure across all provider integration guides

v0.1.0 (Development) - Major Architectural Refactor & Bug Fixes

🎯 NEW: Compatibility Matrix Framework (2025-09-05)

  • Complete Testing Framework: 13 provider compatibility tests with 100% success rate

  • Python Version Support: Full validation across Python 3.11, 3.12, and 3.13

  • Dynamic Generation: Automated maintenance reducing manual work by 75%

  • Official Documentation: Integrated compatibility matrix in Sphinx docs with optimal UX

  • Systematic Workarounds: Professional handling of upstream instrumentor bugs

  • Streamlined Architecture: 25% file count reduction with consolidated documentation

This release represents a comprehensive modernization of the HoneyHive Python SDK with significant architectural improvements and enhanced developer experience.

🔄 Breaking Changes

  • Modernized Architecture: HoneyHiveTracer now supports multiple independent instances

    • HoneyHiveTracer.init() method maintained for backwards compatibility

    • Direct constructor usage also available: HoneyHiveTracer(api_key="key")

    • Each initialization creates a new independent tracer instance

✨ Major Additions

  • Examples Directory Restructure: Organized provider examples into dedicated integrations/ subdirectory with 39% size reduction, improved navigation, and focused approach eliminating external dependencies

  • CSS-Based Dual-Theme System: Automatic light/dark theme detection for Mermaid sequence diagrams with targeted styling for optimal readability across all browsers

  • Documentation Quality Prevention System: Comprehensive error prevention and validation framework

    • Zero Build Warnings: Documentation now builds cleanly without any Sphinx warnings

    • Automated RST Validation: Pre-commit hooks validate structure and formatting

    • Type Safety Enforcement: All code examples use proper EventType enums

    • Code Example Testing: Automated validation ensures correct syntax and imports

  • Documentation Content Improvements: Major cleanup and standardization

    • Divio Architecture Compliance: Complete reorganization following Divio documentation system

    • Decorator-First Approach: Updated examples to emphasize @trace decorators

    • Type-Safe Examples: Replaced string literals with EventType enums

    • Backward Compatibility Documentation: Comprehensive guide for tracer auto-discovery

  • Automatic Tracer Discovery: Enhanced decorator functionality

    • @trace decorator now works without explicit tracer parameter

    • OpenTelemetry baggage-based tracer discovery mechanism

    • set_default_tracer() function for global tracer configuration

    • Maintains backward compatibility with existing code

  • Enhanced Decorator Support: Improved tracing capabilities

    • @trace_class decorator for automatic class-level tracing

    • enrich_span() utility function for adding context to active spans

    • Unified decorator behavior for both sync and async functions

    • Better error handling and span lifecycle management

🔧 Improvements

  • Testing Infrastructure: Comprehensive test coverage improvements

    • Unit tests for registry and tracer discovery mechanisms

    • Integration tests for backward compatibility scenarios

    • Performance testing for multi-instance scenarios

    • Mocking strategies for reliable test isolation

  • Developer Experience: Enhanced tooling and workflows

    • Pre-commit hooks for code quality and documentation validation

    • Strict changelog enforcement for high-frequency development environments

    • Feature synchronization verification

    • Enhanced error messages and debugging information

🐛 Fixes

  • API Endpoint Corrections: Fixed incorrect health check endpoints

  • Documentation Warnings: Resolved 23+ Sphinx build warnings

  • Import Issues: Fixed pylint ungrouped-imports warnings

  • Cross-Reference Links: Corrected broken internal documentation links

Note

Staying Updated

  • GitHub Releases: Watch the releases page for notifications

  • PyPI Updates: Monitor honeyhive on PyPI for new versions

  • Breaking Changes: Major version bumps indicate breaking changes - review the changelog carefully before upgrading

Version Upgrade Guide

Upgrading to Latest Version

# Upgrade to latest version
pip install --upgrade honeyhive

# Or specify a specific version
pip install honeyhive==X.Y.Z

Breaking Changes Checklist

When upgrading across major versions, review:

  1. API Changes: Check for deprecated or removed methods

  2. Configuration Changes: Verify environment variable names and formats

  3. Dependency Updates: Update any instrumentor packages if needed

  4. Import Changes: Update import statements if package structure changed

  5. Behavior Changes: Test critical paths for any behavioral differences

Migration Support

If you need help migrating between versions:

Contributing to the Changelog

For Contributors

When submitting pull requests, update the “Unreleased” section in CHANGELOG.md:

## [Unreleased]

### Added
- New feature description

### Changed
- Changed behavior description

### Deprecated
- Deprecated feature notice

### Removed
- Removed feature description

### Fixed
- Bug fix description

### Security
- Security improvement description

Change Categories

  • Added: New features

  • Changed: Changes in existing functionality

  • Deprecated: Soon-to-be removed features

  • Removed: Removed features

  • Fixed: Bug fixes

  • Security: Security improvements

Writing Good Changelog Entries

  • Be specific: “Fixed trace span duration calculation” vs “Fixed bug”

  • Include impact: “Breaking Change: Removed deprecated trace_event() method”

  • Add context: “Improved performance by 40% for large trace batches”

  • Reference issues: “Fixed #123: Memory leak in async tracing”

Release Process

For Maintainers

The release process follows these steps:

  1. Update Version: Bump version in pyproject.toml

  2. Update Changelog: Move “Unreleased” items to new version section

  3. Create Release: Tag and create GitHub release

  4. Publish Package: Automated publishing to PyPI

  5. Update Documentation: Deploy updated docs with new version

Release Schedule

  • Major Releases: Quarterly (breaking changes, major features)

  • Minor Releases: Monthly (new features, improvements)

  • Patch Releases: As needed (bug fixes, security updates)

  • Pre-releases: Beta versions for testing major changes

Version Numbering

Following Semantic Versioning:

  • Major: Breaking changes (1.0.0 → 2.0.0)

  • Minor: New features, backwards compatible (1.0.0 → 1.1.0)

  • Patch: Bug fixes, backwards compatible (1.0.0 → 1.0.1)

  • Pre-release: Beta versions (1.1.0-beta.1)