Changelog

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

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

Unreleased

0.7.0 - 2025-01-27

Added

Continue CLI Agent Support
- New adapter for Continue CLI (@continuedev/cli) - AI coding assistant with TUI and headless modes
- Command: cn (not continue)
- Installation via npm: npm install -g @continuedev/cli
- Version tested: 1.5.12
- Key features:
  - -p flag for prompt input (different pattern from Claude/Gemini's stdin approach)
  - --model flag for model specification (e.g., gpt-4, claude-sonnet-4-5)
  - --silent flag to strip <think></think> tags from output
  - Authentication via cn login to Continue Mission Control
- Full Agent interface implementation with structured prompts
- Message filtering to prevent echo in multi-agent conversations
- Streaming and non-streaming message support
- Example configurations:
  - examples/continue-coding.yaml - Single Continue agent for refactoring
  - examples/continue-team-coding.yaml - Multi-agent team (Continue + Claude + Gemini)
- Complete test coverage and documentation
- Now supporting 16 AI agent CLIs (up from 15)

Technical Details

New Module: pkg/adapters/continue.go
- Flag-based prompt pattern: cn -p "prompt" instead of stdin
- Implements structured three-part prompts (identity, context, instruction)
- filterRelevantMessages() excludes agent's own messages
- filterStatusMessages() removes Continue CLI status output
- isStatusMessage() detects and filters loading/initialization messages
- Health check with --version flag, fallback to --help
Updated: internal/registry/agents.json
- Added Continue entry with npm install/upgrade/uninstall commands
- Marked requires_auth: true
- Documentation: https://docs.continue.dev/cli
Updated Tests:
- pkg/adapters/adapters_test.go - Added TestContinueAgentInitialization
- internal/registry/registry_test.go - Updated agent count from 15 to 16
Documentation Updates:
- README.md - Added Continue to supported agents list
- CLAUDE.md - Added Continue CLI technical details

Benefits

✅ TUI and Headless Modes: Works in both interactive and automated contexts
🚀 Model Flexibility: Specify any supported model via --model flag
🔇 Clean Output: --silent flag removes internal reasoning tags
🔐 Secure Authentication: Centralized auth via Continue Mission Control
🎯 Multi-Agent Ready: Filters messages for clean conversation participation
📝 Well Documented: Complete examples and troubleshooting guides

[0.6.0] - 2025-10-26

Added

Complete JSON-Only Output Mode (--json flag)
- Pure JSONL stream - every line is a valid JSON object from the very first line
- ALL output (logs, messages, events, diagnostics) emitted as JSON to stdout
- Includes agent messages, system messages, diagnostic logs, and metadata
- Two event types:
  - Conversation events: bridge.connected, conversation.started, message.created, conversation.completed
  - Log events: log.entry for all messages and diagnostic logs
- Enables complete conversation replay and analysis from JSON stream alone
- Perfect for log aggregators, monitoring tools, CI/CD pipelines, and automation
- Usage: agentpipe run --json -a gemini:Bot --prompt "test" | jq
- Benefits:
  - ✅ Real-time streaming (see events as they happen)
  - ✅ Pure JSON from line 1 (no console format logs)
  - ✅ Easy to pipe to jq, log aggregators, monitoring tools
  - ✅ CI/CD friendly
  - ✅ No breaking changes - opt-in via flag
Diagnostic Logs as JSON Events
- All zerolog diagnostic logs (INF, WRN, ERR, DBG) emitted as log.entry events with role: "diagnostic"
- Includes metadata from log fields (agent_id, duration, tokens, etc.)
- Clean separation: role: "diagnostic" for system logs vs role: "agent"/"system"/"user" for chat messages

Fixed

Pure JSON output from first line
- Fixed issue where initial config log appeared as console format instead of JSON
- Moved JSON emitter initialization to cmd/root.go:initConfig() (before any log calls)
- Global stdoutEmitter shared across root and run commands
- No more mixed console/JSON output
Qoder install/upgrade commands
- Fixed exit status 127 error when running agentpipe agents install qoder or agentpipe agents upgrade qoder
- Changed from bash -- to bash -s -- in install/upgrade commands for proper stdin handling
- Updated commands: curl -fsSL https://qoder.com/install | bash -s -- --force
- Added test to verify bash command format in agent registry
- Fixes #26

Technical Details

New Event Type: log.entry in internal/bridge/events.go
- LogEntryData struct with level, agent info, content, role, metadata, and metrics
- LogEntryMetrics struct for duration, tokens, cost
- Supports both chat messages and diagnostic logs in unified format
New Module: internal/bridge/zerolog_json_writer.go
- Custom zerolog writer that parses zerolog JSON output
- Emits diagnostic logs as log.entry events to stdout
- Extracts level, message, and metadata from zerolog fields
New Module: internal/bridge/stdout_emitter.go
- Implements BridgeEmitter interface for stdout JSON output
- Uses same event schemas as HTTP bridge emitter
- Added EmitLogEntry() method for log event emission
New Interface: internal/bridge/interface.go
- BridgeEmitter interface allows both HTTP and stdout emitters
Updated: cmd/root.go
- Added globalJSONEmitter package variable to share emitter across commands
- Modified initConfig() to detect --json flag and initialize JSON emitter immediately
- Zerolog initialized with ZerologJSONWriter before any log calls in JSON mode
Updated: cmd/run.go
- Added --json flag
- Suppresses UI output when --json is set
- Uses globalJSONEmitter from root
- Set JSON emitter on logger when --json flag is used
- Passes nil writers to logger and orchestrator in JSON mode
Updated: pkg/logger/logger.go
- Added jsonEmitter field to ChatLogger
- Added SetJSONEmitter() method
- Modified LogMessage(), LogError(), and LogSystem() to emit JSON events when JSON emitter is set
Updated: pkg/orchestrator/orchestrator.go
- Changed bridgeEmitter field to use BridgeEmitter interface

Migration Notes

This release consolidates and supersedes v0.5.3, v0.5.4, and v0.5.5
No breaking changes - --json flag is opt-in
All existing functionality remains unchanged

[0.5.5] - 2025-10-26 (superseded by v0.6.0)

Fixed

Pure JSON output from first line
- Fixed issue where initial config log appeared as console format instead of JSON
- Moved JSON emitter initialization to cmd/root.go:initConfig() (before any log calls)
- ALL output now pure JSONL from the very first line when using --json flag
- No more mixed console/JSON output - every line is valid JSON
- Global stdoutEmitter shared across root and run commands

Technical Details

Updated: cmd/root.go
- Added globalJSONEmitter package variable to share emitter across commands
- Modified initConfig() to detect --json flag and initialize JSON emitter immediately
- Zerolog initialized with ZerologJSONWriter before any log calls in JSON mode
Updated: cmd/run.go
- Simplified JSON initialization to use globalJSONEmitter from root
- Removed duplicate emitter creation (now created once in initConfig)
- Passes shared emitter to orchestrator and chat logger

[0.5.4] - 2025-01-26 (superseded by v0.6.0)

Added

Complete JSON-only output mode
- ALL output (logs, messages, events) now emitted as JSON to stdout when using --json flag
- Includes agent messages, system messages, diagnostic logs, and metadata
- Pure JSONL stream - every line is a valid JSON object
- Two event types:
  - Conversation events: bridge.connected, conversation.started, message.created, conversation.completed
  - Log events: log.entry for all messages and diagnostic logs
- Enables complete conversation replay and analysis from JSON stream alone
- Perfect for log aggregators, monitoring tools, CI/CD pipelines, and automation

Enhanced

Diagnostic logs as JSON events
- All zerolog diagnostic logs (INF, WRN, ERR, DBG) emitted as log.entry events with role: "diagnostic"
- Includes metadata from log fields (agent_id, duration, tokens, etc.)
- Clean separation: role: "diagnostic" for system logs vs role: "agent"/"system"/"user" for chat messages

Technical Details

New Event Type: log.entry in internal/bridge/events.go
- LogEntryData struct with level, agent info, content, role, metadata, and metrics
- LogEntryMetrics struct for duration, tokens, cost
- Supports both chat messages and diagnostic logs in unified format
New Module: internal/bridge/zerolog_json_writer.go
- Custom zerolog writer that parses zerolog JSON output
- Emits diagnostic logs as log.entry events to stdout
- Extracts level, message, and metadata from zerolog fields
Updated: internal/bridge/stdout_emitter.go
- Added EmitLogEntry() method for log event emission
Updated: pkg/logger/logger.go
- Added jsonEmitter field to ChatLogger
- Added SetJSONEmitter() method
- Modified LogMessage(), LogError(), and LogSystem() to emit JSON events when JSON emitter is set
- Falls back to console/file output when JSON emitter is not set
Updated: cmd/run.go
- Set JSON emitter on logger when --json flag is used
- Reinitialize zerolog with ZerologJSONWriter for diagnostic log conversion
- Ensures all output becomes JSON events

[0.5.3] - 2025-01-26 (superseded by v0.6.0)

Added

JSON-only output format (--json flag)
- Output conversation events as streaming JSONL (one JSON object per line) to stdout
- Clean programmatic output for CI/CD pipelines, monitoring tools, and automation
- Suppresses all UI elements (logo, initialization messages, session summary, agent messages)
- Events emitted: bridge.connected, conversation.started, message.created, conversation.completed
- Matches bridge events format for consistency
- Usage: agentpipe run --json -a gemini:Bot --prompt "test" | jq
- Benefits:
  - ✅ Real-time streaming (see events as they happen)
  - ✅ Easy to pipe to jq, log aggregators, monitoring tools
  - ✅ CI/CD friendly
  - ✅ No breaking changes - opt-in via flag

Technical Details

New Module: internal/bridge/stdout_emitter.go
- Implements BridgeEmitter interface for stdout JSON output
- Uses same event schemas as HTTP bridge emitter
New Interface: internal/bridge/interface.go
- BridgeEmitter interface allows both HTTP and stdout emitters
Updated: pkg/orchestrator/orchestrator.go
- Changed bridgeEmitter field to use BridgeEmitter interface
Updated: cmd/run.go
- Added --json flag
- Suppresses UI output when --json is set
- Passes nil writers to logger and orchestrator in JSON mode

[0.5.2] - 2025-01-26

Fixed

Summary Field Naming Consistency
- Reverted field naming from ShortSummary/Summary back to ShortText/Text in SummaryMetadata
- Rationale: "Summary" is redundant when fields are already in SummaryMetadata struct
- Updated field naming in StateMetadata from ShortSummary/Summary to ShortText/Text for consistency
- JSON field names: short_text and text (inside parent summary object)
- This provides cleaner naming: SummaryMetadata.Text vs SummaryMetadata.Summary
- Updated all references across:
  - internal/bridge/events.go - Event data structures
  - pkg/conversation/state.go - Conversation state persistence
  - pkg/orchestrator/orchestrator.go - Summary metadata initialization
  - cmd/run.go - State metadata population
  - All test files updated accordingly
- All tests pass, no breaking changes to external APIs

[0.5.1] - 2025-01-26

Added

Dual Summary Generation (Short + Full)
- Conversations now generate both short (1-2 sentences) and full summaries in a single LLM query
- Structured prompt format: SHORT: [summary] and FULL: [summary] for reliable parsing
- Consistent field naming across bridge events and conversation state:
  - ShortSummary - Concise 1-2 sentence summary ideal for list views
  - Summary - Comprehensive detailed summary for detail pages
- Fallback logic: If parsing fails, extracts first 1-2 sentences for short summary
- Summary data persisted in conversation state files (~/.agentpipe/states/)
- New GetSummary() method on Orchestrator for programmatic access
- Benefits:
  - ✅ Single API call generates both summaries (cost & time efficient)
  - ✅ Easy parsing with structured markers
  - ✅ Graceful degradation on parse failures
  - ✅ Ready for AgentPipe Web UI integration
CLI Model Specification for --agents Flag
- Three flexible formats for specifying agents via CLI:
  - type - Auto-generated name with default model
  - type:name - Custom name with default model (existing format)
  - type:model:name - Custom name with specific model (NEW)
- Model validation with comprehensive error handling:
  - Validates agent types against registry
  - Checks if agent supports model specification
  - Enforces required model for OpenRouter
  - Prevents model specification for unsupported agents (kimi, cursor, amp)
- Model support matrix:
  - ✅ Optional: claude, gemini, copilot, qwen, factory, qoder, codex, groq, crush
  - ✅ Required: openrouter (API-based agent)
  - ❌ Not supported: kimi, cursor, amp, opencode
- All CLI adapters updated to use --model flag when Config.Model is set
- Examples:
```
# Use default models
agentpipe run -a claude:Alice -a gemini:Bob

# Specify models explicitly
agentpipe run -a claude:claude-sonnet-4-5:Architect \
  -a gemini:gemini-2.5-pro:Reviewer

# OpenRouter with provider/model format
agentpipe run -a openrouter:anthropic/claude-sonnet-4-5:Assistant \
  -a openrouter:google/gemini-2.5-pro:Critic
```

Technical Details

New Module: cmd/model_validation.go
- ModelSupport struct defining support and requirement status
- agentModelSupport map for all agent types
- Validation functions:
  - validateAgentType() - Checks agent type exists
  - validateModelForAgent() - Validates model specification for agent
  - validateModelInRegistry() - Warns if model not in provider registry
  - checkModelRequired() - Enforces required model for certain agents
  - parseAgentSpecWithModel() - Parses all three formats with validation
Updated: cmd/run.go
- parseAgentSpec() now uses parseAgentSpecWithModel()
- Populates agent.AgentConfig.Model field from CLI input
- Comprehensive error messages for invalid specifications
Updated Adapters (9 total):
- pkg/adapters/claude.go - Added --model flag support (SendMessage + StreamMessage)
- pkg/adapters/groq.go - Added --model flag support (SendMessage + StreamMessage)
- pkg/adapters/crush.go - Added --model flag support (SendMessage + StreamMessage)
- Already had support: qwen, factory, qoder, codex, copilot, gemini
Comprehensive Tests: cmd/run_test.go
- TestParseAgentSpec() - 30+ test cases covering all formats
- TestParseAgentSpecWithModel() - Format parsing validation
- TestValidateAgentType() - Agent type validation
- TestValidateModelForAgent() - Model support validation
- Tests for error cases: empty specs, unknown types, unsupported models, required models

Benefits

🚀 Simplified Model Selection: Specify models directly via CLI without YAML
🔍 Smart Validation: Immediate feedback for invalid agent/model combinations
📝 Clear Error Messages: Actionable error messages for misconfigurations
🎯 Type Safety: Compile-time and runtime validation of agent specifications
🔄 Backward Compatible: Existing type:name format still works
🌐 OpenRouter Integration: Seamless model specification for API-based agents

Documentation

README.md: New "Agent specification formats" section with:
- Format descriptions and examples
- Model support matrix table
- Comprehensive CLI examples
- Error examples showing what not to do
Updated run command flags documentation to show all three formats

[v0.6.0] - 2025-10-25

Added

OpenRouter API Support - First API-Based Agent
- New openrouter agent type for direct API integration without CLI dependencies
- Access 400+ models from multiple providers through a unified API
- No CLI installation required - just set OPENROUTER_API_KEY environment variable
- Support for models from Anthropic, OpenAI, Google, DeepSeek, Groq, and many more
- Real-time token usage and accurate cost tracking from API responses
- Streaming and non-streaming message support via Server-Sent Events (SSE)
- Smart model matching with provider registry integration
- Example configurations:
  - examples/openrouter-conversation.yaml - Multi-provider conversation
  - examples/openrouter-solo.yaml - Single agent reasoning task

Technical Details

New Package: pkg/client/ for HTTP client infrastructure
- openai_compat.go - Generic HTTP client for OpenAI-compatible APIs
- Support for streaming (SSE) and non-streaming requests
- Retry logic with exponential backoff (1s, 2s, 4s)
- Bearer token authentication
- Comprehensive error handling with retry strategies
- Context cancellation support
New Adapter: pkg/adapters/openrouter.go
- Implements complete Agent interface using HTTP API
- Converts AgentPipe messages to OpenAI Chat Completions format
- Integrates with provider registry for cost calculation
- Automatic API key detection from $OPENROUTER_API_KEY
- Health check via minimal API request
- GetCLIVersion() returns "N/A (API)" for API-based agents
Comprehensive Test Coverage:
- pkg/client/openai_compat_test.go - HTTP client tests with mocked responses
- pkg/adapters/openrouter_test.go - Agent adapter tests
- Integration tests (skipped without API key)
- 80% code coverage

Benefits

✅ No CLI Dependencies: Use models without installing any CLI tools
🚀 Direct API Access: Lower latency, more reliable than CLI execution
🌍 Comprehensive Model Support: 400+ models from multiple providers
💰 Accurate Pricing: Real token counts from API responses for precise cost tracking
📡 Streaming Support: Real-time response streaming via SSE
🔌 Unified Architecture: Same Agent interface for both CLI and API-based agents
🛤️ Foundation for Future: Paves the way for direct Anthropic API, Google AI API, Groq API, etc.

Architecture Impact

Establishes pattern for API-based agents separate from CLI-based agents
Creates reusable HTTP client for future OpenAI-compatible providers
Maintains backward compatibility - all existing CLI-based agents unchanged
Demonstrates hybrid approach: AgentPipe can use both CLI tools and direct APIs

Documentation

README.md: New "Using OpenRouter (API-Based Agents)" section
Detailed setup instructions with environment variable configuration
Model examples and use case recommendations
Links to OpenRouter documentation and model list

[v0.5.0] - 2025-10-25

Added

Provider Pricing Integration from Catwalk
- Accurate cost estimation using Catwalk provider configs
- Support for 16 AI providers with comprehensive pricing data:
  - AIHubMix, Anthropic, Azure OpenAI, AWS Bedrock, Cerebras, Chutes
  - DeepSeek, Gemini, Groq, Hugging Face, OpenAI, OpenRouter
  - Venice, Vertex AI, xAI, and more
- New agentpipe providers command with subcommands:
  - list - Show all providers and models with pricing
  - show <provider> - Display detailed provider information
  - update - Fetch latest pricing from Catwalk GitHub
- Smart model matching with exact, prefix, and fuzzy matching
- Hybrid config loading: embedded defaults with optional ~/.agentpipe/providers.json override
- Comprehensive test coverage (>80%) for all provider functionality

Changed

Cost Estimation: Refactored EstimateCost() to use provider registry instead of hardcoded prices
- Falls back to $0 with warning for unknown models
- Provides detailed debug logging with model and provider information
- Legacy function preserved as EstimateCostLegacy() for compatibility
Provider Data: Single consolidated JSON file with all pricing info (120KB embedded)
- Auto-generated from Catwalk's 16 provider configs
- Includes model pricing, context windows, capabilities, and more
- Version-tracked with update timestamps and source attribution

Technical Details

New internal/providers/ package with complete provider management
Provider registry with RWMutex for thread-safe concurrent access
HTTP client with retry logic and exponential backoff for Catwalk fetches
Build script (scripts/update-providers.go) to regenerate providers.json
go:embed directive for zero-dependency embedded pricing data
Supports both JSON and human-readable table output formats

Benefits

✅ Accurate Costs: Real pricing from Catwalk, not hardcoded approximations
🔄 Always Current: Simple agentpipe providers update to get latest pricing
🌍 Comprehensive: 16 providers with 100+ models covered
🎯 Smart Matching: Handles model ID variations automatically
📊 Better Insights: More accurate conversation cost tracking
🛡️ Backwards Compatible: Legacy pricing still available if needed

[v0.4.9] - 2025-10-25

Added

Crush CLI Agent Support (Charmbracelet)
- New adapter for Crush CLI (crush command)
- Terminal-first AI coding assistant with beautiful TUI
- Installation via Homebrew: brew install charmbracelet/tap/crush
- Installation via Go: go install github.com/charmbracelet/crush@latest
- Supports stdin-based message passing
- Complete integration with standardized three-part prompt system
- Comprehensive logging and error handling
- Registry entry with install/uninstall/upgrade commands for all platforms
- Multi-provider authentication support:
  - Anthropic (ANTHROPIC_API_KEY)
  - OpenAI (OPENAI_API_KEY)
  - Groq (GROQ_API_KEY)
  - Google Gemini (GEMINI_API_KEY)
  - AWS Bedrock, Azure OpenAI, and others

Technical Details

Implements all required Agent interface methods
Follows established patterns from Claude and Groq adapters
Message filtering to prevent echo in multi-agent conversations
Output cleaning to remove system messages and prompts
Health check via --version flag
Version detection via GitHub releases API (charmbracelet/crush)
Supports both SendMessage and StreamMessage modes
Current latest version: v0.12.3

Benefits

Adds another high-quality terminal-first agent option
Multi-provider support allows flexibility in LLM choice
Beautiful TUI experience from Charmbracelet
MCP (Model Context Protocol) and LSP integration support

[v0.4.8] - 2025-10-25

Fixed

GitHub API Rate Limiting (403 Errors): Resolved by using proper package managers
- Kimi and Qwen were hitting GitHub's API rate limit (60 requests/hour unauthenticated)
- Changed to use the correct source for each package instead of GitHub releases

Added

PyPI Package Manager Support
- New getPyPILatestVersion() function in version.go
- Fetches from https://pypi.org/pypi/{package}/json
- No rate limits, highly reliable
- Used for Python packages like kimi-cli

Changed

Kimi CLI Version Detection
- Package manager: github → pypi
- Package name: MoonshotAI/kimi-cli → kimi-cli
- Now fetches from PyPI (aligns with uv tool install kimi-cli)
- Eliminates GitHub API 403 errors
Qwen Version Detection
- Package manager: github → npm
- Package name: QwenLM/qwen-code → @qwen-code/qwen-code
- Now fetches from npm registry (aligns with npm install -g @qwen-code/qwen-code)
- Eliminates GitHub API 403 errors
Linter Configuration
- Added dupl exclusion for internal/registry/ path
- Allows intentional code duplication across package manager HTTP functions

Benefits

No more API rate limit errors when checking versions
Faster version checks (no rate limiting delays)
More reliable (uses authoritative source for each package)
Aligns version detection with actual installation methods

Technical Details

PyPI API endpoint: https://pypi.org/pypi/{package}/json
npm API endpoint: https://registry.npmjs.org/{package}/latest (existing)
Both have much higher rate limits than GitHub (60/hour)
Tested and verified with agentpipe agents list --outdated

[v0.4.7] - 2025-10-25

Changed

Kimi CLI Version Detection: Updated to use GitHub releases API
- Changed package manager from "uv" to "github" in registry
- Package name now points to "MoonshotAI/kimi-cli" repository
- Uses standard GitHub releases endpoint for version checking
- More reliable than previous custom parsing approach
- Consistent with other GitHub-based agents (Qwen, etc.)

Technical Details

Reuses existing getGitHubLatestRelease() infrastructure
Fetches from https://api.github.com/repos/MoonshotAI/kimi-cli/releases/latest
Properly detects and compares versions for update notifications
Tested with agentpipe agents list --outdated command

Benefits

Improved reliability for version detection
Simpler implementation using existing code
Better alignment with standard practices
More accurate update notifications for users

[v0.4.6] - 2025-10-25

Added

Groq Code CLI Agent Support
- New adapter for Groq Code CLI (groq command)
- Powered by Groq's Lightning Processing Units (LPUs) for ultra-fast inference
- Installation via npm: npm install -g groq-code-cli@latest
- Supports temperature configuration via agent config
- Interactive CLI with stdin-based message passing
- Complete integration with standardized three-part prompt system
- Comprehensive logging and error handling
- Registry entry with install/uninstall/upgrade commands for all platforms
- Authentication via GROQ_API_KEY environment variable or /login command

Technical Details

Implements all required Agent interface methods
Follows established patterns from Claude and Gemini adapters
Message filtering to prevent echo in multi-agent conversations
Output cleaning to remove system messages and authentication prompts
Health check via --version flag
Version detection via registry system
Supports both SendMessage and StreamMessage modes

[v0.4.5] - 2025-10-25

Changed

Kimi Installation: Enhanced install and upgrade commands to explicitly specify Python 3.13
- Install: uv tool install --python 3.13 kimi-cli
- Upgrade: uv tool upgrade kimi-cli --python 3.13 --no-cache
- Ensures correct Python version is used for Kimi CLI deployment
- Updated README with explicit Python version specification
- Prevents version conflicts with other Python installations

[v0.4.4] - 2025-10-25

Added

Dedicated Security Workflows: Trivy and CodeQL now have their own workflows
- Separated security scanning from test workflow
- Improved workflow organization and clarity

Changed

README Badges: Enhanced badge section with additional metrics
- Added downloads badge showing total release downloads
- Added GitHub stars badge with icon
- Improved visual hierarchy with consistent color coding
- All badges properly linked for easy navigation

Fixed

Windows Test Failure: Fixed TestIsInstallable for platform-specific installations
- Removed Ollama from test expectations (Windows-only instructions)
- Test now only checks agents with actual install commands across all platforms
- Resolves GitHub Actions test failures on Windows runners

Improved

CI/CD Organization: Cleaner separation of concerns
- Test workflow focuses on testing and linting
- Security workflows handle vulnerability scanning
- Reduced workflow complexity and interdependencies

[v0.4.3] - 2025-10-25

Added

Kimi CLI Agent Support (Moonshot AI)
- New adapter for Kimi CLI (kimi command)
- Installation via uv tool install kimi-cli (requires Python 3.13+)
- Support for uv package manager-based installation
- Upgrade via uv tool upgrade kimi-cli --no-cache
- Blue-gradient ASCII logo in branding for visual distinction
- Health checks and agent verification
- Interactive-aware error handling with helpful authentication guidance
- Structured prompt building for multi-agent conversations
- Message filtering and context management
- Stream message support with best-effort implementation

Changed

Updated registry test to expect 12 supported agents (added Kimi)
Updated README with Kimi agent documentation and installation instructions

[v0.4.2] - 2025-10-24

Fixed

Qoder Installation/Upgrade: Added --force flag to Qoder install and upgrade commands
- Allows Qoder to be installed/upgraded even if version already exists
- Prevents "Version X.X.X already exists" errors

[v0.4.1] - 2025-10-22

Added

Conversation Summarization: Automatic AI-generated summaries at conversation completion
- Configurable summary agent (default: Gemini, supports all agent types)
- --no-summary flag to disable summaries
- --summary-agent flag to override configured agent
- Summary configuration in YAML config file
- Summary metadata includes agent type, model, tokens, cost, and duration
- Summary tokens and cost factored into conversation totals
- Smart prompt design avoiding meta-commentary
Unique Agent IDs: Enhanced agent identification for multiple agents of same type
- Agent IDs now unique per instance: {agentType}-{index} (e.g., claude-0, claude-1)
- AgentID included in all bridge streaming events
- Allows tracking of multiple agents with same type in single conversation
- AgentID in conversation.started participants list
- AgentID in all message.created events
Event Store: Local event persistence
- Events saved to ~/.agentpipe/events/ directory
- One JSON Lines file per conversation
- Non-blocking async operation
- Debug logging for storage errors

Changed

Bridge Events: Enhanced ConversationCompletedData structure
- Summary field now contains full SummaryMetadata (instead of plain string)
- Includes summary agent type, model, tokens, cost, duration
- Total tokens and cost now include summary metrics
- Duration does not include summary generation time
Streaming Protocol: Updated message event structure
- EmitMessageCreated now requires agentID as first parameter
- MessageCreatedData includes agent_id field
- AgentParticipant includes agent_id field

Fixed

Agent identification for conversations with multiple agents of same type
Cost tracking to include summary generation costs in totals
Thread-safe access to bridge emitter in orchestrator

[v0.4.0] - 2025-10-21

Added

Bridge Connection Events: Automatic connection announcement
- Emit bridge.connected event on emitter initialization
- System info included in connection event
- Synchronous sending ensures reliability
Cancellation Detection: Detect and report conversation interruption
- Emit conversation.completed with status="interrupted" on Ctrl+C
- Distinguish between normal completion and cancellation
- Proper error propagation in orchestrator

Changed

Event Reliability: Improved critical event delivery
- Use synchronous SendEvent for completion and error events
- Prevent truncated JSON payloads on program exit

[v0.3.0] - 2025-10-21

Added

Streaming Bridge: Opt-in real-time conversation streaming to AgentPipe Web
- Stream live conversation events to AgentPipe Web for browser viewing
- Four event types: conversation.started, message.created, conversation.completed, conversation.error
- Non-blocking async HTTP implementation that never blocks conversations
- CLI commands for easy configuration:
  - agentpipe bridge setup - Interactive configuration wizard
  - agentpipe bridge status - View current bridge configuration (with --json flag support)
  - agentpipe bridge test - Test connection to AgentPipe Web
  - agentpipe bridge disable - Disable streaming
- System info collection: OS, version, architecture, AgentPipe version, Go version
- Configuration via viper config file or environment variables
- Build-tag conditional defaults (dev vs production URLs)
- Privacy-first design: disabled by default, API keys never logged, clear disclosure
- Production-ready with retry logic (exponential backoff) and comprehensive tests (>80% coverage)
- Agent participants tracked with CLI version information
- Conversation metrics: turns, tokens, cost, duration

Changed

BREAKING: Extended Agent interface with new GetCLIVersion() method
- All agent adapters now implement version detection
- Uses internal registry for version lookup
- Required for streaming bridge agent participant data
- Custom agent implementations must add this method

Improved

Configuration: Added BridgeConfig struct to support streaming bridge settings
- Bridge enabled status, URL, API key, timeout, retry attempts, log level
- Defaults applied automatically in config parsing
- Environment variable overrides supported

Fixed

Thread Safety: Added RWMutex for safe concurrent access to orchestrator bridge emitter
Linting: Fixed non-constant format string error in orchestrator error handling

[v0.2.2] - 2025-10-20

Added

JSON Output Support: Added --json flag to agentpipe agents list command
- Regular list mode outputs structured JSON with agent details
- Outdated mode outputs version comparison data in JSON format
- Works with all existing filters: --installed, --outdated, --current
- Clean JSON structure with appropriate omitempty fields
- Example: agentpipe agents list --json
- Example: agentpipe agents list --outdated --json
- Useful for programmatic integration and automation

Improved

Agent List Output: Enhanced parallel version checking for both human-readable and JSON outputs
Code Organization: Refactored version row type for better reusability

v0.2.1 - 2025-10-20

Added

OpenCode CLI Agent: Complete integration for SST's OpenCode terminal-native AI coding agent
- Full adapter implementation with non-interactive opencode run mode
- npm package support: opencode-ai@latest
- Quiet flag for non-interactive execution
- Comprehensive documentation and troubleshooting
- Now 11 supported AI agent CLIs
Referral Links Section: New dedicated section in README to support project development
- Qoder referral link for users to support ongoing development
- Clear explanation of how referral links help fund the project

Fixed

Amp CLI: Updated to support npm-based installation and automated upgrades
- Changed from manual-only to npm install -g @sourcegraph/amp
- Now supports agentpipe agents upgrade amp
Codex CLI: Fixed npm package name for correct version detection
- Corrected from @openai/codex-cli to @openai/codex
- Added homebrew installation option: brew install --cask codex
- Automated upgrades now work correctly

Improved

Documentation: Enhanced installation instructions for multiple agents
- Added npm and homebrew options where applicable
- Updated Prerequisites section with all installation methods
- Added OpenCode to adapter reference implementations
Tests: Updated registry tests to reflect Amp's new installable status

v0.2.0 - 2025-10-20

Added

Agent Upgrade Command: New agentpipe agents upgrade subcommand for easy updates
- Upgrade individual agents: agentpipe agents upgrade claude
- Upgrade multiple agents: agentpipe agents upgrade claude ollama gemini
- Upgrade all installed agents: agentpipe agents upgrade --all
- Automatic detection of installed agents for selective upgrades
- User confirmation prompts before performing upgrades
- Cross-platform support (darwin, linux, windows)
Automated Version Detection: Complete version checking for all 10 supported agents
- npm registry integration (Claude, Codex, Gemini, Copilot, Amp, Qwen)
- Homebrew Formulae API integration (Ollama)
- GitHub Releases API integration (Qwen fallback)
- Shell script parsing for version extraction (Factory, Cursor)
- JSON manifest parsing (Qoder)
- Replaces all "manual install" placeholders with actual version numbers
Multiple Package Manager Support: Extensible version checking architecture
- npm: Query npm registry API for latest versions
- homebrew: Query Homebrew Formulae API for latest versions
- github: Query GitHub Releases API for latest releases
- script: Parse shell install scripts for VER= or DOWNLOAD_URL= version patterns
- manifest: Fetch and parse JSON manifests with "latest" field
Parallel Version Checking: Dramatically improved performance with concurrent API calls
- Goroutine-based concurrent version fetching
- Buffered channels for result collection
- Performance improvement: ~10+ seconds → ~3.7 seconds for 10 agents
- Thread-safe result aggregation

Fixed

npm 404 Errors: Corrected package names for npm-based agents
- Claude: @anthropic-ai/claude-cli → @anthropic-ai/claude-code
- Codex: @openai/codex-cli → @openai/codex
- Gemini: @google/generative-ai-cli → @google/gemini-cli
Ollama Version Detection: Enhanced to work without running Ollama instance
- Now parses version from warning messages (e.g., "Warning: client version is 0.12.5")
- Improved containsVersion() and extractVersionNumber() logic
- No longer requires Ollama server to be running

Improved

UI/UX Enhancements: Better table display for agent version information
- Removed redundant "Status" column from outdated agents table
- Rebalanced column widths for better readability
  - Agent: 15 → 12 characters
  - Installed Version: 20 → 24 characters
  - Latest Version: 20 → 24 characters
  - Total width: 80 → 85 characters
- Changed upgrade instructions from "install" to "upgrade" for clarity
Agent Registry Metadata: Complete package manager information for all agents
- Factory: Uses script-based version detection from https://app.factory.ai/cli
- Amp: Uses npm registry @sourcegraph/amp
- Cursor: Uses script-based version detection from https://cursor.com/install
- Qoder: Uses manifest from qoder-ide.oss-ap-southeast-1.aliyuncs.com
- All agents now have upgrade commands defined for current OS

v0.1.5 - 2025-10-19

Fixed

Linting Errors: Fixed golangci-lint errors in doctor.go
- Fixed gofmt formatting (struct field alignment)
- Fixed prealloc warning (pre-allocated slices with known capacity)
- CI/CD pipeline now passes all quality checks

v0.1.4 - 2025-10-19

Added

Doctor Command JSON Output: Programmatic agent detection for web interfaces
- --json flag for structured output in JSON format
- Complete system diagnostics in machine-readable format
- Agent detection with availability, authentication, and version info
- Perfect for dynamic UI generation (e.g., agentpipe-web)
- Outputs: system_environment, supported_agents, available_agents, configuration, summary
- Each agent includes: name, command, path, version, install/upgrade commands, docs, auth status
- Clean JSON output (logo suppressed when using --json flag)

Improved

Documentation: Added comprehensive JSON output format documentation to README
- Usage examples for both human-readable and JSON modes
- Field-by-field JSON structure explanation
- Use cases for programmatic consumption

v0.1.3 - 2025-10-19

Added

Factory CLI Agent Support: Full integration with Factory.ai's Droid coding agent
- Non-interactive exec mode with droid exec command
- Autonomy level configuration (--auto high) for multi-agent conversations
- Structured prompt delivery with clear context sections
- Smart message filtering (excludes agent's own messages)
- Comprehensive logging and error handling
- Optional model specification via config
- Agent-native software development with Code Droid and Knowledge Droid
- Installation: curl -fsSL https://app.factory.ai/cli | sh
- Documentation: https://docs.factory.ai/cli

Improved

Doctor Command: Added Factory CLI detection with installation and upgrade instructions
README: Updated with Factory CLI support and troubleshooting section
Architecture Documentation: Added Factory to supported agents list and visual diagrams
Agent Count: Now supporting 10 AI agent CLIs (up from 9)

v0.1.1 - 2025-10-19

Fixed

Windows Test Compatibility: Fixed timer resolution issues causing test failures on Windows
- Windows timer granularity (~15.6ms) caused time.Since() to return 0 for very fast operations
- Increased mock agent delay to 20ms in TestConversationWithMetrics to ensure measurable durations
- Test now passes reliably on all platforms (Windows, macOS, Linux)
Windows File Permission Tests: Added runtime OS detection to skip Unix-specific permission checks
- TestState_Save now correctly skips file permission verification on Windows
- Tests properly handle platform differences in file permission models

Changed

Go Version Requirement: Downgraded from Go 1.25.3 to Go 1.24.0 for broader compatibility
- Maintains compatibility with golangci-lint v1.64.8
- All GitHub Actions workflows updated to use Go 1.24
- go.mod updated to reflect Go 1.24 requirement

Documentation

Added comprehensive documentation of Windows-specific testing challenges in CLAUDE.md
Documented timer resolution requirements for cross-platform test development
Updated development guide with platform compatibility considerations

v0.1.0 - 2025-10-16

Added

Agent Type Indicators: Message badges now show agent type in parentheses (e.g., "Alice (qoder)")
- Helps users quickly identify which agent type is responding
- Displayed in all message badges in both TUI and CLI output
- Agent type automatically populated from agent configuration
Branded TUI Logo: Enhanced TUI with colored ASCII sunset gradient logo
- Consistent branding across CLI and TUI modes
- Shared branding package for code reuse
- Beautiful sunset gradient colors using ANSI 24-bit color codes
TUI Search Feature: Press Ctrl+F to search through conversation messages
- Case-insensitive search through message content and agent names
- n/N navigation between search results
- Visual feedback with match count and position
- Auto-scroll to current search result
Agent Filtering: Use slash commands to filter messages by agent
- /filter <agent> to show only specific agent's messages
- /clear to remove active filter
- System messages always visible regardless of filter
Help Modal: Press ? to view all keyboard shortcuts
- Organized by category (General, Conversation, Search, Commands)
- Toggle on/off with ? or Esc
- Comprehensive documentation of all keybindings
Amp CLI Agent Support: Advanced integration with Sourcegraph's Amp coding agent
- Thread management for efficient conversations
- Smart message filtering (excludes agent's own messages)
- Structured prompt delivery with clear context sections
- Streaming support with thread continuation
- Reduces API costs by 50-90% vs traditional approaches
Qoder CLI Agent Support: Full integration with Qoder agentic coding platform
- Non-interactive print mode with qodercli --print
- Enhanced context engineering for comprehensive codebase understanding
- Intelligent agents for systematic software development tasks
- Built-in tools (Grep, Read, Write, Bash) for file operations
- MCP integration support for extended functionality
- Permission bypass with --yolo flag for automated execution

Improved

Enhanced HOST vs SYSTEM Distinction: Clearer visual separation in message display
- HOST messages now formatted like agent messages with badge, newline, and indented content
- SYSTEM messages remain inline format for announcements
- HOST badge uses distinctive purple color (#99)
- Makes conversation context clearer by distinguishing orchestrator prompts from system notifications
Gemini Adapter Reliability: Improved error handling for process exit issues
- Now accepts valid output even when Gemini CLI doesn't exit cleanly
- Distinguishes between real API errors (404, 401) and harmless process termination
- Enhanced output cleaning to filter error traces and stack dumps
- Significantly reduces false failures in multi-agent conversations
Standardized Agent Introduction: All agents now receive complete conversation history when first coming online
- Complete message delivery: Agents receive ALL existing messages (system prompts + agent messages)
- No message loss: Correctly identifies orchestrator's initial prompt vs. agent announcements
- Correct topic extraction: Finds system message with AgentID="system" as the conversation topic
- Clear labeling: System messages are explicitly labeled as "SYSTEM:" in conversation history
- Structured three-part prompt format:
  1. AGENT SETUP (first): Agent's name, role, and custom instructions
  2. CONVERSATION TOPIC (second): Initial orchestrator prompt highlighted prominently
  3. CONVERSATION SO FAR (third): All existing messages (announcements + responses)
Amp Agent Context Awareness: Restructured prompt delivery with thread management
- Uses amp thread new and amp thread continue for efficient communication
- Smart message filtering: Automatically excludes Amp's own messages from being sent back to it
- Only sends messages from OTHER agents and system messages (Amp already knows what it said)
- Thread management: Reduces API costs and response times by 50-90%
- Automatic thread ID tracking and incremental message sending
- Enhanced logging with prompt previews (first 300 chars) for debugging
Session Summary: Now displayed for all conversation endings, not just CTRL-C interruptions
- Shows summary when conversation completes normally (max turns reached)
- Shows summary when interrupted with CTRL-C
- Shows summary even when conversation ends with an error
- Includes total messages, tokens, time spent, and cost for all endings

Fixed

Inconsistent Agent Badge Colors: Fixed race condition causing first message to have grey badge
- Now ensures agent color is assigned before badge style is retrieved
- Agent name badges now consistently show the assigned color from first message onward
- Improved visual consistency in both TUI and CLI output
TUI Display Corruption: Fixed stderr output interfering with TUI rendering
- Removed all fmt.Fprintf(os.Stderr, ...) calls from TUI code
- Metrics and logging no longer corrupt the TUI alt-screen display
- Silent error handling in TUI mode while maintaining conversation panel visibility
Agent Prompt Response: Fixed critical bug where agents weren't properly responding to orchestrator's initial prompt
- Changed prompt header from passive "CONVERSATION TOPIC" to directive "YOUR TASK - PLEASE RESPOND TO THIS"
- Makes it clear the initial prompt is a direct instruction, not passive context
- Agents now immediately engage with the topic instead of asking "what would you like help with?"
Amp Thread Creation Pattern: Fixed empty response issue with Amp agent
- Previously: Created thread with prompt, received empty response
- Now: Create empty thread first, then send prompt via thread continue
- Matches Amp CLI's expected pattern where thread new only returns thread ID
- Amp now correctly responds to initial prompts on first turn
Agent Introduction Logic: Fixed orchestrator prompt detection
- Correctly distinguishes between orchestrator messages (AgentID="system") and agent announcements
- Agent announcements are system messages from specific agents, not the conversation topic
- All agents now receive the orchestrator's initial prompt in the "YOUR TASK" section
Codex Non-Interactive Mode: Fixed terminal compatibility errors with Codex agent
- Uses codex exec subcommand for non-interactive execution
- Parses JSON output to extract agent messages cleanly
- Automatically bypasses approval prompts with safety flags
- No more "stdout is not a terminal" errors in multi-agent conversations
Standardized All Adapters: Applied consistent interaction pattern across all 8 adapters
- All adapters (Amp, Claude, Codex, Copilot, Cursor, Gemini, Qoder, Qwen) now use identical:
  - Three-part structured prompts (Setup → Task → History)
  - Message filtering to exclude agent's own messages
  - Comprehensive structured logging with timing and metrics
  - Proper error handling with specific error detection
- Ensures reliable, consistent behavior across all agent types
Orchestrator Identification: Changed orchestrator messages from "System" to "HOST" for clarity
- Initial conversation prompt now uses AgentID="host" and AgentName="HOST"
- Distinguishes orchestrator messages from system announcements (agent joins, etc.)
- All 8 adapters updated to recognize both "system"/"System" and "host"/"HOST" for backwards compatibility
- Makes it clear who is presenting the initial task vs. system notifications

v0.0.16 - 2025-10-15

Added - Production-Ready Features

Prometheus Metrics: Comprehensive observability with 10+ metric types
- HTTP server with /metrics, /health, and web UI endpoints
- Ready for Grafana dashboards and Prometheus alerting
- Track requests, durations, tokens, costs, errors, rate limits, retries
- OpenMetrics format support
Middleware Pipeline: Extensible message processing architecture
- 10+ built-in middleware (logging, metrics, validation, filtering, sanitization)
- Custom middleware support for transforms and filters
- Error recovery and panic handling
- Chain-of-responsibility pattern implementation
Conversation State Management: Save/resume conversations
- Save/resume conversations from JSON state files
- agentpipe resume command with --list flag
- Automatic state directory (~/.agentpipe/states/)
- Full conversation history, config, and metadata preservation
Export Functionality: Multi-format conversation export
- Export to JSON, Markdown, or HTML formats
- Professional HTML styling with responsive design
- XSS prevention with HTML escaping
- agentpipe export command

Added - Reliability & Performance

Rate Limiting: Token bucket algorithm per agent
- Configurable rate and burst capacity
- Thread-safe implementation with ~60ns overhead
- Automatic rate limit hit tracking in metrics
Retry Logic: Exponential backoff with smart defaults
- 3 retries with 1s initial delay, 30s max, 2.0x multiplier
- Configurable per orchestrator
- Retry attempt tracking in metrics
Structured Error Handling: Typed error system
- AgentError, ConfigError, ValidationError, TimeoutError, etc.
- Error wrapping with context
- Better error classification for metrics
Config Hot-Reload: Development workflow enhancement
- Watch config files for changes with viper.WatchConfig
- Thread-safe reload with callbacks
- --watch-config flag for development mode

Added - Docker Support

Multi-stage Dockerfile (~50MB final image)
docker-compose.yml with metrics server on port 9090
Health checks and graceful shutdown
Volume mounts for configs and logs
Non-root user for security
Complete documentation in docs/docker.md

Added - Testing & Quality

Comprehensive Test Coverage: 200+ tests
- 86+ unit tests across orchestrator, adapters, logger, errors, ratelimit, config, conversation
- 15 integration tests for end-to-end conversation flows
- 25+ benchmark tests for performance regression detection
- TUI component tests with race detection
- All tests passing with concurrent access validation
Documentation: Complete docs/ directory
- architecture.md - System design and patterns
- contributing.md - Contribution guidelines
- development.md - Development setup and workflows
- troubleshooting.md - Common issues and solutions
- docker.md - Docker deployment guide

Added - Developer Experience

Interactive Init Command: Configuration wizard
- Guided prompts for all configuration options
- Agent selection and configuration
- Orchestrator mode and settings
- Automatic file creation
Structured Logging: Zerolog-based logging
- JSON and pretty console output
- Contextual fields for debugging
- Integration across orchestrator, adapters, and commands
- Maintains fmt.Fprintf for TUI display
Enhanced CLI: New commands and flags
- agentpipe export - Export conversations
- agentpipe resume - Resume saved conversations
- agentpipe init - Interactive config wizard
- --save-state, --state-file, --watch-config flags

Improved

Godoc comments on all exported types and functions
0 linting issues with golangci-lint
Thread-safe implementations throughout
Proper resource cleanup and leak prevention

v0.0.15 - 2025-10-14

Added

GitHub Copilot CLI Integration: Full support for GitHub's Copilot terminal agent
- Non-interactive mode support using --prompt flag
- Automatic tool permission handling with --allow-all-tools
- Multi-model support (Claude Sonnet 4.5, GPT-5, etc.)
- Authentication detection and helpful error messages
- Subscription requirement validation

Improved

Graceful CTRL-C Handling: Interrupting a conversation now displays a session summary
- Total messages (agent + system)
- Total tokens used
- Total time spent (intelligently formatted: ms/s/m:s)
- Total conversation cost
- All messages are properly logged before exit
Total Time Tracking in TUI: Statistics panel now shows cumulative time for all agent requests

Fixed

Resource Leak Fixes:
- Fixed timer leak in cursor adapter (using time.NewTimer with proper cleanup)
- Message channel now properly closed on TUI exit
- Dropped messages now logged to stderr with counts
- Orchestrator goroutine lifecycle properly tracked with graceful shutdown

v0.0.9 - 2025-10-12

Added

Cursor CLI Integration: Full support for Cursor's AI agent
- Automatic authentication detection
- Intelligent retry logic for improved reliability
- Optimized timeout handling for cursor-agent's longer response times
- JSON stream parsing for real-time response streaming
- Robust error recovery and process management

v0.0.8 - 2025-10-10

Added - TUI Improvements

Real-time Activity Indicators: Visual feedback showing which agent is currently responding
Enhanced Metrics Display:
- Response time shown in seconds with 1 decimal precision (e.g., 2.5s)
- Token count for each response
- Cost estimate per response (e.g., $0.0012)
- Total conversation cost tracking in Statistics panel
Improved Message Formatting:
- Consolidated headers (timestamp and name only shown when speaker changes)
- Proper multi-paragraph message handling
- Clean spacing between messages
- No extra newlines between paragraphs from same speaker
Configuration Improvements:
- TUI now properly honors all configuration settings
- Config file path displayed in Configuration panel
- Dual output support (logs to file while displaying in TUI)
- Metrics display controlled by show_metrics config option

Improved

Better Error Handling: Clearer error messages for agent failures and timeouts
Improved Health Checks: More robust agent verification before starting conversations
Cost Tracking: Automatic calculation and accumulation of API costs
Metrics Pipeline: End-to-end metrics flow from orchestrator to TUI display

Performance & Reliability

Optimized Message Handling: Reduced memory usage and improved message rendering performance
Better Concurrency: Proper goroutine management and channel handling
Graceful Shutdowns: Clean termination of agents and proper resource cleanup

User Experience

Intuitive Panel Navigation: Tab-based navigation between panels
Real-time Feedback: Instant visual indicators for agent activity
Clean Message Display: Smart consolidation of headers and proper paragraph formatting
Cost Transparency: See exactly how much each conversation costs

FilesExpand file tree

CHANGELOG.md

Latest commit

History

CHANGELOG.md

File metadata and controls

Changelog

Unreleased

0.7.0 - 2025-01-27

Added

Technical Details

Benefits

[0.6.0] - 2025-10-26

Added

Fixed

Technical Details

Migration Notes

[0.5.5] - 2025-10-26 (superseded by v0.6.0)

Fixed

Technical Details

[0.5.4] - 2025-01-26 (superseded by v0.6.0)

Added

Enhanced

Technical Details

[0.5.3] - 2025-01-26 (superseded by v0.6.0)

Added

Technical Details

[0.5.2] - 2025-01-26

Fixed

[0.5.1] - 2025-01-26

Added

Technical Details

Benefits

Documentation

[v0.6.0] - 2025-10-25

Added

Technical Details

Benefits

Architecture Impact

Documentation

[v0.5.0] - 2025-10-25

Added

Changed

Technical Details

Benefits

[v0.4.9] - 2025-10-25

Added

Technical Details

Benefits

[v0.4.8] - 2025-10-25

Fixed

Added

Changed

Benefits

Technical Details

[v0.4.7] - 2025-10-25

Changed

Technical Details

Benefits

[v0.4.6] - 2025-10-25

Added

Technical Details

[v0.4.5] - 2025-10-25

Changed

[v0.4.4] - 2025-10-25

Added

Changed

Fixed

Improved

[v0.4.3] - 2025-10-25

Added

Changed

[v0.4.2] - 2025-10-24

Fixed

[v0.4.1] - 2025-10-22

Added

Changed

Fixed

[v0.4.0] - 2025-10-21

Added