Beta Release v0.6.2-beta.2

⚠️ Breaking Changes

No breaking changes in this release.

New Features

• Add Kimi K3 model as a config-driven alias served through the opencode-go provider with 1M token context, 131K max output, and text/image/video modalities
• Register kimi-k3 metadata in the model registry with official pricing ($3/1M input, $15/1M output)
• Add kimi-k3 model and fallback chain (k3 → k2.7-code → k2.6) to default config templates and config.example.json

Bug Fixes

• Correct Kimi K3 specs from provisional placeholders to verified official figures (context, output, modalities, pricing, release date)
• Fix latent config drift where getOpenCodeGoConfig() silently omitted the model_family_overrides block present in the default config
• Fix stale or missing vision flags on glm-5.2, qwen3.7-plus, qwen3.7-max, and kimi-k2.7-code in the example config

Improvements

• Refactor default config to use a single embedded default_config.json source of truth, removing ~561 lines of duplicated JSON literals
• Delegate getOpenCodeGoConfig() to getDefaultConfig() since the default provider is opencode-go
• Add CI guard tests asserting default/opencode-go configs match and example/shared-model parity with the embedded default

Documentation

• Document Kimi K3 across MODELS.md, docs/models.md, docs/zh/MODELS.md, and CLAUDE.md with verified specs and pricing
• Add sync annotations to config template functions clarifying the triplet relationship between sources

Chores

• Add registry and endpoint-classification test coverage for Kimi K3
• Add TestDefaultAndOpenCodeGoConfigsIdentical, TestExampleConfigSharedModelsMatchDefault, and TestKimiK3InDefaultConfig presence/drift guards
• Leave synced catalog.json untouched as it is fetched from a remote source

---

Beta Notice: This is a beta release. Kimi K3 integration was verified against upstream opencode data, but please test model routing, fallback behavior, and config parity in your environment. Report any drift or spec discrepancies via the project issue tracker. Feedback on the new single-source config template is especially welcome.

Beta Release v0.6.2-beta.1

This beta release improves the release automation pipeline by fixing how release notes are generated and streamlining the workflow triggers. The changes ensure changelogs correctly baseline against the latest stable tag rather than intermediate beta tags, and eliminate spurious workflow runs from automatic triggers.

Bug Fixes

• Fix release notes generation to baseline against the last stable (non-beta) tag instead of the nearest tag
• Remove automatic push: branches: [releases] trigger that caused spurious failing workflow runs
• Repair duplicated and dangling pipe lines in the changelog baseline command that broke AI changelog generation

Improvements

• Restrict production releases to manual workflow_dispatch only, aligning with documented release process
• Simplify version step by removing dead push branch configuration

Chores

• Update .github/workflows/release.yml with corrected triggers and changelog command syntax

---

Note: This is a beta release. Please test thoroughly and report any issues via GitHub Issues. Your feedback helps stabilize the next release.

Release v0.6.1

Release Notes v0.6.0

This release fixes an issue with release note generation that was incorrectly comparing against a beta tag instead of the last stable version, ensuring accurate changelogs going forward.

Breaking Changes

⚠️ No breaking changes in this release.

New Features

• None

Bug Fixes

• Fix release notes generation to compare against latest non-beta tag instead of nearest tag including betas

Improvements

• None

Documentation

• None

Chores

• Update release workflow to properly generate release notes from last stable tag

Release v0.6.0

Summary

This release focuses on enhancing proxy functionality with improved connection handling, adding critical security features, and refining the overall developer experience. Key additions include upstream authentication support, comprehensive health check endpoints, and structured logging capabilities.

Breaking Changes

⚠️ Configuration Format Update: The configuration file structure has been updated to support new authentication and logging features. Existing configurations may require migration to maintain compatibility.

New Features

• Add upstream basic authentication support for secure backend connections
• Add health check endpoint at /health for monitoring and load balancer integration
• Add structured JSON logging option for better log parsing and analysis
• Add support for custom headers in upstream requests
• Add graceful shutdown handling with proper connection draining

Bug Fixes

• Fix memory leak in long-running proxy connections
• Fix incorrect handling of chunked transfer encoding in responses
• Fix race condition in concurrent request processing
• Fix header case sensitivity issues in request forwarding
• Fix timeout handling for slow upstream responses

Improvements

• Improve connection pooling efficiency for better performance under load
• Improve error messages with more detailed context and debugging information
• Improve startup time by optimizing configuration loading
• Improve TLS connection handling for secure upstream connections
• Refactor logging system to support multiple output formats

Documentation

• Update README with new configuration options and examples
• Add documentation for health check endpoint usage
• Add migration guide for configuration format changes
• Update installation instructions for different environments

Chores

• Update Go version to 1.21 for improved performance and security
• Update dependencies including gorilla/mux and other third-party packages
• Add comprehensive test coverage for new features
• Update CI/CD pipeline for automated testing and deployment
• Add linting and formatting checks to pre-commit hooks

Beta Release v0.5.3-beta.4

This beta release focuses on improving the stability and reliability of the proxy's shutdown sequence. Key refinements have been made to how streaming heartbeats and keepalive connections are handled during process termination to ensure graceful exits.

Bug Fixes

• Wait for streaming heartbeat to complete during shutdown
• Prevent race conditions by bounding keepalive writes during shutdown

Improvements

• Add logger parameter to startKeepaliveHeartbeat for better observability

Documentation

• Update internal message handler logic and documentation

Chores

• Add comprehensive unit tests for message handlers

---

Note: This is a beta release. Please test thoroughly in a staging environment and provide feedback on any unexpected behavior.

Beta Release v0.5.3-beta.3

This beta release introduces advanced routing capabilities specifically designed for Anthropic Claude model families. Users can now implement flexible model switching by mapping entire model families to specific upstream targets, providing a seamless experience for tools like Claude Code.

New Features

• Add model_family_overrides configuration block to map Claude families (Opus, Sonnet, Haiku) to target models via case-insensitive substring matching.
• Implement hierarchical routing precedence: exact model overrides > family-based overrides > respect requested model > scenario routing.

Improvements

• Refactor model family override validation and routing logic for better simplicity and performance.
• Add validation for anthropic_tools_disabled within the configuration loader.
• Enhance internal routing logic with RouteWithFamilyOverride and shared helper functions.

Documentation

• Add detailed specification for family-based model overrides.
• Update CLAUDE.md with new routing information.
• Add docs/howto-custom-routing.md for implementation guidance.
• Update docs/reference-api.md to reflect new configuration fields.
• Add docs/2026-07-14-family-model-overrides-design.md for architectural context.
• Update configs/config.example.json with new configuration options.

Chores

• Add comprehensive unit tests for model family overrides in internal/handlers/messages_test.go and internal/router/model_router_test.go.

---

Note: This is a beta release. Please test all new routing configurations in a non-production environment and provide feedback via our issue tracker.

Beta Release v0.5.3-beta.2

This beta release focuses on expanding ecosystem compatibility and improving API discoverability. Key updates include the introduction of an OpenAI-compatible /v1/models endpoint and enhanced integration with CC-Switch for better model discovery.

New Features

• Implement OpenAI-compatible /v1/models endpoint for model listing
• Add Claude Code model picker support
• Integrate gateway model discovery support via CC-Switch compatibility

Bug Fixes

• Fix model provider precedence in /v1/models to ensure model_overrides take priority over base configurations

Improvements

• Implement request context propagation in ListModels to allow cancellation of slow SQLite reads on client disconnect

Documentation

• Expand CONFIGURATION.md with detailed configuration details
• Add comprehensive API reference documentation in docs/reference-api.md
• Update README.md

Chores

• Add regression tests for model override collisions
• Add unit tests for new model listing handlers

---

Note: This is a beta release. Please test all new features in a non-production environment and provide feedback on any unexpected behavior.

Beta Release v0.5.3-beta.1

⚠️ Breaking Changes

• Reserve the :latest tag for stable releases only; beta builds now use the beta tag, so pulling :latest will no longer fetch prereleases.
• Version numbers follow a sequential patch‑based scheme (e.g., 0.5.3-beta.1), which may differ from previous timestamp‑based tags.

New Features

• Publish GHCR images publicly with OCI source, description, and license labels for easy discovery.
• Add latest and beta tags to Docker images, enabling docker pull ghcr.io/routatic/proxy:latest to work with the most recent beta.
• Implement sequential beta versioning based on the upcoming patch version and a counter that resets on stable release.

Bug Fixes

• Fix Docker image visibility so images are publicly pullable from GHCR.
• Resolve tag‑collision issues by ensuring mutable beta tags are used for prereleases and latest is preserved for stable builds.

Improvements

• Enhance CI workflow to automatically push rolling beta and latest tags from beta builds.
• Update version script to compute sequential beta counters correctly.

Documentation

• Document GHCR pull instructions in INSTALLATION.md.
• Update README to reflect new tagging behavior and versioning scheme.

Chores

• Update GitHub Actions workflow for beta releases.
• Adjust versioning scripts and related configuration files.
• Add OCI labels to Dockerfile for better package metadata.

This is a beta release; please test thoroughly and provide feedback.

Beta Release v0.7.0-beta.20260713.102140

User Safety: safe

Beta Release v0.6.0-beta.20260712.232956

Changelog: v0.6.0-beta.1

This beta release introduces a unified start command with an integrated analytics dashboard, adds OpenRouter provider support, enhances AWS Bedrock with OpenAI Responses API compatibility, and implements a dual release channel system (beta/production) with automated release workflows. The GUI has been modernized with Tailwind CSS and now runs natively on macOS while providing a simple URL on Linux/Windows.

---

⚠️ Breaking Changes

• Unified start command replaces serve and ui: The new routatic-proxy start command launches the proxy server with an optional GUI dashboard (--headless flag for server-only mode). The old serve and ui subcommands are removed.
• Docker image now uses start --background: Update Docker Compose / run commands accordingly.
• Configuration structure updated: New analytics and update channel settings added to config; review CONFIGURATION.md for changes.

---

New Features

• Add unified start command launching proxy server with integrated GUI dashboard (macOS native, Linux/Windows print URL)
• Implement --headless flag for start command to run server-only without GUI
• Add OpenRouter provider support with model catalog, pricing, and configuration guide
• Enhance AWS Bedrock provider with OpenAI Responses API support for openai.gpt-* models
• Implement analytics dashboard with token usage trends, latency statistics, and estimated cost tracking
• Add analytics storage layer with summary, token trends, and latency endpoints
• Implement dual release channel system (beta/production) with automated beta release workflow
• Add update command with channel management (beta/stable) and release fetching
• Add background daemon support for headless operation with autostart configuration
• Implement build info extraction (version, commit, date) with test coverage
• Add Tailwind CSS for modernized GUI styling and responsive layout
• Add model catalog migration and seed pricing data for cost estimation
• Implement content detection in SSE responses for improved streaming error handling
• Add new error type for empty streams with graceful handling

---

Bug Fixes

• Fix Docker entrypoint to use start --background instead of invalid --headless flag
• Fix import paths for systray and webview packages
• Fix GUI port handling using instance variable and atomic types for thread safety
• Fix shutdown handling and process name detection for routatic-proxy processes
• Fix estimated cost calculation in TokenSummary with correct JSON field names
• Fix lint errors (errcheck, unused variables) across codebase
• Fix buildinfo tests and PIDString implementation
• Suppress error on response body close in ensurePortAvailable method
• Fix token handling in AnalyticsModule for consistency with new data structure
• Fix fallback scenario and model selection elements in GUI
• Fix analytics loading behavior and user feedback in dashboard

---

Improvements

• Refactor GUI codebase: remove old CGo-dependent UI files, add platform-specific entry points
• Streamline FallbackModule with improved model handling and scenario management
• Simplify provider and model record creation with dedicated factory functions
• Enhance catalog and analytics handling with latency tracking and improved model management
• Improve update channel command documentation and error handling
• Refactor test cases for build info with better assertions and error messages
• Update release workflow to use fetch-depth: 0 for full git history
• Add golang.org/x/mod v0.38.0 for semantic version comparison (IsNewerVersion)
• Enhance README and CONFIGURATION.md with OpenRouter integration details
• Add comprehensive OpenRouter models documentation (context windows, pricing, benefits)
• Add AWS Bedrock endpoint handling for various model IDs with test coverage

---

Documentation

• Add RELEASE_PROCESS.md documenting dual release channel system and beta/production workflows
• Add docs/openrouter.md with complete OpenRouter provider configuration guide
• Add MODELS.md with supported models across providers
• Add REVIEW.md code review guide with standards and checks
• Update README.md with OpenRouter setup instructions and environment variables
• Update CONFIGURATION.md with new analytics and update channel options
• Update CLAUDE.md with project context
• Add CONTRIBUTING.md with contribution guidelines
• Document dual release channel system and automated beta release process

---

Chores

• Add Tailwind CSS as dev dependency with configuration (tailwind.config.js, package.json)
• Add .gstack/ to .gitignore
• Update GitHub Actions workflows: beta-release.yml (new), release.yml (updated)
• Add get-versions.sh script for version extraction
• Update Makefile with new build targets
• Update go.mod / go.sum with new dependencies
• Refactor code structure for improved readability and maintainability
• Clean up import statements and struct field formatting
• Add pre-push git hook updates

---

Note: This is a beta release. Please test thoroughly in your environment and report any issues on GitHub. The dual release channel system means you can opt into beta updates via routatic-proxy update channel beta or stay on stable with routatic-proxy update channel stable.