Skip to content

v0.4.0

Choose a tag to compare

@bharat2808 bharat2808 released this 22 Jul 01:09
· 135 commits to main since this release

Features

  • Completion API adaptor for OpenAI Responses-compatible server (Chat Completions-only providers)
  • Flexible preset system for any combination of proxy commands with background execution
  • Session-based inline image persistence and caching for auto-routed images
  • Native Ollama image generation backend via /api/generate
  • Separated image generation config so presets work independently
  • Async skill index prewarming for faster startup
  • Filtering of repeated large Codex instruction blocks

Bug Fixes

  • Stabilized streamed Responses lifecycle ordering on [DONE], EOF, and upstream failures
  • Scoped image auto-routing to active user turn only, honoring auto_route_image = false
  • Hardened inline image persistence with signature validation, size limits, and cache expiry
  • Fixed local Ollama capability discovery when /api/tags omits capabilities
  • Deduplicated translated function tools with turn-local definition precedence
  • Fixed completion adaptor error handling
  • Preserved launcher, adaptor, and runtime override settings across restart
  • Verified and safely replaced proxy listeners on restart
  • Restart proxy by default when starting preset or switching to Ollama
  • Fixed preset behavior and config-driven dedupe errors