github 1jehuang/jcode v0.6.0

latest releases: v0.10.2, v0.10.1, v0.10.0...
one month ago

What's New in v0.6.0

152 commits since v0.5.13 - a major release with new features, stability improvements, and a large-scale codebase refactor.

Features

  • File diff mode - Scroll-synced full file view with change highlights and deleted line interleaving
  • Reasoning effort control - /effort command and Alt+arrow keys to adjust model reasoning effort remotely
  • Stream keepalive - Pong heartbeats during provider calls and streaming prevent client disconnects on long operations
  • Cross-provider model switching - Switch between providers seamlessly when credentials are available
  • Configurable provider profiles - Add custom OpenAI-compatible provider endpoints
  • Proactive and semantic compaction - Smarter conversation compaction modes
  • Background task notifications - Get notified when background tasks complete
  • /poke command - Resume model when todos are incomplete
  • /cache TTL command - Control cache time-to-live
  • Notification line - Separate notification line from status line in TUI
  • 32k max output tokens - Doubled from 16k across all providers
  • GPT-5.4 support - Model aliases, routing, and recommendations

Fixes

  • Ctrl+C/D during processing - Now interrupts the current operation instead of quitting
  • Swarm deadlocks - Fixed summary, stop, and session cleanup deadlocks
  • Swarm coordinator - Headless sessions no longer auto-claim coordinator; zombie session detection
  • Server spawn hang - Fixed with ready-pipe notification and readiness waiting
  • Usage API rate limiting - Increased cache TTL and added 429 backoff
  • Selfdev reload - Preserves live session model; fixed repo detection from cwd ancestors
  • Session persistence - Lossless session save/restore
  • Mermaid rendering - Fixed side-panel rendering bugs
  • Provider failover - Robust fallback chain with eager auth checks and expanded error patterns
  • WebSocket payloads - Strip stream/background fields from response.create payloads
  • Default reasoning effort - Changed OpenAI default from xhigh to high

Security

  • Hardened socket and temp file permissions
  • Secured registry file and credential permissions
  • Security preflight checks in CI
  • Isolated auth paths and Copilot config under JCODE_HOME

Performance

  • Optimized TUI scrolling latency and file-diff rendering
  • Optimized TUI rendering for long chat histories
  • Trimmed syntect YAML loading dependency

Infrastructure

  • Major codebase refactor - 60+ commits extracting server, TUI, and app modules into focused files
  • Comprehensive test coverage for mermaid rendering, SVG helpers, and WebSocket payloads
  • Stress test scripts for profiling session creation/destruction
  • Remote build improvements with git metadata preservation

Full Changelog: v0.5.13...v0.6.0

Don't miss a new jcode release

NewReleases is sending notifications on new releases.