Skip to content

CHANGELOG.md

Latest commit

 

History

History
182 lines (147 loc) · 12.4 KB

CHANGELOG.md

File metadata and controls

182 lines (147 loc) · 12.4 KB

Changelog

All notable changes to enowX Coder are documented here.

[0.2.6] — 2026-04-25

Token Optimization — 99% Reduction for Anthropic-format Gateways

  • Prompt caching now works for custom providers: Added api_format field (openai | anthropic) to providers — custom gateways using Anthropic Messages API now get prompt caching, reducing prompt tokens from ~11,700 to ~0 on cache hits
  • Chat history sliding window: Chat path now applies the same context trimming as agent path — max 20 message pairs, 32K char budget, per-message truncation, html:preview blocks stripped. Prevents token bloat on long sessions
  • uses_anthropic_format() method: Centralized routing logic replaces scattered provider_type == "anthropic" || provider_type == "enowxlabs" checks across chat service and agent runner

Gateway SSE Compatibility Fix

  • Event-line fallback for SSE parsing: Some Anthropic-compatible gateways omit the "type" field from SSE data payloads. Parser now tracks the preceding event: line and uses it as fallback — fixes empty responses from proxies like LiteLLM, Claude Desktop gateway, and enowX Labs gateway
  • Applied to both chat SSE parser (chat_service.rs) and agent tool SSE parser (runner.rs)

Non-Streaming Fallback for Unsupported Models

  • Auto-retry without streaming: When a gateway returns an empty stream (message_start → message_stop with no content blocks), the request is automatically retried with stream: false and the full response is parsed synchronously
  • Fixes blank responses for models where the gateway doesn't support streaming (e.g. claude-opus-4.6 on certain proxies)
  • Applied to both chat path and agent path (with tool call support)

Endpoint Resolution Fix for Custom Gateways

  • Preserve /v1 path for custom providers: Previously, all non-Anthropic providers had /v1 stripped from their base URL when building the Anthropic endpoint, resulting in host/messages instead of host/v1/messages. Now only the built-in enowxlabs provider strips /v1; custom gateways keep their full path
  • Fixed in chat service, title generation, and agent runner

Model Listing for Custom Providers

  • Custom providers can now list models: Previously, unknown provider_type slugs (e.g. user-created "my-gateway") returned "Unknown provider type" error. Now routes by api_format — Anthropic-format providers hit {base_url}/models with correct auth headers
  • fetch_anthropic_models now accepts a configurable base URL and auth scheme instead of hardcoding api.anthropic.com

Provider Settings UI

  • API Format selector: New toggle (OpenAI / Anthropic) in Settings for custom providers — choose Anthropic for Claude-compatible gateways to enable prompt caching and correct message serialization
  • Selector shown in both "Add Provider" form and existing provider detail panel
  • Built-in providers (enowxlabs, anthropic) auto-set to Anthropic format

Database

  • Migration 20260424000_provider_api_format.sql: Adds api_format TEXT NOT NULL DEFAULT 'openai' column to providers table. Existing anthropic and enowxlabs providers auto-updated to 'anthropic' format

[0.2.5] — 2026-04-23

Excalidraw Canvas — Collaborative Whiteboard

  • New feature: Full Excalidraw whiteboard integrated as a Canvas tab — switch between Chat and Canvas via segmented control in header
  • Drawings persist per project in SQLite — auto-save every 1 second, auto-load on project switch
  • Dark/light theme syncs with app theme toggle
  • Full Excalidraw toolbar: shapes, arrows, text, freehand, colors, layers, export to disk

AI Canvas — Generate Diagrams with AI

  • New feature: AI prompt bar floating at bottom of canvas — describe what to draw, AI generates Excalidraw elements
  • AI generates native Excalidraw JSON (rectangles, ellipses, diamonds, text, arrows with bindings)
  • Context-aware editing: Existing canvas elements are sent to AI as context — ask "change color of 2021 to yellow" and AI modifies only that element while preserving everything else
  • Provider and model selector in canvas prompt bar (shared with chat)
  • Merge behavior: AI returns full element set for edits, appends for new drawings

Conversation Memory — Smart Context Management

  • New feature: AI now remembers previous messages in the same chat session
  • Loads user messages + agent outputs from DB, interleaved chronologically
  • Sliding window: Max 20 recent message pairs to prevent token overflow
  • Strip html:preview blocks: Widget HTML (5-10KB each) compressed to ~300-500 byte summaries preserving chart data, labels, colors, metrics
  • Token budget: History capped at ~32K chars (~8K tokens), prioritizing recent messages
  • Smart truncation: User messages max 500 chars, assistant max 1500 chars, cut at word boundary
  • summarize_html_widget() extracts Chart.js datasets, labels, colors, metric values from HTML for targeted edits

Provider System Overhaul

  • Enable/disable toggle: Per-provider on/off switch in Settings detail panel — disabled providers hidden from chat dropdown
  • Add Provider flow: "Custom" replaced with "+ Add Provider" button — create unlimited custom providers with name, base URL, API key, model
  • User-created providers appear in sidebar list with enable/disable dot indicator
  • Delete provider resets selection to enowX Labs
  • is_enabled column added to providers table (migration)
  • Manual model add: Input field in Available Models section — type model name, click Add, auto-enabled
  • Manual models persist in DB and appear in list even after Settings reopen (merged with API-fetched models)

Token Optimization

  • MAX_REACT_ITERATIONS reduced: 20 → 10 — prevents excessive API calls per request while still allowing complex tasks
  • Auto-select provider + model on app start: If only 1 provider enabled, auto-selected as default. First available model auto-selected. No manual selection needed.
  • Provider dropdown shows as label when only 1 provider (no unnecessary dropdown)

Dark Mode — Warm Claude Palette

  • Replaced cold pure black (#0a0a0a) with warm Claude-inspired tones
  • Background: #1a1816 (warm dark brown), Surface: #201e1b, Text: #e8e4de (warm off-white)
  • Accent color: #D4845A (terracotta/copper) matching light mode
  • Borders, links, focus rings all use warm tones
  • Danger colors updated to Claude red palette (#E24B4A)

Markdown Table Fix

  • fixMarkdownTables() rewritten — fixes malformed tables from streaming
  • Auto-corrects separator row column count to match header
  • Splits crammed rows (multiple rows on one line)
  • Injects missing separator rows without duplicating between data rows
  • Applied to both agent streaming output and final chat messages

Scroll Bounce Fix

  • Removed gradient overlay conditional mount/unmount (caused layout shift loop)
  • Added isAutoScrolling guard — scroll events from auto-scroll ignored
  • Throttled streaming scroll to 80ms intervals
  • requestAnimationFrame for scroll-to-bottom to avoid layout thrashing

UI Polish

  • Left sidebar font sizes increased: project names, session titles, labels from 12px to 13px
  • Model dropdown: max-height: 360px with scroll for long model lists
  • Footer bar removed for cleaner layout
  • Right sidebar toggle button in header (hidden when sidebar open)
  • enowX Flux tooltip on hover with description and token usage info

[0.2.0] — 2026-04-22

enowX Flux — Generative UI System

  • New feature: enowX Flux — AI generates interactive charts, diagrams, SVG illustrations, dashboards, and widgets inline in chat
  • Integrated Claude.ai's reverse-engineered design system (CSS variables, SVG color ramps, pre-styled form elements, light/dark mode)
  • html:preview code blocks render as live interactive iframes with full JS/CSS support
  • Chart.js, D3, Canvas, and CDN libraries supported inside previews
  • Debounced iframe rendering during streaming (400ms) to prevent flicker
  • Auto-resize iframe to match content height via direct DOM measurement + ResizeObserver
  • Re-measure on window resize for responsive behavior
  • Expand preview to fullscreen modal with source code view and copy button
  • Toggle button in header to enable/disable Flux (saves ~3K tokens per request when off)
  • Custom tooltip on hover explaining what enowX Flux does
  • Enhanced PREVIEW_GUIDE with full Claude guidelines: diagram types, illustrative patterns, interactive examples, UI components, art/illustration

Welcome Screen

  • Welcome screen with greeting + quick action chips (Write, Learn, Code, Personal, Brainstorm)
  • Chips prefill prompt in input bar on click
  • Auto-create project + session when user sends first message (no need to open folder first)
  • Input bar always enabled — no more "Open a folder to start chatting"

Chat Improvements

  • Streaming output: Agent responses now stream token-by-token with blinking cursor, instead of appearing all at once
  • Auto-rename chat: After first AI response, LLM generates a short 2-5 word title (e.g. "Greeting", "React Auth Setup")
  • Copy response: Copy button below each AI response with model name and generation duration
  • Agent info relocated: Model name and duration moved to bottom of response, agent type stays at top
  • Markdown table fix: fixMarkdownTables() utility fixes malformed tables from streaming — auto-corrects separator column count, splits crammed rows, injects missing separators
  • Scroll bounce fix: Removed gradient overlay mount/unmount loop, added isAutoScrolling guard, throttled streaming scroll (80ms)

Sidebar

  • Left sidebar toggle: Click logo icon to collapse/expand with smooth CSS grid animation (0.25s cubic-bezier)
  • Right sidebar toggle: Toggle button inside sidebar header + in chat header when closed
  • Sidebar font size: Increased project names, session titles, and labels from 12px to 13px for readability
  • Footer removed: Cleaner layout, sidebar toggles moved to header

Settings & Providers

  • Custom provider fix: Added "Default Model" input field for Custom/Ollama providers — fixes silent save failure
  • Error feedback: Save errors now shown as red inline messages instead of silent console.error
  • Base URL save: Can now save base URL even when provider doesn't exist yet (auto-creates)
  • Model dropdown scroll: Fixed dropdown clipping — added max-height: 360px with scrollbar for long model lists

Header & Navigation

  • Rename chat: Click pencil icon → inline edit with Enter/Escape
  • Three-dot menu: New Chat + Delete Chat options
  • Delete confirmation: Custom dark-themed confirm dialog (replaced broken native window.confirm)
  • Theme toggle: Dark/light mode switch in header
  • enowX Flux toggle: ON/OFF pill button with status tooltip

Theme & Design

  • Dark mode warm palette: Replaced cold pure black (#0a0a0a) with warm Claude-inspired tones (#1a1816 bg, #e8e4de text, #D4845A accent)
  • Light/dark preview sync: iframe previews now use class-based theme switching (.light/.dark) instead of prefers-color-scheme media query — syncs with app toggle
  • Tauri dialog permissions: Upgraded from dialog:allow-open to dialog:default for full dialog support

Bug Fixes

  • Agent request fix: Fixed parameter mismatch — frontend now wraps run_agent params in request object matching Rust RunAgentRequest struct
  • API key pre-flight: Added validation check before API calls — clear error message instead of cryptic 401
  • First message missing: Fixed race condition where setActiveSessionId triggered DB reload that wiped optimistic user message
  • Clippy warnings: Fixed all Rust clippy warnings — dead code removal, parameter struct refactoring, simplified patterns

Infrastructure

  • Tauri capabilities: Added dialog:default permission for ask/confirm/message dialogs
  • Generate title command: New generate_title Tauri command — LLM generates chat title via non-streaming API call

[0.1.0] — Initial Release

  • Tauri desktop app with Rust backend + React/TypeScript frontend
  • Multi-agent system: Orchestrator, Planner, Coder FE/BE, Security, UX, UI, Tester, Reviewer, Researcher, Librarian
  • Provider support: enowX Labs, OpenAI, Anthropic, Gemini, Ollama, Custom (OpenAI-compatible)
  • Chat with streaming responses
  • Tool execution: read_file, write_file, list_dir, search_files, run_command, web_search
  • Settings modal with provider configuration and model management
  • Project and session management
  • Agent configuration per agent type