Changelog

Major AI engine upgrade, autonomous shell execution, paste-proof input, session resilience, and a 41% reduction in system prompt tokens.

• improvedAI engine upgrade — significantly faster responses and higher output quality
• newAutonomous shell command execution — Lore can run commands without pausing to confirm
• newProject auto-fix and modernization on lore init — stale patterns cleaned up automatically
• newPaste-proof multi-line input — paste any amount, submit with Ctrl+D
• newGraceful cancel with Ctrl+C — stops current request cleanly without exiting
• newSession resilience — broken or interrupted responses are never written to disk
• newFriendly error messages — clear, actionable output on API and auth failures
• newOnboarding welcome message on first launch
• improvedSystem prompt optimized — 41% token reduction with zero capability loss
• improvedEntry point auto-wiring — Lore detects and uses the correct main file automatically

Streaming responses, deep code review, agentic file reading, smart context selection, LORE.md, shell execution, multi-line input, and autonomous mode.

• newStreaming responses: text appears word by word in real-time
• newDeep code review: reads actual source files before giving analysis
• newAgentic file reading: AI requests and reads specific files mid-conversation (up to 3 rounds)
• newSmart context selection: only relevant wiki pages and code sent per request (40-50% token savings)
• newLORE.md: drop a file in your project root with coding rules Lore follows every session
• newAuto-learned project insights saved to .lore/insights.json
• newAgentic loop: chains multiple steps automatically via auto-apply
• newMulti-line input with Alt+Enter, instant paste for large prompts
• newSession exit summary: duration, messages, tokens, cost, files changed
• new/commit command for AI-generated commit messages
• new/pr command to create pull requests with AI descriptions
• new/sh command to run shell commands directly from Lore
• new/auto-sh to toggle automatic execution of AI-suggested commands
• newAuto-apply mode: press [A]lways to skip confirmations for the session
• newAuto-fix mode: self-check failures auto-retry (on by default)
• newPost-apply summaries with test commands and next steps
• newGoogle OAuth sign-in: one command, pick your account, done
• newSession persistence: zero data loss, crash recovery
• newOne-liner install scripts for Linux/macOS and Windows
• improvedNew AI engine with major quality upgrade
• improvedTUI flicker fixes, smoother rendering, response time display
• improvedInterlocking knot logo, header redesign
• improvedInput box dims to gray while waiting for API response
• improvedSmoother mouse wheel scrolling (1 line at a time)
• improvedSelf-check upgraded to multi-command pipeline per project type
• improvedResponse separator visibility bumped for better readability
• fixSession data no longer lost on unexpected close or crash

Full SaaS infrastructure live. Real authentication, payment processing, free tier, and cross-platform binary releases available to everyone.

• newFree tier: 300K tokens/month, all features, no credit card required
• newFour subscription tiers: Free ($0), Starter ($8/mo), Builder ($15/mo), Pro ($20/mo)
• newFree tier registration endpoint (sign up with email only)
• newReal authentication with Supabase license validation
• newCLI auth commands: lore auth login, status, logout
• newLemon Squeezy payment integration with webhook-based license provisioning
• newServer-side usage tracking and token cap enforcement per tier
• newGoReleaser + GitHub Actions for automated cross-platform builds
• newPublic binary distribution via GitHub Releases (6 platforms)
• newLanding page, documentation, and changelog deployed on Cloudflare Pages
• fixMonthly usage counters now auto-reset on month boundary
• improvedAPI key moved from hardcoded config to encrypted secrets
• improvedToken caps updated: Starter 2M, Builder 6M, Pro 12M
• improvedEvery feature available on every tier, tokens are the only differentiator

Complete safety layer: diff preview, rollback, self-check, and budget enforcement.

• newDiff preview with visual line-by-line diffs before applying changes
• newInteractive apply/preview/reject/teach options on every proposed change
• newGit-based rollback with automatic snapshot commits
• newlore history and lore rollback commands
• newBudget hard caps with monthly enforcement and 80% warnings
• newSelf-check: auto-runs build after applying changes, offers fix or undo on failure
• newRobust JSON parser with character-walker for malformed model output
• improvedCLI visual polish: ASCII cat logo, styled input box, color-coded token meter

Teach mode and learning document system with PDF export.

• newTeach mode with 3 depth levels: brief, standard, deep
• newAuto-saved learning docs in .lore/learnings/ with YAML front-matter
• newPDF export with embedded fonts (Inter, JetBrains Mono), code blocks, TOC
• newlore export-learning command (single or combined export)
• fixShared CleanJSON utility for reliable parsing across all endpoints

The core wiki memory system. Project scanning, context loading, session compilation, and wiki commands.

• newlore init: smart project scan with file structure analysis
• newWiki context auto-loaded on REPL start (index + conventions + architecture)
• newDynamic wikilink loading mid-session
• newWiki compile on exit: conversation distilled into wiki updates
• newWiki commands: /wiki status, search, read, lint
• newStale page detection and wiki health indicators

First working prototype. CLI and round-trip chat.

• newBackend API deployed
• newGo CLI with Cobra + Bubbletea + Lipgloss + Glamour
• newInteractive REPL with chat, token meter, slash commands
• newFull round-trip: CLI to AI engine and back
• newSystem prompts for identity, teaching, wiki operations
• newTOML config with environment variable support