— CHANGELOG
Product updates
How LabDraft has evolved, release by release — and what’s next on the bench.
— ROADMAP
Where LabDraft is going
What LabDraft already does, what is being improved now, and the product areas worth strengthening next. Future items are directions, not scheduled promises.
Completed10
- Multi-Provider AI GatewayFree-only routing, health checks & failover
- Engineering & General GeneratorsSingle and bulk practical workflows
- Engineering Bulk ModesSequential Accurate and parallel Fast runs
- Bulk Recovery & Partial ExportRetry, manual completion & reserved pages
- PDF & DOCX ExportBrowser-built manuals with code and figures
- Accounts & Private LibrarySave, search, favourite & restore practicals
- Subject Catalog & Universal Search278 subjects, aliases & deep links
- Reference PDF GuidanceBranding, structure and visual hints
- Code Studio & DiagnosticsCode workflows and provider health tools
- Responsive App ShellSidebar, themes and mobile navigation
In Progress4
- Generation ReliabilityProvider recovery, timeouts and clearer errors
- Output Quality ChecksStructure, code and evidence validation
- Export ConsistencyPDF/DOCX parity, pagination and typography
- Account Usage AccuracyDurable daily and monthly signed-in counters
Future Improvements6
- Editable Saved PracticalsRevise sections without regenerating everything
- Bulk Session RecoveryResume interrupted work after a refresh
- Reference-Guided LayoutsCloser bounded matching from uploaded samples
- Per-Figure ControlsRegenerate or replace one figure independently
- Mobile Workstation PolishImprove dense bulk and preview controls
- Broader Verified ExecutionVerify more languages with real runtimes
— RELEASES
Release history
- v2.0.9Current
Traffic Readiness & Interface Polish
- Clearer privacy information explains current data handling
- Shared AI limit support can fail closed when a distributed store is configured, with a launch check for traffic preparation
- Document uploads receive a ZIP expansion check before extraction and account routes avoid unnecessary session refreshes
- Scrollbars follow the site theme across pages and nested controls; Code Studio's terminal appearance picker now uses the same styled, keyboard-accessible menu
- v2.0.8
Code Studio Terminal & Readability
- Code Studio offers macOS, Windows Terminal, PowerShell, VS Code and Kali Linux window appearances without changing execution behavior
- R, Ruby, Lua and Kali Linux Commands join the Code Studio technology picker
- Editor and terminal contrast is clearer in both themes; output images contain only the raw result on a black canvas
- v2.0.7
Search, Interactive Code & Library Polish
- Global search now indexes the complete Engineering subject catalog, including Kali Linux, image processing and account destinations, with relevant generator deep links
- Selecting an unambiguous programming subject can suggest its language while preserving every later manual language choice
- Programming generation adds Auto, User input and Fixed sample behavior; calculators and other app-like tasks must read and validate runtime input instead of hard-coding an example
- Generation progress now reports observable request/checking states and actual elapsed time without invented section progress or completion estimates
- The homepage opens the saved Library directly and rotates six subject examples from a randomized starting practical
- The global catalog expands to 278 subjects with image processing, Python OOP, privacy, AI security, mobile security, cloud security, IoT security and penetration-testing topics
- Explicit saves, structured Library views, signed-in quota checks, monthly usage summaries and terminal-environment consistency complete the account-history reliability work
- v2.0.6
Bulk Input & Capacity Recovery
- Multi-paragraph Engineering aims now use a consistent 2,000-character limit across forms, batch preflight, and the generation API
- Bulk batches check every resolved aim before sending requests, with clear feedback for invalid items
- An optional Automatic recovery setting waits through short free-provider cooldowns and retries while keeping pending items out of failure recovery
- The bulk side queue follows the permanent parsed manifest, including pasted practicals
- v2.0.5
Bulk State & Free AI Failover
- Pending bulk slots remain in the generation queue, while recovery controls appear only for failed practicals
- Provider-specific content refusals can fail over to another approved free provider without disabling the first provider for unrelated requests
- OpenRouter's fallback uses a verified zero-priced free model after the previous free model became unavailable
- v2.0.4
Bulk Generation Status Hotfix
- Pending bulk practicals are no longer displayed as generation failures while their requests are still running
- The bulk result panel now reports active generation honestly and keeps PDF/DOCX actions disabled until the run settles
- Retry and manual-completion controls appear only for genuinely failed slots; permanent numbering and reserved-page recovery remain unchanged
- The production AI model configuration is unchanged: Groq GPT-OSS 120B remains primary with the existing free fallbacks
- v2.0.3
Generator & Export Correctness
- Searchable subject, language, environment, and technology selectors now commit cleanly without reopening after focus restoration
- Generated results are session-only and automatically hidden when generation-defining inputs no longer match
- Java, shell, C, and C++ execution metadata now stays consistent across prompts, previews, terminal figures, PDF, and DOCX exports
- Algorithm practicals now include assumptions, complexity and edge-case counts; generated output is labelled verified only after successful sandbox execution
- Bulk PDF and DOCX export now preserves every requested slot: failures can be retried or completed manually and receive clearly labelled reserved pages in partial exports
- Supabase migration history now matches the audited 14-migration production schema, allowing Git preview reconciliation without replaying schema changes
- Tool-dependent terminal examples reject common evidence contradictions and remain clearly labelled as simulated or unverified until independently observed
- Missing or corrupt figures are omitted cleanly while PDF and DOCX text, numbering, pagination, and problem/aim structure remain coherent
- Engineering and General generators remain overflow-free across compact 320px phones, tablets, and desktop workstations
- v2.0.2
Reliability & Responsive Polish
- Free-only AI routing now fails closed through the approved provider chain instead of silently selecting paid capacity
- Provider health, task-aware routing, bulk distribution, and quota handling were tightened with a sanitized manual verification command
- A protected daily read-only database health request keeps the free Supabase project active without fake users, sessions, or writes
- Homepage product cards, the feedback form, and the shared header now remain collision-free from compact 320px phones through tablet layouts
- Feedback categories now use stable responsive rows, accessible pressed states, and mobile-sized touch targets
- v2.0.1
P1 Generator Experience & Visual Polish
- Execution Environment & Shell Picker — Polished combobox picker supporting Auto-detect, Bash / Shell, PowerShell, CMD, Terminal, Kali Linux, and IDE / Code Editor with secondary OS descriptions
- Terminal Style Figure Skins — Restored macOS and Windows terminal visual figure skins with smart shell-matching defaults and manual overrides across Single and Bulk flows
- Green Volumetric Lighting — Minimal, cinematic atmospheric green/sage ambient illumination with slow GPU-accelerated drift and complete prefers-reduced-motion support
- Language Picker Overhaul — Searchable language combobox with keyboard navigation and interactive technology stack chip tag editor
- Configuration Panel Architecture — Segmented 3-tab configuration panel (Execution & Environment, Structure, Formatting & References) with dynamic natural height expansion
- Full-Width Generator Workstation — Responsive full-width layout with compact subordinate badge practical number input
- Preserved Compatibility — 100% backward compatibility maintained across AI generation prompts, steps parsing, PDF, and DOCX export pipelines
- v2.0.0
User Accounts, Subject Catalog & Complete Platform Rebuild
- Note: v2.0.0 represents the complete platform rebuild shipped continuously since v1.0.6, bringing user accounts, cloud storage, a 268-subject catalog, and parallel generation.
- Accounts & Cloud Library — Authentication and cloud-library infrastructure introduced in v2.0.0
- Accounts & Library — Personal user dashboard (/dashboard) and cloud-saved practicals library (/dashboard/practicals) with version history, revision restoration, and favourites
- Accounts & Library — v1 Practicals REST API with 8 endpoints for cloud practical management and secure authorization
- Global Subject Catalog — 268 verified global subjects across 18 engineering and natural science branches
- Global Subject Catalog — Typeahead subject search matching names, abbreviations (DAA, OOP, SEPM), and aliases, plus private custom subject creation
- Interface & Shell — Global app shell featuring a retractable overlay sidebar, three-zone header, theme controls, and background grain texture
- Generation Engine — Choice between sequential single-voice ('Accurate') and parallel multi-provider ('Fast') bulk generation with up to 3 concurrent runs
- Generation Engine — Redesigned workstation UX with deferred preview, animated form-to-preview transitions, and real-time StageBar progress
- Exports & Figures — Monospaced Courier-Bold code blocks across all exported PDF and Word DOCX documents
- Exports & Figures — Choice of macOS and Windows Terminal visual skins for command-line output figures
- Product Tour & Polish — Interactive 5-step tutorial page (/about) and streamlined 55% shorter homepage fold
- v1.0.6
Product Experience Update
- Universal search with a Ctrl/Cmd+K command palette
- Results pre-select the whole form — nothing to re-pick
- Developer dashboard: provider health, metrics, live requests
- One reusable Generation Progress component across every generator
- 35 new General Practical subjects across three new categories
- Rebalanced hero so all product cards clear the fold
- Brand polish — deeper burnished-copper accent
- v1.0.5
Premium Design System
- Unified drafting-sheet surfaces across all three generators
- Interactive illustrated hero with per-card parallax
- Trust marquee and blueprint wordmark on the homepage
- Comment-free generated code, enforced end to end
- v1.0.4
Quality & Resilience
- Per-provider attempt timeouts so one slow backend can't stall a request
- Daily-quota vs burst rate limits handled differently
- Malformed output cools the offending provider instead of retrying it
- Environment-aware terminal output and auto-generated tool-skin screenshots
- v1.0.3
AI Gateway
- Multi-provider support: Groq, Gemini, OpenRouter, NVIDIA
- Automatic failover with per-provider health cooldowns
- Structured per-attempt logging that never records prompts
- General Practical Lab — first release
- v1.0.2
UI Polish
- Certificate and index front matter for exports
- Premium circular-reveal theme transition
- Distributed rate limits, spend guard and CSP hardening
- Redesigned PDF borders and multi-column index
- v1.0.1
Generation Improvements
- Syllabus file upload and DOCX export
- Bulk generation with combined-manual output
- Engineering-paper design system and theming
- Section-selection checkboxes with a dynamic prompt schema
- v1.0.0
Initial Release
- AI-generated programming practical write-ups
- Aim to theory, algorithm, code and output
- Client-side PDF generation