Fixed broad user permissions

Bug Fixes

Fixed job preferences to prioritize resume job title over fragmented keywords
Updated notification branding with Offlyn monogram icon
Improved notification title format: Offlyn Apply — 3 New Jobs

UI Improvements

Redesigned Resume Keywords section to be collapsible
Zen, minimal design with single-row collapsed state
Shows match score at a glance, expands for details

Technical

Enhanced preference engine with topSearchPhrases tracking
Backward compatible with existing user data
Full Chrome + Firefox support

updated windows ai setup

updated AI setup instructions

Scheduled Job Discovery

Background job searches run automatically at configurable intervals (1, 4, 8, 12, or 24 hours)
Learns your preferences from search and apply history to find relevant jobs
New jobs are cached and displayed when you open the Find Jobs page
Browser Notifications

Get notified when new matching jobs are found
Click notifications to go directly to the jobs page
Badge indicator shows count of new jobs
Preference Learning

Extension learns from your search terms and applied jobs
Automatically derives preferred keywords, locations, salary ranges
Improves job recommendations over time
Draggable Widget

In-page widget can now be dragged anywhere on the page
Position persists during the session
Panel repositions automatically to avoid clipping off-screen
Fill Mode Progress Ring

When auto-filling forms, widget minimizes to a compact pill
Circular progress ring shows fill progress
Less intrusive, zen-like experience during form filling

windows model updated

Updated Model

Updated the AI model

AI Setup Flow: We've significantly improved the first-time setup experience. The extension now provides platform-specific installers (signed .pkg for macOS, .bat for Windows) that handle Ollama installation automatically. This reduces user friction from complex terminal commands to simple download-and-click.

Multi-Profile System: Added comprehensive profile management allowing users to maintain separate profiles for different job types (e.g., "Senior Engineer", "Product Manager"). Each profile maintains its own resume data, preferences, and compatibility scoring.

Job Discovery Integration: Integrated with Adzuna API for job discovery. Users can find relevant positions based on their profile data. API calls are made client-side with proper rate limiting.

Enhanced Form Filling: Improved compatibility with major job boards including Greenhouse, Workday, and others. The extension now handles more complex form structures and custom fields.

Privacy & Security: All AI processing happens locally via Ollama. No user data is sent to external services except for job search queries to Adzuna (which only include search terms, not personal information).

windows support

Version 0.6.0

Improved permissions: removed unused permissions for a leaner install
Updated in-extension privacy policy for accuracy
General stability and code cleanup

• One-click AI setup — install Ollama and required models directly from the extension, no terminal needed
• Workday autofill improved — smarter field detection, fixed year/company/school fields being filled with wrong values, Field of Study autocomplete now works correctly
• Shopping sites — Offlyn now shows a friendly prompt on checkout pages and only fills name/address (never credit card fields)
• Netflix & Greenhouse — fixed skipped fields including self-ID questions and company-specific questions
• Dark mode — troubleshooting modal now fully night-mode friendly
• AI Setup page redesigned with clearer on-device privacy messaging

Offlyn Apply — Privacy-First Job Application Assistant

Auto-fill job applications in one click. Generate tailored AI cover letters. Track every application automatically. All AI runs locally on your machine via Ollama — your data never leaves your browser.

─── WHAT IT DOES ───────────────────────────────────

AUTO-FILL APPLICATIONS
Detects application forms across 50+ ATS platforms — Greenhouse, Lever, Workday, Ashby, iCIMS, BambooHR, Taleo, SmartRecruiters, and more. Fills every field using your saved profile. Smart field trust tiers ensure cookie banners, chat widgets, and tracking pixels are never touched.

AI COVER LETTERS
Generates a cover letter tailored to each specific role using your resume and the job description. Runs on your local Ollama instance — no API keys, no subscriptions.

GRAPH MEMORY & LEARNING
Builds a private knowledge graph of your answers over time. Learns from every correction you make. The more you use it, the better it knows your preferences for work authorisation, salary, location, and custom fields.

APPLICATION TRACKER
Auto-logs every application you submit — job title, company, date. Deduplicates confirmation pages so your tracker stays accurate.

RESUME CHAT
Ask your resume questions in plain English. Powered by local RAG (Retrieval-Augmented Generation) — no data leaves your device.

DATA EXPLORER
Visual graph explorer showing what the extension has learned: your profile nodes, answer relationships, corrections, and field mappings. Includes Overview, Focus, and Advanced modes.

COMING SOON
DS-160 immigration form filling · DMV & driver's licence renewals · USCIS I-485 · KYC & identity verification

─── PRIVACY ────────────────────────────────────────

✓ All AI runs locally via Ollama — zero external API calls
✓ Data stored in chrome.storage.local only
✓ No accounts, no telemetry, no analytics
✓ 100% open source

─── REQUIREMENTS ───────────────────────────────────

• Ollama installed locally for AI features (ollama.ai)
• Auto-fill works independently without Ollama
• No subscription or account required

Version 0.3.0

Offlyn Apply v0.2.5 — Release Notes
March 3, 2026
Bug Fixes & Reliability
Storage resilience overhaul (onboarding)
The profile save and storage repair flows have been significantly hardened to handle transient Firefox storage errors without unnecessary data loss.
Transient-error fast retry — on any storage read/write failure, the extension now waits 500ms and retries once before escalating. Most Firefox disk I/O and quota-check errors resolve within this window.
New retryStorageRead utility — storage reads are now retried up to 3 times with 300ms between attempts before declaring failure. Profile load on the onboarding init path uses this first, falling back to repair only after all retries are exhausted.
Smarter repairStorage — the repair sequence now backs up the user's profile and all dashboard metrics before any destructive operation, and restores them after a full clear. Users who previously lost their profile during a storage repair will no longer experience data loss.
Targeted resume key cleanup — the repair sequence now removes all resume-related keys (resumeFile, resumeFileMeta, resumeChunk_*, resume_embeddings, resume_chunks, rag_context) as the first escalation step, reserving a full clear as a true last resort.
Migration skips instead of repairs — a failure during the resume file migration check now skips silently and retries next open, rather than triggering a destructive repair on what is likely a transient read error.
Visual
Updated toolbar icons — refreshed branding across all sizes (16×16, 32×32, 48×48, 96×96, 128×128).

Adaptive Learning System
The extension now learns from every form submission, not just corrections. When you submit an application, it snapshots all field values at high confidence and uses them to improve future fills on similar forms (learning-system.ts).
Context-Aware Answer Storage
Multiple variations of answers (e.g. "Why this company?", cover letters, availability) can now be stored per context — industry, company size, seniority level — and the right variation is selected automatically (context-aware-storage.ts).
Tracking Badge
A subtle frosted-glass pill appears bottom-left when a job page is detected, showing the job title and company. It auto-dismisses after 2 seconds and can be manually closed.
Sync Service Foundation
Infrastructure for optional backend syncing of profile and contextual data is now in place for a future cloud sync feature (sync-service.ts).

Live Parse Feed
The onboarding resume upload now shows a live feed as it reads your resume — you can see actual content being extracted word-by-word with a blinking cursor, giving clear visual feedback that parsing is in progress.
Smarter Multi-Page Form Handling
The autofill engine now detects clicks on "Next", "Continue", and "Apply" buttons and automatically waits for the page state to transition before re-scanning for new fields. This means multi-step application forms fill correctly without manual intervention.
Workday Integration
Native support for Workday application forms — the extension now detects Workday pages, identifies the current step, and runs dedicated handlers for complex sections like Work Experience, Education, Languages, and the Skills tag input.
Resume Parse Progress Streaming
Resume parsing now broadcasts real-time progress stages and percentages from the background to the UI, so you always know how far along the AI is.
Application Stats in Popup
The popup now loads live stats (total applications, currently interviewing) directly from storage and displays them on open.
Dark Mode Toggle in Popup
A dark mode toggle button is now available directly in the popup header. ( Thank you @Carlos for the suggestion )
Bug Fixes
Fixed missed click events on buttons with icon/span children (clicks now bubble up to the correct interactive element)
Improved error messages for resume parse failures with better classification
Removed legacy pdf.worker.mjs in favour of the newer worker file

Version 0.2.0 introduces a reinforcement learning engine that improves autofill accuracy over time based on user corrections. A new settings page, compatibility widget, and Ollama configuration panel make the extension easier to set up and use. The UI has been refreshed with the updated Offlyn brand design across all pages. Also includes a new help page, privacy policy page, smarter field matching, and significant stability improvements.

v0.1.2 — Workday support & better error messages

• Workday support — Autofill now works on Workday job applications, including Work Experience, Education, Languages, and Skills sections that the generic engine couldn't reach
• Clearer AI setup errors — When Ollama models aren't installed, the extension now shows the exact terminal commands to run (ollama pull llama3.2) instead of a generic error
• Better PDF parsing — Improved PDF resume loading with fallback handling and clearer error messages if the PDF library fails to load
• Smarter field label detection — Added Workday-specific label detection (data-automation-id attributes) so fields are correctly identified and filled