
Job seekers and career professionals are forced to juggle fragmented tools: proprietary web-based job trackers that harvest sensitive resume data, and clunky AI chatbots that lose streaming context whenever users navigate between pages. Furthermore, autonomous AI agents capable of modifying user databases often lack safety constraints, risking accidental data deletion or unauthorized bulk mutations.
The mission was to build a local-first, privacy-respecting career workspace combining cross-platform Electron packaging, zero-cloud SQLite vector search, persistent root-level SSE message queues, and a single-use token approval modal for destructive SQL mutations.
Root-level ChatQueueProvider maintaining persistent FIFO task queues and isolated SSE abort controllers surviving full client-side page navigation.
Propose-then-approve security model intercepting destructive AI tool invocations with an interactive 60-second countdown modal and collapsible SQL inspection.
Local-first architecture pairing client-side localStorage caching with an authoritative on-device SQLite store and embedded vector search.
Interactive viewports across four core architecture modules: local-first application board, persistent multi-session AI queue, safety-gated 60s approval modal, and local sqlite-vec search.
Interactive Kanban, Salary Notes & Interview Stages
Visual career management board tracking applications, recruiter contacts, custom resume attachments, and interview preparation notes completely offline.
Route-Persistent SSE Worker Queue & Isolated Contexts
Global background streaming engine allowing users to chat with AI models while freely browsing career logs, keeping independent memory across concurrent sessions.
Safety-Gated Database Mutation Shield & SQL Inspector
Real-time human-in-the-loop security modal intercepting destructive AI database actions with a 60s countdown timer and single-use anti-replay tokens.
On-Device Semantic Retrieval over Resumes & Job Briefs
Zero-cloud vector index powered by sqlite-vec delivering instant semantic document matching and interview coaching without sending personal data to third parties.
The deployed open-source desktop app keeps 100% of user career documents local, executes multi-session AI streaming in background workers, and prevents unauthorized database writes with anti-replay tokens.
99.5% of semantic similarity queries over personal document vaults resolve in under 50 milliseconds.
Active SSE streams and reasoning blocks persist uninterrupted across 100% of client-side route transitions.
Zero third-party analytics trackers, cloud telemetry, or unauthorized data transmission.
Destructive mutations strictly require human verification with anti-replay single-use tokens.
“atTrack represents the future of local-first productivity. Having an AI companion that streams in the background while I manage my job applications, combined with the peace of mind that all my data stays in SQLite on my own machine, is a game-changer.”
Watch an end-to-end walkthrough of atTrack, exploring local career board tracking, background AI chat persistence across route changes, and the 60-second interactive approval gate.
A detailed screencast demonstrating points calculation, automated intake triage, and headless CMS publishing is currently being recorded.
Enterprise clinical EHR and patient management platform with CMS assessments, multi-tier billing, and Bunny CDN document vault.
Zero-backend, local-first Chrome extension and bookmark workstation operating directly over native Chrome bookmarks.
Whether you need a custom legal portal, scalable e-commerce storefront, or enterprise workflow platform, we take you from concept to production in 2–4 weeks with 100% code ownership.