Skip to main content

Career · desktop app

collab-finder — job hunt desktop app

Desktop app for high-fit job hunting. You approve before anything hits your master CV or apply files. Hunt screens, pack checks, pipeline, local database. Not a chat app.

For Recruiters and engineers who want job-hunt screens, not another chat window.

  • Evaluate roles, prepare, generate apply PDFs on your machine
  • Check pack files without a terminal
  • Track prep and post-apply outcomes
  • Local database; credentials stay on the machine
  • Named apply PDFs via the CV generate CLI (kanithanj.cv)

Surfaces

  • Discover / Mission / Sweden / Xplore hunt screens
  • Preferences — pack file checks
  • Pipeline — hunt progress
  • CV generate CLI (kanithanj.cv)
  • Local database on this machine
Source on GitHubGitHub release

Product

  • What you do

    Hunt loop

    Open Discover, Mission, Sweden, or Xplore. Paste a job link or description in Quick Target. Evaluate fit, prepare notes, generate an apply CV, mark Applied. Reopen a saved row to load fit and prep from the local database without another model call.

  • Shipped · Preferences

    Pack file checks

    Shows whether your pack files on disk are ready. Evaluate and Next 10 read those files. They are not built into the app. Missing files make Evaluate use placeholder text.

    Example Pack status on Preferences. Not a live machine.

  • Shipped · Pipeline

    Pipeline

    Tracks where each role stands before and after you apply. Applied rows stay visible even when Mission lists get long.

    Example Fixed prep stages and outcome labels. No live counts.

  • Durable state

    Local database

    Stores opportunities, prep, events, and history on this machine. Secrets stay in the OS keyring. No cloud sync. Local backup optional.

  1. Find roles
  2. Evaluate and prepare
  3. Pack files on disk
  4. Generate apply CV (kanithanj.cv)
  5. Track outcome

Job hunting spreads across tabs and chat logs. This app keeps a local record. Chat helps; screens do the work.

Tech and architecture

  • Tauri
  • Rust
  • React
  • TypeScript
  • SQLite

Architecture

  • Tauri desktop shell (Rust + React): UI calls Rust for secrets and file work
  • Local database (SQLite, WAL mode) holds opportunities, prep, events, and pipeline dates
  • Master CV on the public portfolio site is never changed by apply — job edits merge when you generate a PDF

Key components

  • Hunt screens: Discover rail and Quick Target; Mission career boards; Sweden JobTech; Xplore live X plus a guarded cycle
  • Preferences: pack health checks, CV generate CLI install, rank packs. Health is local file inspection
  • Pipeline: dedicated opportunity query, prep status, and separate post-apply outcomes
  • CV generate CLI (kanithanj.cv): status, list, generate, sync. Writes PDFs. Never mutates the master CV on the public portfolio site

Data flow

  1. Opportunity in
  2. Ledger row
  3. CV packet from packs or textarea
  4. Evaluate + Prepare
  5. Human review
  6. Export pack / generate apply CV
  7. Pipeline outcome

Promote to the public site is a separate sync of the master CV. Apply PDFs come from packs; the portfolio PDF uses the master CV only.

Tradeoffs

  • Human approval before promote slows automation and prevents silent CV corruption. That friction is intentional
  • Tauri desktop adds Rust surface area versus a pure web app. The payoff is OS keychain boundaries and a real offline ledger
  • Hunt screens own the work. Quest chat exists but is not the primary UI

Testing and ops

  • CI quality gates ship with the 1.0.0 cut
  • Pack health unit tests cover missing and placeholder identity files
  • Pipeline verify runner plus merge-status tests keep applied rows from being dropped
  • Per-job CV merge behavior is covered in this portfolio repo so PDF featured projects stay deterministic
  • Ops stay boring: local SQLite, packs under ~/.local/share/collab-finder, CLI sync for the master CV

More shipped walkthroughs in this set, or back to the landscape.

All shippedBuilding landscape