Applications

A bring-your-own-data, bring-your-own-LLM job application tracker backed by your own Google Sheet—now as a native desktop app.

Key Features

• Applications in your Google Sheet: track applications with status history stored in your own spreadsheet.
• AI auto-fill: fill the form from a pasted job posting with Gemini, OpenAI, Anthropic, or any OpenAI-compatible server (including local LLMs via LM Studio).
• Analytics: status and company breakdowns, a draggable pipeline Sankey, and a status timeline grouped by day, week, or month.
• Printable report for any date range.
• Desktop-native: runs as an Electron app with secure local token storage and file-based logging.

System Design

flowchart LR
  subgraph electron["Electron App"]
    main["Main Process<br/>(Node.js)"]
    renderer["Renderer Process<br/>(React SPA)"]
    preload["Preload Script<br/>(contextBridge)"]
  end
  subgraph local["Local Storage"]
    tokens[("Encrypted Tokens<br/>(safeStorage)")]
    logs[("Log Files<br/>(daily rotation)")]
    config[("Spreadsheet Config")]
  end
  google["Google OAuth, Sheets,<br/>Drive APIs"]
  llm["LLM Provider<br/>Gemini / OpenAI / Anthropic / LM Studio"]
  browser["System Browser"]

  renderer <-->|IPC| preload <-->|IPC| main
  main --> tokens
  main --> logs
  main --> config
  main -->|OAuth loopback| browser -->|auth code| main
  renderer -->|"fetch (REST)"| google
  main -->|"fetch (proxied via IPC)"| llm

Loading

Details on architecture and data flow are in .claude/DESIGN.md.

Local Dev Prerequisites

• Node.js >= 22 and npm
• A Google Cloud project with the Sheets and Drive APIs enabled
• OAuth credentials:
  • OAuth 2.0 Client ID of type "Desktop app" (includes client secret)
  • OAuth consent screen with your account added as a test user

Configuration & Environment Variables

Set these in .env (template: .env.example). electron-vite inlines them into the build, so rebuild or restart npm run dev after changing them. Because the client secret ends up in out/ and dist/, both are gitignored.

Variable	Required	Notes
MAIN_VITE_GOOGLE_CLIENT_ID	Yes	Desktop OAuth Client ID
MAIN_VITE_GOOGLE_CLIENT_SECRET	Yes	Desktop OAuth Client Secret

LM Studio Configuration

Self-hosted models are called from the app's main process. Nothing LLM-related is bundled with the application.

1. Load the model: in LM Studio's Developer tab, load a chat model (7B+). Set Context Length to 8192 or more. From CLI: lms load <model> --context-length 8192.
2. Start the server. CORS settings don't matter: requests come from the app's main process.
3. Connect the app: in Settings → AI Provider, pick Self-hosted (OpenAI-compatible), set URL to http://localhost:1234/v1/chat/completions, click Discover, and select the model.
4. Leave Structured Output toggle off. The JSON schema is sent with every request as response_format.
5. Nothing to set for thinking. Requests send reasoning_effort: "none" (LM Studio) and chat_template_kwargs.enable_thinking: false (vLLM, llama.cpp). The schema constrains only the answer, not reasoning, so a model left thinking can loop until the request times out (240s). Output is capped at 2048 tokens.

To check a model outside the app (expect reasoning_tokens: 0 and JSON in content):

curl -s localhost:1234/v1/chat/completions -H 'Content-Type: application/json' -d '{
  "model": "<model>",
  "messages": [
    {"role": "system", "content": "Extract the job title and company."},
    {"role": "user", "content": "Acme Corp is hiring a Senior Go Engineer in Austin."}
  ],
  "temperature": 0,
  "max_tokens": 2048,
  "reasoning_effort": "none",
  "response_format": {
    "type": "json_schema",
    "json_schema": {
      "name": "job_posting_extraction",
      "strict": true,
      "schema": {
        "type": "object",
        "properties": {"position": {"type": "string"}, "companyName": {"type": "string"}},
        "required": ["position", "companyName"],
        "additionalProperties": false
      }
    }
  }
}'

Operational Runbook

Initial Setup

# Clone and install dependencies
git clone git@github.com:runyanjake/applications.git && cd applications
npm install

GCP OAuth Setup

1. APIs & Services → Library: enable the Google Sheets API and Google Drive API.
2. Google Auth Platform → Branding / Audience / Data Access:
   • User type: External (or Internal on Workspace).
   • Scopes: userinfo.email, userinfo.profile, .../auth/spreadsheets, .../auth/drive.readonly.
   • Test users: add your Google account.
   • While the app is in Testing, refresh tokens expire after 7 days, so expect to sign in weekly. Publish app (unverified, with a warning screen and a 100-user cap) removes that limit.
3. Clients → Create client: type Desktop app. No redirect URIs are needed, because desktop clients accept any http://127.0.0.1 loopback port (the app uses 8085). Copy the Client ID and Client Secret.

Development

# Configure credentials (see Configuration above)
cp .env.example .env

# Start Electron in development mode (hot reload for the renderer;
# restart after changing main/preload code or .env)
npm run dev

Linting & Type Checking

npm run lint
npm run typecheck

Building

# Compile main, preload and renderer into out/
npm run build

# Run the compiled build without packaging
npm run start

# Package for distribution (output in dist/)
npm run package           # current platform
npm run package:mac       # macOS (.dmg, .zip)
npm run package:win       # Windows (.exe, portable)
npm run package:linux     # Linux (.AppImage, .deb)

Installing / Running the Packaged App (macOS)

# Run directly from the build output
open "dist/mac-arm64/Job Application Tracker.app"

# Or install: open the .dmg and drag the app to Applications
open dist/Job\ Application\ Tracker-*-arm64.dmg

The app is unsigned. Builds you make locally run without a Gatekeeper prompt. A copy downloaded or transferred from elsewhere must be opened once via right-click → Open. On first sign-in, macOS may ask for Keychain access; this is safeStorage encrypting the stored tokens.

Common Operations

# View application logs (macOS)
tail -f ~/Library/Application\ Support/job-application-tracker/logs/app-$(date +%F).log

# View application logs (Linux)
tail -f ~/.config/job-application-tracker/logs/app-$(date +%F).log

# Clear stored credentials (macOS)
rm -rf ~/Library/Application\ Support/job-application-tracker/secure-data/

# Clear stored credentials (Linux)
rm -rf ~/.config/job-application-tracker/secure-data/

# Filter error logs
jq -c 'select(.level == "error")' ~/Library/Application\ Support/job-application-tracker/logs/*.log