Upwork Job Matcher

AI-powered Chrome extension that scores Upwork job listings against your profile using a hybrid scoring system — rules-based analysis combined with optional LLM intelligence.

---

Screenshots

Job Cards with Score Badges

Score badges and dollar-hint flag pills are injected directly into Upwork job cards:

Extension Popup

Quick stats and extension status at a glance:

Overview	Stats

Settings Panel

Configure your profile, AI provider, scoring rules, and display preferences:

Profile	AI Provider	Scoring & Rules

Logs & Token Usage	Display

---

Features

• Hybrid Scoring — Combines fast rules-based analysis (budget, proposals, location, posting time) with deep LLM evaluation (skills match, experience alignment, project scope)
• Real-Time Badge Injection — Score badges and colored flag pills appear directly on Upwork job cards as you browse
• Smart Budget Analysis — Sophisticated hourly rate and fixed-price comparison against your personal rate/range, with fallback thresholds
• 5 AI Providers — OpenAI, Anthropic (Claude), Google Gemini, Z.ai, and custom OpenAI-compatible endpoints
• Stack-Based Optimization — Skip LLM analysis for jobs that don't match your tech stack keywords, saving API costs
• Custom Scoring Rules — Define your own rules with operators (<, >, range, contains, in) and point values
• Adjustable Weights — Tune the rules-to-LLM weight ratio (default: 40% rules, 60% LLM)
• Request Queue — Background service worker manages up to 3 concurrent LLM calls with automatic queuing
• Token Usage Tracking — Monitor API costs per provider and model
• Dark Mode — Full dark mode support across settings and popup
• On/Off Toggle — Floating toggle button on Upwork pages to enable/disable scoring instantly
• Manual Analysis — Click to trigger LLM analysis on optimization-skipped jobs
• Detailed Tooltips — Hover over score badges to see rules score, AI score, final weighted score, and reasoning

---

How It Works

1. Scrape — The content script observes job cards on Upwork via MutationObserver and extracts title, budget, skills, location, proposals, and posting time
2. Rules Score — Applies inlined rules-based scoring (proposals, budget comparison, client location, posting recency) → 0–100
3. LLM Score — Sends job data + your profile to an AI provider for skills and experience matching → 0–100
4. Combine — Final weighted score: (rulesScore × rulesWeight + llmScore × llmWeight) / 100
5. Display — Injects a score badge with separate R: and AI: chips plus colored flag pills into each job card

---

Installation

Manual Install (Developer Mode)

1. Download or clone this repository
2. Open chrome://extensions/ in Google Chrome
3. Enable Developer mode (toggle in the top-right corner)
4. Click Load unpacked
5. Select the project root folder (the folder containing manifest.json)

The extension icon will appear in your toolbar. Navigate to upwork.com/nx/find-work/ to start scoring jobs.

---

Configuration

Quick Start

1. Click the extension icon → Settings (gear icon)
2. Set your profile — Skills, experience, preferred hourly rate, and fixed-price range
3. Add an API key — Choose a provider (OpenAI recommended) and paste your key
4. Browse jobs — Visit Upwork's find-work or search pages; scores appear automatically

Settings Sections

Section	Description
Profile	Your skills, experience level, target countries, hourly rate, and fixed-price range
AI Provider	Provider selection, API key, model picker (with dynamic model fetching)
Scoring	Rules/LLM weight slider, custom rules editor, stack keywords
Display	Score thresholds (green/yellow), dark mode toggle
Logs	Optional request logging (up to 200 entries), token usage per provider

---

Scoring System

Rules-Based Scoring (40% by default)

Raw points (max 130, normalized to 0–100):

Factor	Criteria	Points	Flag
Proposals	< 5	30	Low Competition
	5–14	20	—
	15–29	10	—
	30–49	5	High Competition
	50+	0	High Competition
Hourly Rate (user rate set)	Your rate below client minimum	0	Below Minimum
	Budget friendly (low → midpoint)	30	Budget Friendly
	Near top of range (midpoint → high)	25	Near Top of Range
	Above market (high → 1.5× high)	15	Above Market
	Premium (> 1.5× high)	5	Premium
Hourly Rate (no user rate)	$30–80/hr	30	Budget Match
	$81–150/hr	20	Budget High
	$20–29/hr	15	Budget Low
	<$20/hr	0	Budget Low
Fixed Price (user range set)	Below your minimum	0	Not Viable
	Your min → midpoint	20	Acceptable
	Midpoint → your max	30	Good Fit
	Above your max	35	Premium Opportunity
Fixed Price (no user range)	$100–500	30	Budget Match
	$501–1,000	20	Budget High
	>$1,000	15	Budget High
	<$100	5	Budget Low
Client Location	US, UK, Australia	25	Top Location
	Canada	15	Good Location
	Western Europe	10	OK Location
	Other	0	—
Posting Time	< 1 hour	15	Just Posted
	1–6 hours	10	Recent
	6–24 hours	5	—
	> 1 day	0	Old Post

LLM Scoring (60% by default)

When enabled, job data and your profile are sent to the selected AI provider for deep analysis:

• Skills match — How well the job's required skills align with your profile
• Experience alignment — Whether the experience level matches yours
• Project scope — Feasibility and relevance of the project description

Final Score

Final = (rulesScore × rulesWeight + llmScore × llmWeight) / 100

Score Range	Color	Meaning
80–100	Green	Strong match — apply
50–79	Yellow	Moderate match — review
0–49	Red	Weak match — likely skip

---

Supported AI Providers

Provider	Models	Notes
OpenAI	GPT-4o, GPT-4o Mini, GPT-3.5 Turbo	Models fetched dynamically from API
Anthropic	Claude 3.5 Sonnet, Claude 3.5 Haiku, Claude 3 Haiku	Models fetched dynamically
Google Gemini	Gemini 2.0 Flash, Gemini 1.5 Pro, Gemini 1.5 Flash	Models fetched dynamically
Z.ai	Z1 Pro, Z1, Z1 Mini	Models fetched dynamically
Custom	Any OpenAI-compatible endpoint	User provides URL and model name

Models are fetched dynamically from each provider's API when you select a provider in settings.

---

Development

Project Structure

upwork-extension/
├── manifest.json                 # Extension manifest (MV3)
├── background/
│   └── service-worker.js         # LLM API calls, request queue, token tracking
├── content/
│   ├── content.js                # Job card scraping, rules scoring, badge injection
│   └── content.css               # Badge and flag pill styles
├── settings/
│   ├── settings.html             # Configuration UI
│   ├── settings.js               # Settings logic, model fetching, live preview
│   └── settings.css              # Settings styles (light + dark mode)
├── popup/
│   ├── popup.html                # Extension popup
│   ├── popup.js                  # Popup logic (stats, status)
│   └── popup.css                 # Popup styles
├── shared/
│   ├── constants.js              # Storage keys, defaults, provider configs
│   └── scoring.js                # Pure rules-based scoring functions
├── icons/
│   ├── icon16.png
│   ├── icon48.png
│   └── icon128.png
└── LICENSE.md                    # MIT License

Tech Stack

• Vanilla JavaScript — ES6 modules, no build process or frameworks
• Chrome Extension Manifest V3 — Service worker, content scripts, chrome.storage API
• No dependencies — All code is loaded directly by Chrome

Development Workflow

1. Load the extension via chrome://extensions/ → Load unpacked
2. Edit files in your editor — changes are reflected immediately
3. Reload components:
   • Service worker: Click "Reload" on the extension card (or the service worker link)
   • Content script: Refresh the Upwork page
   • Settings page: Close and reopen the settings tab

Debugging

Component	How to Debug
Content script	Inspect a job card element → Console shows [UJM] prefixed logs
Service worker	chrome://extensions/ → "Service worker" link opens DevTools
Settings	Right-click settings page → Inspect

Key Constraints

• Content scripts cannot import ES modules — all constants used by content/content.js must be inlined (see STORAGE_KEYS in both shared/constants.js and content/content.js)
• chrome.storage.sync has a 100KB quota (8KB per item)
• Service worker terminates after 30s idle — queue state resets
• MutationObserver may fire multiple times per DOM change — a WeakSet tracks processed cards

---

Contributing

Contributions are welcome! Here's how to get started:

1. Fork this repository
2. Create a branch: git checkout -b feature/your-feature-name
3. Make your changes following the patterns below
4. Test by loading the extension in Chrome and verifying on upwork.com
5. Commit with a clear message: git commit -m "Add: description of change"
6. Push and open a Pull Request

Coding Conventions

• Vanilla JS only — no external dependencies
• ES6 module syntax for background and settings; inlined code for content scripts
• Use STORAGE_KEYS constants from shared/constants.js — never string literals
• When adding a new storage key, inline it in content/content.js as well (content scripts can't import modules)
• Follow existing patterns for API calls, storage reads/writes, and DOM injection

Adding a New User Setting

1. Add storage key to shared/constants.js → STORAGE_KEYS
2. Inline the same key in content/content.js → STORAGE_KEYS
3. Add UI field to settings/settings.html
4. Add DOM ref in settings/settings.js → initDOMRefs()
5. Add load logic in settings/settings.js → loadSettings()
6. Add save logic in settings/settings.js → saveSettings()
7. Add a chrome.storage.onChanged listener in content/content.js if the content script needs to react to changes

Adding a New AI Provider

1. Add to shared/constants.js → PROVIDER_URLS, PROVIDER_MODELS, DEFAULT_MODELS
2. Add a fetch*Models() function in background/service-worker.js
3. Add the provider case in background/service-worker.js → callLLM() switch statement
4. Add the provider option in settings/settings.html provider radio group

---

License

This project is licensed under the MIT License — see LICENSE.md for details.