Job Circular 0 chars
Master CV 0 chars
WORKFLOW:
Paste both → Gap Analysis → Tailor CV → Cover Letter → Overleaf → PDF → Apply ✓
│
BD ISP blocked?
Use OpenRouter / Mistral / HuggingFace via Vercel Proxy →
Gap Analysis
Run Gap Analysis from the Input tab to see keyword matching
LaTeX CV
Run "Tailor CV →" from the Input tab to generate your optimised LaTeX
Cover Letter
Run "Cover Letter" from the Input tab to generate your tailored letter
Provider comparison
| Provider | Cost | Quality | Where to get key | Good for |
|---|---|---|---|---|
| Claude (Anthropic) | Paid | Best | console.anthropic.com | Best LaTeX accuracy |
| ChatGPT (OpenAI) | Paid | Excellent | platform.openai.com/api-keys | GPT-4o for CVs |
| Gemini (Google) | Free tier | Very good | aistudio.google.com | Free + generous quota |
| Groq | Free tier | Good | console.groq.com | Fastest response, free |
| Chrome AI (Gemini Nano) | 100% Free | Basic | chrome://flags (no key) | Fully offline, private, on-device |
| OpenRouter (via Proxy) | Free tier | Good | openrouter.ai/keys | Many free models, BD-friendly |
| Mistral (via Proxy) | Free tier | Good | console.mistral.ai | Fast European AI, BD-friendly |
| HuggingFace (via Proxy) | 100% Free | Moderate | huggingface.co/settings/tokens | No key needed, BD-friendly |
| Ollama (local) | Free | Depends on model | ollama.com (install app) | Fully offline, private |
🖥 Chrome Built-in AI — Gemini Nano (100% free, runs on your device)
What it is: Google's Gemini Nano runs directly inside Chrome on your PC — no internet needed, no API key, completely private. Best for gap analysis and cover letters. For complex LaTeX CV generation it may hit the context window limit (~4 000 tokens), so use Gemini or Groq for that task.
Step 1 — Requirements (must meet all)
Chrome 127+ Dev, Beta or Canary (not stable release)
Windows 10/11 or macOS 13 (Ventura)+, or Linux
Storage: at least 22 GB free on Chrome profile drive
GPU: > 4 GB VRAM — OR — CPU: 16 GB RAM + 4 cores
⚠ Chrome for Android / iOS / standard ChromeOS: not supported yet.
Step 2 — Enable two flags in Chrome
1. Open a new tab and go to:
chrome://flags2. Search for:
prompt-api-for-gemini-nano → set to Enabled3. Search for:
optimization-guide-on-device-model → set to Enabled BypassPerfRequirement4. Click Relaunch at the bottom of the page
Step 3 — Download the Gemini Nano model
1. Go to:
chrome://components2. Find Optimization Guide On Device Model
3. Click Check for update — model downloads in background (~1.5 GB)
4. Wait for download to finish, then refresh this page
Step 4 — Verify & use
Select Chrome AI — ON-DEVICE free in the provider bar
Status badge turns green: ✓ Gemini Nano ready
No key, no internet needed — click Tailor CV / Gap Analysis
To debug: open DevTools Console → type
await LanguageModel.availability() → should return "available"
Brave Leo note: Brave Leo cannot be called from a web page — it has no JavaScript API exposed to web content. It is a browser sidebar tool only. However, if you run Ollama locally and configure it in Brave's BYOM settings, you can use that same Ollama instance here via the Ollama — LOCAL free provider option.
🚀 Vercel Proxy — Free AI for Bangladesh (ISP bypass)
Why? Some AI APIs (OpenRouter, Mistral, HuggingFace) are blocked or throttled from BD ISPs. A free Vercel serverless proxy bridges your browser → Vercel → AI provider — completely bypassing the ISP block. Your API key travels inside HTTPS, never exposed.
Architecture — How the proxy works
Step 1 — Deploy proxy to Vercel (one time)
1. Create free account at vercel.com
2. New Project → "Deploy from template" or import GitHub repo
3. Create folder
api/ and paste proxy.js inside it4. Add
vercel.json at the root5. Click Deploy → copy your *.vercel.app URL
6. Paste the full endpoint in "Proxy Endpoint URL":
https://your-app.vercel.app/api/proxy7. Click Test ▶ to confirm it's working before running
Step 2 — Use proxy providers here
Select OpenRouter, Mistral, or HuggingFace from the provider dropdown
Enter your Vercel Proxy URL (saved automatically in browser)
Get a free API key from the provider (HuggingFace: optional)
Paste key → Save → run Gap Analysis or Tailor CV
Free Vercel hobby plan: 100 GB-hrs/month · 100k function invocations · plenty for CV tailoring
OpenRouter — best free model selection
1. Sign up at openrouter.ai → API Keys → Create
2. Free models: llama-3.2-3b, llama-3.1-8b, gemma-3-4b, deepseek-r1 (all :free suffix)
3. Select OpenRouter in provider bar, paste proxy URL + key
HuggingFace — completely free, no credit card
1. Sign up at huggingface.co (free)
2. Settings → Access Tokens → New token (read) — OR leave key blank
3. Select HuggingFace in provider bar, paste proxy URL
4. Note: cold-start delay ~20s on first use (model loading)
Setup for each provider
Gemini (recommended for free)
1. Go to aistudio.google.com
2. Sign in with Google → Get API Key
3. Paste key here, select Gemini provider
4. Free tier: 15 requests/min, plenty
Groq (fastest, free)
1. Go to console.groq.com → Sign up free
2. API Keys → Create new key
3. Paste key here, select Groq
4. Use llama-3.3-70b model for best results
Ollama (fully local, free)
1. Install from ollama.com
2. Run: ollama pull llama3.2
3. Select Ollama provider here
4. No key needed — runs on your PC
Compile LaTeX → PDF
A. Overleaf.com → New → Blank → Paste → Recompile
B. Local: pdflatex cv.tex (run twice)
C. VSCode + LaTeX Workshop extension
Mistral (via Proxy, free tier)
1. Sign up at console.mistral.ai → API Keys
2. Select Mistral in provider bar
3. Enter your Vercel Proxy URL + Mistral API key
4. Models: mistral-small-latest, open-mistral-nemo
Master CV plain-text format (copy this template)