ApplyVista — AI-Powered Job Application Assistant

ApplyVista is an intelligent, offline-first job application assistant that generates ATS-ready skills summaries, cover letters, recruiter emails, and follow-up emails — all grounded in your own resume and project documents using a Retrieval-Augmented Generation (RAG) pipeline.

It also discovers engineers at companies you're applying to via GitHub, and runs an automated job scout every 6 hours that searches LinkedIn and Seek for new opportunities and sends email alerts.

---

Architecture

┌─────────────────────────────────────────────┐
│  Next.js Frontend (localhost:3000)          │
│  shadcn/ui + Tailwind                       │
│  Pages: / (apply) + /scout (job scout)      │
└─────────────┬───────────────────────────────┘
              │ HTTP
┌─────────────▼───────────────────────────────┐
│  FastAPI Backend (localhost:8000)           │
│  ├── POST /jd         ← LangGraph pipeline  │
│  ├── POST /export/docx|pdf                  │
│  ├── GET  /download/{filename}              │
│  ├── GET  /scout/jobs                       │
│  ├── GET  /scout/status                     │
│  └── POST /scout/run                        │
└──────┬──────────────────────────────────────┘
       │
┌──────▼──────────────────────────────────────┐
│  LangGraph Pipeline                         │
│  extract_metadata → [generate_docs ||       │
│  github_search] → save_excel                │
└──────┬──────────────────────────────────────┘
       │                     ┌─────────────────┐
       │                     │  APScheduler    │
       │                     │  Every 6 hours  │
       │                     │  ↓              │
       │                     │  Job Scout      │
       │                     │  (DuckDuckGo)   │
       │                     │  ↓              │
       │                     │  Email Alert    │
       └─────────────────────┴─────────────────┘

---

Features

RAG Document Generation

• Indexes your CV and project docs into ChromaDB
• Retrieves the most relevant context per query
• Generates 6 tailored outputs per job description:
  • Skills & Keywords
  • Cover Letter
  • Application Emails (3 variants)
  • ATS Resume Summary
  • "Why this role?" short response
  • Cold recruiter outreach email

LangGraph Orchestration

• Stateful graph replaces the old sequential pipeline
• extract_metadata → parallel fan-out → save_excel
• generate_docs and github_search run concurrently

GitHub Engineer Discovery

• Searches GitHub for engineers at the company in the JD
• Returns name, email, bio, location, top languages, repo count
• Exports results to a 2-sheet Excel file

Job Scout Agent

• Runs automatically every 6 hours via APScheduler
• Searches LinkedIn and Seek using DuckDuckGo
• Scores each result by keyword overlap with your skills
• Deduplicates against previously seen jobs
• Appends new jobs to a persistent Excel file
• Sends a Gmail email alert listing new findings

Next.js Frontend

• Paste a JD → click Generate → 7 tabs appear instantly
• Each tab has an editable text area + Download DOCX / PDF
• Engineers tab shows a GitHub profile table + Excel download
• /scout page shows job scout status, jobs table, and a manual trigger button

Local LLM (Fully Offline)

• Powered by Ollama (Llama 3.2 3B by default)
• No API keys, no cloud calls, no cost, 100% private

---

Tech Stack

Component	Purpose
Next.js 16	Frontend (React, App Router)
shadcn/ui + Tailwind	UI components
FastAPI	Backend REST API
LangGraph	Pipeline orchestration
LangChain	RAG chains and prompt building
Ollama	Local LLM runtime
ChromaDB	Vector store
SentenceTransformers	Embeddings (all-MiniLM-L6-v2)
APScheduler	Job scout scheduling
duckduckgo-search	Job search queries
openpyxl	Excel export
python-docx / fpdf	DOCX and PDF export

---

Project Structure

ApplyVista/
├── app/
│   └── api.py                  ← FastAPI app (all endpoints)
├── src/
│   ├── rag_pipeline.py         ← Core RAG logic
│   ├── langgraph_pipeline.py   ← LangGraph graph
│   ├── github_agent.py         ← GitHub engineer discovery
│   ├── job_scout_agent.py      ← Job search + email alerts
│   ├── scheduler.py            ← APScheduler setup
│   ├── config.py               ← Paths and model settings
│   ├── text_utils.py           ← NLP utilities
│   └── profile_config.py       ← Your private profile (gitignored)
├── frontend/
│   ├── app/
│   │   ├── page.tsx            ← Main JD input + output tabs
│   │   └── scout/page.tsx      ← Job scout dashboard
│   ├── components/
│   │   ├── JDInput.tsx
│   │   ├── OutputTabs.tsx
│   │   ├── DocTab.tsx
│   │   ├── EngineersTable.tsx
│   │   └── ScoutDashboard.tsx
│   └── lib/api.ts              ← Typed fetch helpers
├── data/
│   ├── job_rag/
│   │   ├── profile_docs/       ← Drop your CV and project docs here
│   │   ├── chroma_db/          ← Vector store (auto-generated)
│   │   └── seen_jobs.json      ← Scout dedup store
│   └── outputs/                ← Generated Excel, DOCX, PDF files
├── .env.example
└── requirements.txt

---

Setup

1. Clone

git clone https://github.com/sheikhmunim/ApplyVista.git
cd ApplyVista

2. Python environment

python -m venv .venv
# Windows
.\.venv\Scripts\activate
# macOS / Linux
source .venv/bin/activate

pip install -r requirements.txt

3. Ollama

ollama pull llama3.2:3b
ollama serve

4. Profile config

Copy the template and fill in your details:

cp src/profile_config\(template\).txt src/profile_config.py

Drop your CV (PDF) and any project docs into data/job_rag/profile_docs/.

5. Environment variables

cp .env.example .env

Edit .env:

# Required
OLLAMA_HOST=http://localhost:11434
LLM_MODEL=llama3.2:3b

# Optional — GitHub engineer discovery
GITHUB_TOKEN=ghp_your_token_here

# Optional — Job scout email alerts (Gmail)
SMTP_HOST=smtp.gmail.com
SMTP_PORT=587
SMTP_USER=your_gmail@gmail.com
SMTP_PASSWORD=your_app_password     # Gmail App Password, not account password
ALERT_EMAIL=your_gmail@gmail.com

# Job scout search config
SCOUT_ROLE=ML Engineer
SCOUT_LOCATION=Melbourne

Gmail App Password: myaccount.google.com → Security → App passwords (requires 2FA)

6. Frontend dependencies

cd frontend
npm install

---

Running

Terminal 1 — Backend:

uvicorn app.api:app --reload --port 8000

Terminal 2 — Frontend:

cd frontend
npm run dev

URL	Description
http://localhost:3000	Main app
http://localhost:3000/scout	Job scout dashboard
http://localhost:8000/docs	Swagger API explorer

---

API Endpoints

Method	Endpoint	Description
POST	/jd	Run full LangGraph pipeline
POST	/export/docx	Convert text to DOCX
POST	/export/pdf	Convert text to PDF
GET	/download/{filename}	Download a file from outputs/
GET	/scout/jobs	Get all scouted jobs as JSON
GET	/scout/status	Scheduler status + job count
POST	/scout/run	Manually trigger the scout
POST	/chat	RAG chat over your profile docs
GET	/health	Health check

---

Workflow

1. Paste a job description into the input box
2. Click Generate Documents and wait (~30–60s depending on hardware)
3. Six document tabs appear — each is editable
4. Download any tab as DOCX or PDF
5. The Engineers tab shows GitHub profiles at the company + Excel download
6. Open /scout to see jobs found automatically or click Search Now

---

License

MIT License

---

Author

Sheikh Abdul Munim Master of Artificial Intelligence — RMIT University

LinkedIn · GitHub