Job Search Optimiser

Monorepo with three related projects focused on job search, hiring workflows, and developer portfolio evaluation. Each subproject runs independently and has its own dependencies and runtime.

Subprojects

• BOT_Detector: Bot detection project (see folder for details).
• crawler_agent: Crawls GitHub repos and technical articles, then uses an LLM to generate review feedback.
• Hiring_Assistant/cv-chat-ai: TalentScout hiring assistant chatbot (FastAPI backend + React/Vite frontend).
• HR_Project: JSO HR Intelligence Agent (CLI workflow over a local SQLite candidate database).

Top-Level Structure

• BOT_Detector/
• crawler_agent/
• Hiring_Assistant/
• HR_Project/

Quick Start

🛡️ Part B — Sentinel-JSO: AI Risk Monitoring Agent

Sentinel-JSO is an automated security agent that protects the JSO platform from fake accounts, bot activity, and fraudulent recruitment. It assigns every user a dynamic risk score using a 4-signal weighted model (behavior · account linkage · content · device/IP), auto-escalates high-risk accounts to the Super Admin dashboard, and powers a LangChain + Gemini AI investigation pipeline that lets admins query any flagged account in plain English and receive a full explainable report. Built with Flask, SQLite, and a Faker-generated dataset of 200 users, 1,800+ activity logs, and 15 bot clusters.

→ See /BOT_Detector/README.md for full setup and documentation.

1) crawler_agent

What it does

• Fetches GitHub repo metadata, READMEs, and recent commits
• Optionally fetches and summarizes technical articles
• Produces review feedback using a configurable LLM provider

Entry point

• crawler_agent/Github Crwaler Agent/main.py

Config

• crawler_agent/Github Crwaler Agent/config.yaml

Run (PowerShell)

cd "crawler_agent/Github Crwaler Agent"
python main.py

Notes

• Required Python packages are inferred from imports in crawler_agent/ (examples: requests, PyYAML, newspaper3k, langchain-*).
• Update config.yaml with your GitHub username, provider, model, and API keys.

---

2) Hiring_Assistant (TalentScout)

What it does

• Screens candidates by collecting required info
• Generates 3 to 5 technical questions based on candidate tech stack
• Supports OpenAI, Anthropic, and Gemini providers
• See /Hiring_Assistant/README.md for full setup and documentation.

Backend

cd "Hiring_Assistant/cv-chat-ai/backend"
python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt
uvicorn main:app --reload --port 8000

Frontend

cd "Hiring_Assistant/cv-chat-ai"
npm install
# create .env in this folder with:
# VITE_API_URL=http://localhost:8000
npm run dev

---

3) HR_Project (JSO HR Intelligence Agent)

What it does

• Conversational HR assistant to query a local SQLite candidate database
• Supports SQL, semantic search, hybrid search, comparisons, explanations, and stats
• Uses Anthropic Claude for intent classification and SQL generation
• See /HR_Project/README.md for full setup and documentation.

Run

cd "HR_Project"
python setup_db.py
python main.py

Configuration

• Set ANTHROPIC_API_KEY in a .env file or environment variables.

Known Issues (as of this repo state)

• HR_Project/main.py and HR_Project/orchestrator.py import modules from database/, agents/, and utils/ packages that are not present in this repo. The code is currently in a flat file layout. You may need to update imports (for example, from setup_db import setup_database, from orchestrator import HRAgent, from config import ...) or recreate those package directories.
• HR_Project/config.py points DB_PATH to ../database/jso_hr.db, but the database file is located at HR_Project/jso_hr.db by default.

---

Development Notes

• This repo contains multiple independent runtimes. Create separate virtual environments for Python projects and avoid sharing dependencies across them unless pinned.
• The Hiring_Assistant/cv-chat-ai frontend and backend are designed to run in two terminals.
• The crawler_agent uses direct API calls and LLM providers. Avoid committing real API keys into the repository.

License

No top-level license file found. Individual subprojects may define their own licenses.