Janda_AI-Job-Search-Agent

Janda is a sophisticated, autonomous AI job search and application assistant — a highly practical use case that merges multi-agent orchestration, Retrieval-Augmented Generation (RAG) pipelines, LLM reasoning, resume/CV comparison, and web scraping/search APIs built using open-source and free tools.

Key Components

Component : Functionality

1. Personal Knowledge Base 1. Resume, project docs, LinkedIn profiles, and cover letters to learn the user’s expertise
2. Web Scraper 2. Pulls jobs from websites based on filters like location, modality, and date
3. RAG Engine 3. Matches jobs to user profiles and scores them by relevance
4. LLM Interface 4. Generates a tailored resume, cover letter, and answers application questions
5. UI/UX Layer 5. Lets the user review job matches and request assets

Summary: Tools & Technologies

Module ==> Tech Stack (Free)

1. Knowledge Base ==> LlamaIndex + FAISS + InstructorXL + MongoDB
2. Scraping ==> Playwright + BeautifulSoup
3. LLM Inference ==> Mistral-7B via Hugging Face Transformers
4. Agent Framework ==> CrewAI (multi-agent) + LangChain (tools & logic)
5. Resume/Cover Letter Gen ==> LangChain + Prompt Templates + Knowledge Base
6. RAG + QA ==> LlamaIndex (Production => Haystack) + Sentence Transformers
7. UI ==> Streamlit / Gradio / FastAPI + React

SETUP

1. For Docker setup, bake with COMPOSE_BAKE=true docker compose up --build

Run CI locally

To reproduce the GitHub Actions CI steps locally, use a Python 3.11/3.12 virtualenv and run the following commands:

# create a virtualenv (optional but recommended)
python -m venv .venv
source .venv/bin/activate

# upgrade pip and install dev deps
python -m pip install --upgrade pip
pip install -r requirements-dev.txt

# lint (flake8)
flake8 || true

# static types (mypy)
mypy || true

# run test suite
pytest -q

# security scan (bandit)
bandit -r backend/agents -n 5 || true

Notes:

• The || true entries mirror the CI behavior where we run linters/type-checkers/bandit in best-effort mode; remove || true to make these steps fail on errors.
• Use Python 3.11 or 3.12 to match the CI matrix.