How dutta-sujoy/ai-resume-analyzer works
Project type and purpose
• This is a single-page, full-stack web application built with Streamlit (frontend/UI) and Python (backend/orchestrator).
• Purpose: let users upload resumes (PDFs), parse them, call Google Generative AI for analysis and matching against job descriptions, then display actionable feedback.
File structure
• Top-level files only:
• .env – stores API key for Google Generative AI
• README.md – project overview and tech stack
• app.py – the Streamlit app that ties everything together (UI + backend logic)
• requirements.txt – lists dependencies (Streamlit, pdfplumber, pytesseract, google-generative-ai client, etc.)
• resume_analyzer.ipynb – Jupyter notebook used for prototyping parsing/AI calls
• No separate “frontend” and “backend” directories: it’s a monolithic Python process that handles both UI and orchestration.
README for technical insights
• Tech Stack confirms:
• Frontend/UI – Streamlit
• Backend logic – plain Python in app.py
• AI service – Google Generative AI (Gemini) via API
• PDF parsing – pdfplumber, fallback OCR with pytesseract
• Config – .env for securing API keys
• Workflow steps called out: parse → AI analysis → match with job description → feedback.
Main components and their interactions
a. Components
1. User’s browser (renders Streamlit UI)
2. Streamlit app (app.py)
– UI layer: file-upload widget, text inputs, result display
– Controller layer: orchestrates parse/analysis/matching
3. PDF Parser module
– pdfplumber for direct text extraction
– pytesseract fallback for scanned images
4. AI Service (external)
– Google Generative AI (Gemini) API endpoint
5. Environment/Config
– .env holds API_KEY, loaded at startup
6. (Optional) Notebook/Prototyping (resume_analyzer.ipynb) – not in production, but useful for design discussion
b. Relationships & Data Flow
1. User → Streamlit UI: uploads PDF, pastes job description
2. Streamlit app → Parser module: extract raw text
3. Parser module → Streamlit app: returns resume text
4. Streamlit app → Google Generative AI API (with API key from .env):
– Summarization and evaluation calls
– Matching calls for job-description alignment
5. AI API → Streamlit app: JSON responses with scores, suggestions
6. Streamlit UI → User: formatted results (tables, charts, course links)
c. Architectural patterns & design principles
• Monolithic serverless-style process running in a single container/host
• Clear separation of concerns: UI controls vs. parsing logic vs. external AI calls
• Externalize secrets via .env, avoid hard-coding keys
• Fallback pattern: try pdfplumber first, then OCR
d. Key technologies & libraries
• Streamlit for rapid UI
• pdfplumber + pytesseract for parsing
• google-generative-ai client for AI interactions
• python-dotenv or built-in os.environ for config