Skip to content

features.md

Latest commit

 

History

History
99 lines (70 loc) · 6.91 KB

features.md

File metadata and controls

99 lines (70 loc) · 6.91 KB

ResumeForge — Features File

Purpose: Give this file to any AI assistant along with your current resume to generate strong, accurate resume bullet points for the ResumeForge project. Developer: Rishabh Singh | rawatrishi3@gmail.com | github.com/coder-Rishi05

RESUME ENTRY HEADER (copy-paste ready)

ResumeForge — AI-Powered Resume Analyser | React.js, Node.js, Express.js, MongoDB, Groq API (LLaMA 3.3 70B), Multer, Brevo API | Live Demo

FEATURE LIST (for AI to generate bullets from)

Core Pipeline

  • User uploads PDF resume + enters job role → backend extracts text → sends to Groq AI → returns structured JSON analysis + cover email
  • Full pipeline: PDF upload → text extraction → AI processing → structured response → email delivery — all in a single user action
  • Stateless API design — no database, no stored files, no user accounts. Resume processed in-memory and discarded after response

PDF Handling

  • Implemented Multer with memoryStorage — PDF stored entirely in req.file.buffer, never written to disk
  • PDF-only file filter enforced at middleware level (application/pdf MIME check) with 5MB size limit
  • Used pdf2json library to extract raw text from buffer with custom safeDecode for URL-encoded characters

AI Integration — Groq / LLaMA 3.3 70B

  • Engineered a dual-purpose prompt: generates both a tailored job application email AND section-wise resume analysis in a single API call
  • Prompt instructs model to return strict JSON only — no markdown, no preamble — with full schema definition in the prompt
  • AI output includes: overall score (0-100), section scores (0-10 each), strengths, critical gaps, missing ATS keywords, suggested rewrites (current vs improved side-by-side)
  • Post-processing pipeline: strips markdown fences → regex extracts JSON object → validates required fields → normalises email body formatting
  • Error handling: distinct HTTP status codes for each failure stage (422 extraction, 502 AI failure, 500 email failure)
  • Model config: temperature 0.5 (deterministic output), max_tokens 3000

Email Delivery — Brevo API

  • Integrated Brevo transactional email API (migrated from Nodemailer/Gmail SMTP for production reliability)
  • AI-generated subject line and email body sent directly to HR with PDF resume as base64 attachment
  • Email subject format: "Application for [Job Role] — [Candidate Name]" — generated dynamically by AI from resume content

Frontend — React.js

  • Built multi-page React SPA (Vite) with React Router v6 — 4 pages: Home, Analysis, EmailPreview, Success
  • Passed data between pages using React Router navigate state — no Redux, clean linear flow
  • Built collapsible section cards for Analysis page — each section expandable with score, strengths, suggestions (priority-tagged High/Medium/Low), and side-by-side rewrite comparisons
  • Implemented global dark/light theme system using React Context API — CSS custom properties toggled via data-theme attribute on document.documentElement, persisted in localStorage
  • Custom file upload component with PDF validation, file size display, and visual feedback states
  • Centralized Axios API layer in lib/api.js for all multipart/form-data requests

UI/UX

  • Designed warm editorial UI: DM Sans (body) + DM Mono (labels/tags), off-white light theme with gold (#e8b84b) accent
  • Mouse-tracking ambient glow effect on hero section using mousemove event and CSS radial gradient
  • Animated ticker banner with CSS keyframe animation (translateX(-50%))
  • Responsive grid layouts (3-col feature grid → 2-col → 1-col at breakpoints)
  • Smooth theme transitions (0.25s ease) on all color properties

METRICS / QUANTIFIABLE FACTS (use in bullets)

  • 2 API routes: /api/resume/analyse and /api/email/send
  • 4 frontend pages in the user flow
  • 6 feature categories displayed on landing: Section Scores, ATS Check, Email Draft, Rewrites, Gap Analysis, Send Direct
  • AI model: LLaMA 3.3 70B (70 billion parameter model via Groq)
  • Section analysis covers: Technical Skills, Projects, Experience, Education, Contact/ATS
  • Each section scored 0-10 with: summary, strengths, suggestions (3 priority levels), before/after rewrites
  • Overall score: 0-100 aggregate

INTERESTING TECHNICAL DECISIONS (use for interview stories / resume bullets)

  1. memoryStorage over diskStorage — chose in-memory PDF handling because Render has an ephemeral filesystem; disk writes would be lost on dyno restart. Also eliminates cleanup logic.

  2. Brevo over Nodemailer — migrated mid-project when Gmail SMTP app password caused delivery issues in production. Brevo gives reliable transactional delivery without 2FA complications.

  3. Single AI call for both email + analysis — instead of two separate Groq API calls (one for email, one for analysis), combined both into one prompt. Saves latency and API credits.

  4. Strict JSON-only prompt — explicitly instructed LLaMA to return no markdown, no preamble, no explanation. Added post-processing (regex extraction + field validation) as a safety net.

  5. React Router state over Redux — for a linear 4-step flow (Home → Analysis → EmailPreview → Success), passing state via navigate is cleaner than global state management.

SUGGESTED RESUME BULLETS (ready to use or adapt)

Strong versions:

  • Architected a full-stack AI resume analyser using Node.js/Express and Groq's LLaMA 3.3 70B; engineered a dual-purpose prompt that generates section-wise analysis (scored 0–10) and a tailored HR cover email in a single API call
  • Implemented Multer memoryStorage for PDF ingestion — file processed entirely in-memory buffer and never written to disk, ensuring compatibility with Render's ephemeral filesystem
  • Built 3-stage error pipeline (422/502/500) with independent logging per stage, reducing debugging time for PDF extraction, AI, and email failures
  • Migrated email delivery from Nodemailer/Gmail SMTP to Brevo transactional API mid-project to resolve production delivery failures; integrated base64 PDF attachment in API payload
  • Designed React multi-page flow (Vite + Router v6) with collapsible analysis cards showing priority-tagged suggestions (High/Medium/Low) and side-by-side bullet rewrites
  • Implemented global theme system via React Context API and CSS custom properties — data-theme toggled on document root with localStorage persistence and 0.25s transition on all color tokens

HOW TO USE THIS FILE

When asking an AI to write resume bullets for ResumeForge, say:

"I have attached my ResumeForge project README and features file. Based on this, write [X] strong resume bullet points for a [Backend/Fullstack] Developer role. Focus on [specific area: AI integration / backend architecture / frontend / etc]. Keep bullets achievement-oriented and specific."

The AI will use the metrics, technical decisions, and feature list above to generate accurate, non-generic bullets.