Skip to content

AI_USAGE.md

Latest commit

 

History

History
73 lines (57 loc) · 4.85 KB

AI_USAGE.md

File metadata and controls

73 lines (57 loc) · 4.85 KB
Raw

AI-Assisted Development

This document provides a transparent overview of how artificial intelligence tools are used in the development, architecture, testing, and maintenance of the Resume Builder project.

1. Governance & Responsibility Statement

While AI tools assist in code generation, refactoring, test corpus design, and documentation, human developers remain strictly responsible for:

  • Reviewing and validating all code modifications.
  • Verifying security boundaries, rate limiting policies, and secret isolation.
  • Executing automated test suites and confirming zero regressions.
  • Maintaining privacy standards and ensuring user data confidentiality.

2. AI Tools Used in Development

The following AI tools and models are confirmed for development and application features:

  • Google Antigravity AI Coding Assistant: Used for pair programming, systematic debugging, test harness generation, and architectural analysis.
  • Google Gemini API (gemini-3.5-flash): The production AI model integrated into the backend Job Match evaluation service.

3. How AI Is Used

  1. Architecture & Threat Modeling:
    • Designing multi-layered server rate limiting pipelines, cryptographic SHA-256 caching, and circuit breakers.
  2. Systematic Debugging & Root-Cause Analysis:
    • Diagnosing upstream API model transitions (gemini-2.5-flash deprecation to gemini-3.5-flash).
    • Resolving TypeScript compiler flags and downlevel iteration options.
  3. Comprehensive Test Suite Generation:
    • Creating the 35-scenario Universal Resume Importer test corpus with positive, negative, and edge cases.
    • Building 41-case security audits for IP rate limiting and payload validation.
    • Developing 47-point SEO and structured data verification scripts.
  4. Technical Documentation:
    • Drafting architecture diagrams, data flow schemas, Architecture Decision Records (ADRs), and developer workflows.

4. Development & Milestone Log

2026-08-14

  • Task: Core Resume Editor, Live A4 Preview & Universal Custom Section System
  • AI Tool: Antigravity Assistant
  • Problem: Need an interactive A4 resume builder with drag-and-drop reordering, custom section schemas, and native PDF/Word export without server dependencies.
  • AI Assistance: Implemented Zustand store with localStorage persistence, @dnd-kit sortable lists, @react-pdf/renderer vector PDF builder, and docx native OpenXML generation.
  • Developer Action: Reviewed layout fidelity, verified print margins, and validated cross-browser responsiveness.
  • Result: Complete client-side resume drafting workspace with 1:1 live preview.

2026-08-15

  • Task: 10-Stage Universal Resume Importer
  • AI Tool: Antigravity Assistant
  • Problem: Need a client-side parser to ingest existing PDF and Word resumes without external cloud OCR or LLM dependencies.
  • AI Assistance: Created 10-stage deterministic parsing pipeline with pdfjs-dist, mammoth, spatial column reconstruction, 50+ alias semantic registry, and 4-factor confidence scoring. Built a 35-scenario automated test corpus.
  • Developer Action: Verified layout reconstruction across multi-column, sidebar, and academic CV formats.
  • Result: 100% client-side resume importer passing 43/43 comprehensive tests.

2026-08-16

  • Task: Gemini Job Match AI Integration & Security Hardening
  • AI Tool: Antigravity Assistant & Google Gemini API
  • Problem: Upstream model gemini-2.5-flash returned 404 on newer API keys, and the public endpoint required protection against quota exhaustion and abusive traffic.
  • AI Assistance: Transitioned model to gemini-3.5-flash, bounded output tokens to 4096, implemented sliding window IP rate limiting (3/10m, 10/24h), IP concurrency locks, token-bomb sanitizers, SHA-256 caching (5m TTL), and quota circuit breakers. Added real-time cooldown UI.
  • Developer Action: Validated live HTTP 200 responses, inspected error handling under 429 quota exhaustion, and verified secret key isolation.
  • Result: Production-ready Job Match feature passing 41/41 security tests and 22/22 integration tests.

2026-08-16

  • Task: Production SEO & Structured Data Architecture
  • AI Tool: Antigravity Assistant
  • Problem: Need public search engine discoverability for the landing page while enforcing strict noindex protection on the private editor workspace.
  • AI Assistance: Built centralized SEO configuration (src/seo/siteConfig.ts), dynamic usePageSeo hook, Schema.org JSON-LD generators (WebApplication, WebSite, FAQPage), robots.txt, sitemap.xml, and an automated 47-test SEO validation suite.
  • Developer Action: Verified Open Graph preview tags, live canonical domain mapping (https://naphix-resume.netlify.app/), and validated zero user PII leakage.
  • Result: Complete SEO foundation passing 47/47 tests and compiling cleanly with 0 build errors.