Skip to content

System Architecture

Overview

HireMinds follows a modern decoupled architecture:

  • Frontend: Next.js 14 App Router (React, Tailwind CSS).
  • Backend: FastAPI (Python 3.11+).
  • Database: PostgreSQL with pgvector for semantic embeddings.
  • Caching & Rate Limiting: Redis.
  • Voice AI: WebRTC Speech-to-Speech Pipeline / WebSocket fallback.

High-Level Architecture Diagram

flowchart TD
    Browser["Candidate/Recruiter Browser (Next.js)"]
    API["FastAPI Backend"]
    AI["HireMinds Voice AI (WebRTC S2S)"]
    DB[("PostgreSQL + pgvector")]
    Redis[("Redis Cache")]

    Browser <-->|"HTTP/REST & WebSockets"| API
    Browser <-->|"WebRTC Audio"| AI
    API <-->|"Prompt Context & Webhooks"| AI
    API <-->|"Read/Write"| DB
    API <-->|"Pub/Sub"| Redis

AI Architecture

HireMinds leverages multiple AI/LLM models:

  • ATS Semantic Matching: pgvector embeddings for resumes and job descriptions.
  • Resume Parsing: General LLM data extraction (resume_parser.py).
  • Voice Agent: Fused Speech-to-Speech WebRTC pipeline to eliminate traditional STT->LLM->TTS latency.
  • Context Injection: Candidate GitHub analysis and parsed skills are injected into the dynamic system prompt (build_system_prompt()).

Note

All AI API keys and secrets must be configured via environment variables. User data is not used for training.