Frequently Asked Questions (FAQ)

Product Questions

Q: Can I integrate HireMinds with my existing ATS (like Workday or Greenhouse)? A: Currently, HireMinds serves as a standalone ATS. Integrations with external Job Boards and ATS platforms are on our roadmap.

Q: How does the AI score candidate interviews? A: Scoring is based on the transcript generated during the session. The ScoringService evaluates behavioral responses using the STAR method (star_compliance), confidence (confidence_signal), and technical accuracy. A candidate must have at least 4 conversational turns to receive a score.

Q: What happens if a candidate's internet disconnects mid-interview? A: The system retains the partial transcript. The session_watchdog runs every 15 minutes. If an interview is stuck in IN_PROGRESS for over 30 minutes with fewer than 4 turns, it flags the session for manual review and automatically refunds the recruiter's credits.

Technical Questions

Q: Can I run HireMinds fully locally? A: Yes, if using local STT/TTS and local LLMs, though the standard configuration points to cloud APIs for production stability.

Q: Why does the older voice agent fallback take a few seconds to respond? A: The legacy WebSocket fallback uses server-side Voice Activity Detection (VAD) which waits for 2000ms of silence before processing the audio. The primary WebRTC pipeline uses client-side VAD, which reduces this delay to ~200ms.

Q: Does HireMinds store candidate video? A: Currently, no. HireMinds stores the audio transcript, proctoring events, and optionally audio recordings (if enabled), but does not record continuous video.

Q: What is the vector type error in PostgreSQL? A: HireMinds uses semantic ATS matching, which requires the pgvector extension. Run CREATE EXTENSION vector; in your PostgreSQL database to resolve this issue.