brightagent
AI-powered conversational interface for recruiting agencies to search and manage talent via natural language queries against their Avionté ATS.
14 commits over 2 days
What it is
Bright Agent is a custom chat interface for recruiting agencies that allows recruiters to ask natural language questions about their talent pool (e.g., "Find all React developers in New York") and receive context-aware AI responses. It integrates with Avionté ATS to replicate and synchronize talent data, uses vector embeddings for semantic search across resumes and profiles, and includes role-based access control, real-time webhooks, and analytics dashboards.
Why I built it
Built to replace traditional keyword-based ATS search with an AI-first conversational interface that allows recruiters to interact with their talent database more naturally. The tight Avionté integration and semantic search suggest this was custom-developed for a specific recruiting agency ("Bright Talent" per README) to modernize their talent discovery workflow and capture expert recruiter knowledge.
Tech stack
Next.js 14 (App Router), React 18, TypeScript, tRPC, Prisma ORM, PostgreSQL with pgvector (Supabase), NextAuth.js (Google OAuth), Tailwind CSS, deployed on Vercel. Vector embeddings power semantic search across talent data.
Timeline & effort
Last active Jul 2, 2025 — 14 commits across 2 active days (spanning 101 days).
Development spanned 101 days (late May through early September 2025) but with only 2 distinct active days and 14 commits total. The commit pattern—1 initial commit in May, then 13 commits in a September burst—suggests short intensive work sessions rather than sustained effort. Evidence points to a prototyping phase followed by a cleanup sprint.
Did it turn out useful?
The CLEANUP_PLAN.md reveals phases 1-3 (infrastructure stabilization, code architecture cleanup, performance/reliability) are marked complete, but phases 4-5 (feature completion, production hardening) are "ready to start." Multiple TODO items remain (Vercel env updates, production error testing). The aggressive refactoring documentation and incomplete production steps indicate a functional prototype undergoing stabilization rather than a fully shipped product. The private repo, zero stars, and "proprietary for Bright Talent" license suggest internal/client work still in refinement.