What problem does it solve? Recruiters and hiring teams need a single entry point that decides where and how deeply to search for candidates across local contact indexes and hosted Powerset networks, without manually choosing backends, writing queries, or managing search runs. ## Core Features & Use Cases - Intelligent Routing: Records a decision.json that routes each query by surface (people, company, SQL, contacts), backend (local DuckDB or Powerset TurboPuffer/Postgres), depth (fast or deep), and mode (interactive or auto). - Fast Search Pipelines: Runs prepare-and-execute flows over local DuckDB indexes or hosted TurboPuffer retrieval with LLM filtering and reranking, including a deterministic direct-lookup fast path for names, emails, and LinkedIn URLs. - Deep JD Search: Turns a pasted job description or job-posting URL into a result-driven pond loop with human scoring in a web viewer, model-crafted follow-up queries, and a final shortlist.csv. - Use Case: Paste a senior backend engineer job posting and get a reviewed, multi-round candidate shortlist from your Powerset network, with every query and filter confirmed before any spend. ## Quick Start Ask the agent to search for candidates, for example: search for senior software engineers in San Francisco using my Powerset network.
