All projects

Agentic chat

Agentic chat is a multi-agent AI system with intelligent query routing, pgvector-based retrieval and semantic caching, long-term memory, document intelligence, and tool-integrated research workflows.

Agentic chat was built as a multimodal AI workspace that can decide when to use long-term memory, retrieval, vision, or external tools instead of forcing every query through the same response path. It supports deep research, web search, YouTube analysis, URL context injection, and Google Workspace actions through LangGraph-based orchestration.

Highlights

  • Orchestrates multi-step workflows with LangGraph across web search, YouTube analysis, Google Workspace actions, and deep-research planning rather than relying on a single tool call.
  • Built an intelligent context router across memory, documents, vision, hybrid, and tool-only modes, with semantic caching that reduced API costs by 40%.

Tech stack

Next.jsTypeScriptOpenAIMem0PostgreSQLLangGraphRAGAgentic ResearchAgents

More projects

Additional work across AI products, developer tooling, and full-stack systems.

Browse all