Chuckie

What It Is

Chuckie is a personal AI system designed around the idea that useful agents need memory, structure, and connective tissue, not just a chat interface. Each agent operates within a shared environment: a persistent task queue, a common tool layer, and a Postgres database with pgvector for semantic retrieval.

Agents

Planner breaks goals into structured task sequences and coordinates work across other agents. It maintains state across sessions so long-running projects don't require re-briefing.

Research monitors sources, synthesizes information, and produces structured briefs. It ingests from RSS feeds, arXiv, SSRN, and selected data APIs, then surfaces what's relevant based on an evolving topic registry.

Money tracks financial data, synthesizes market context, and supports investment-related decisions. It handles portfolio monitoring, signal tracking, and periodic summaries of developments across markets and sectors the user cares about.

Execution Layer

The tool layer provides agents with structured access to external services: web search, document parsing, structured data extraction, and output formatting. Tools are typed and consistent across agents, so outputs from one can feed directly into another without reformatting.

All task state persists in Postgres. Long-running jobs survive session boundaries. Outputs are stored and retrievable by other agents downstream.

In Progress

Current work connects Chuckie to live external inputs: email, calendar events, and real-time data feeds. The goal is a system that operates continuously with selective human input, rather than one that requires invocation for each task.