Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!
At its simplest, an AI Agent is a step up from a standard chatbot. While a chatbot is designed to talk, an agent is designed to act. Think of it like the difference between a travel guide who tells you about Paris (Chatbot) and a personal assistant who actually books your flights, reserves the hotel, and handles the dinner reservations (Agent).
The Core Components¶
An AI agent doesn’t just respond to a prompt. It follows a loop of reasoning to accomplish a goal. Most agents operate using these four pillars.
Perception: Taking in data from the environment (user prompts, web searching, or file reading).
Brain (Reasoning): Using a Large Language Model (LLM) to break a complex goal into smaller, logical steps.
Memory: Remembering past interactions or storing information found during a task to use later.
Tools: The ability to “handshake” with other software (like your email, calendar, or a web browser) to execute actions.
How it Works: The “Loop”¶
Unlike a single-turn conversation, an agent often works in an autonomous loop until the job is done.
Goal Setting: You give the agent a high-level objective (e.g., “Research the best laptop under $1,000 and email a summary to my boss”).
Planning: The agent decides it needs to search the web, compare specs, and then access your email tool.
Action: It executes the search and reads the results.
Self-Correction: If the first search doesn’t work, it tries a different query.
Completion: It sends the email and notifies you it’s finished.
Agents¶
localgpt¶
localgpt is a local AI assistant.
hindsight¶
hindsight is an agent memory system built to create smarter agents that learn over time. Most agent memory systems focus on recalling conversation history. Hindsight is focused on making agents that learn, not just remember.