Report Abuse

Basic Information

This repository is a collection of experimental AI agent implementations and personal explorations in agent development. It provides toy example agents that demonstrate agentic programming concepts such as conversational behavior and retrieval-augmented research workflows. The project uses a lightweight environment and dependency workflow managed with uv and expects local environment variables for model endpoints, API keys, embedding model names and a local database directory. Included examples are a conversational Therapist agent that maintains memories across sessions and a Startup Researcher that queries a search API and generates markdown outputs. The README includes quick start instructions, example run commands, and guidance for configuring credentials and model names in a local .env file.

Links

Categorization

App Details

Features
Contains multiple small agent examples implemented as learning prototypes. The Therapist agent is a conversational agent that saves short-term interactions to a SQL-backed local database and builds long-term summarized memory with a vector store. The Startup Researcher agent integrates a web search API to collect information and produce a markdown research report. The codebase uses DSPy for agent structure, an exa search integration for external retrieval, chroma-like vector lookups for memory, and uv for environment and dependency management. Examples are runnable from the command line with simple uv run commands and configuration is driven by a provided .env.example.
Use Cases
The repo serves as a practical starting point for developers and researchers who want runnable, minimal examples of agent patterns such as conversational memory and retrieval-augmented generation. It shows how to wire environment-managed dependencies, configure model and API credentials, persist short-term conversation history, and implement a basic long-term memory lookup. The Startup Researcher illustrates how to call an external search API and convert results into structured markdown output. The examples can be executed locally to inspect behavior, iterate on prompts and memory strategies, and serve as templates for building more sophisticated agent prototypes or experiments.

Please fill the required fields*