llm-agents-python

LangGraph multi-agent orchestrator template — dynamic Supervisor routing, 5 example specialist agents, and optional human-in-the-loop review. POC tier of the Multi-Agent Orchestrator Template family: in-memory state, no auth, minutes to first run.

Byte-level proof that a refactor changed nothing: golden-census records a program's stdout, stderr, exit codes and written files with the machine normalized away, and compares them after a change. Built for LLM coding agents; Python 3.11+, no dependencies.

Homebrew tap for golden-census: brew install bunkanoyume/golden-census/golden-census

An air-gapped, Text-to-Pandas enterprise data intelligence platform built with Streamlit and local LLMs. Mitigates code injections via isolated namespace execution sandboxes and handles chaotic column configurations natively in real-time.

Reproduce how LLM agents fail in production (duplicate charges, infinite loops, prompt injection) with no API key, then fix each with a tested guard you can use in your own agent loop.

LLM-driven computer-use agent that learns legacy bank UI workflows once, then replays them deterministically — with safety guardrails, structured evidence, and human escalation.

CrewFlow is a production-ready multi-agent AI workflow system built using CrewAI and Python. This project demonstrates how multiple AI agents collaborate to solve complex tasks such as research, analysis, and reporting. It showcases agent orchestration, task delegation, and modular AI system design used in real-world enterprise applications.

Algorithmic crypto trading system for Binance evolving into an AI-driven agent. Real-time signals, risk-managed execution, and a roadmap to sentiment-aware, self-tuning strategies.