self-correcting

The self-correcting memory layer for AI agents. Zero-dependency Python memory and MCP server. Supersede, revert, or forget a value deterministically. Verifiable erasure, witness-backed tamper-evident receipts, EU AI Act ready.

Ultimate vibe coding plugins for Hermes AI agent — Effect-ts, LSP, Semble code search & Graphify knowledge graph + t/s status bar. 23 tools, 5 plugins, stdlib core.

Most AI writing is fluent slop. Writing Agent turns a topic into a publishable, well-sourced article or book that argues a real thesis and self-critiques until it isn't slop - autonomous, model-agnostic, local-first.

Adaptive vector search with self-correcting embeddings. Fixes semantic collapse in RAG systems via spectral chelation, dynamic dimension masking, and neural adaptation.

A deterministic pipeline that diagnoses, explains, and safely auto-fixes failures in LLM agent systems.

CodegniPy is a groundbreaking Python library that elevates AI to a first-class citizen of the language. It introduces a cognitive computing engine where deterministic code and non‑deterministic LLM reasoning coexist seamlessly, enabling you to write Python with intent rather than just instructions.

GraphNews is a multi-agent system that acts as a self-correcting editorial team, generating newsletters

Agentic forecasting application with autonomous model selection, multi-indicator prediction, and self-correcting pipelines

Self-correcting data-analyst agent: ask a dataset a question in plain English and get runnable Python, SQL, or an Excel formula. Built on LangGraph + a local Qwen2.5-Coder LLM, with a generate→execute→reflect loop that fixes its own errors (measured: 75%→94% success).

An agentic, self-correcting machine learning experimental loop built locally on Apple Silicon using MLX.

Self-correcting Python coding agent: plan → generate → execute tests → fix → repeat. LangGraph loop with sandboxed subprocess execution. Supports Claude and GPT-4o.

An autonomous AI agent that writes Python code, catches its own errors, and self-corrects.

Multimodal agentic RAG on NVIDIA NIM: a LangGraph agent that routes retrieved figures through a Nemotron vision-language model and validates grounding with an LLM-as-judge benchmark.

A vision-language model closes a self-correcting compositing loop over a real-time 3D parallax scene: it stages assets, critiques its own snapshot against a strict validated contract, fixes itself, and renders an animated camera dolly.

Comparative study of two Agentic AI architectures for automated data science: hidden-tool agents vs transparent code-generating agents. Built with CrewAI, OpenAI GPT-4o, tested on Titanic & House Prices datasets.

A governed, self-correcting agentic system: eval-first loan review with human-gated authority, bounded self-tuning, and a sandboxed ops agent that files PRs

Self-correcting 5-pillar financial AI with continuous improvement cycle. Production: boseongkang.github.io/newstrend