Epicenter is a monorepo platform for building local-first, privacy-focused productivity applications with optional cloud sync. It encompasses a suite of apps including Honeycrisp (notes with Yjs CRDT sync), Whispering (voice transcription/dictation with local and cloud providers), Matter (typed markdown vaults with SQLite projection), Local Mail (Gmail mirror with offline triage), Local Books (QuickBooks mirror with CLI tools), Vocab (AI language tutor), Skills (agent instruction management), and the Epicenter desktop host (Tauri + Bun sidecar). The shared platform provides a CRDT-based data layer (Yjs with Durable Object authority), portable blob storage (IndexedDB, OPFS, S3), an OAuth/infrastructure-agnostic auth system, an agent loop framework with tool catalogs, and a server framework built on Hono for Cloudflare Workers or Bun. The architecture emphasizes offline-first, typed error handling (Result/defineErrors pattern), compile-time platform seams (browser/Tauri/desktop-host), and rigorous ADR-driven design with evidence benchmarks.
This repository contains a minimal coding agent designed to work with local Ollama models. The agent operates in a loop, processing user requests by inspecting and modifying the workspace through a set of structured tools (file operations, shell commands, etc.). It implements six core components: live repo context collection, prompt shaping with cache reuse, structured tools with validation and permissions, context reduction and output management, transcript and memory persistence, and bounded delegation to subagents. The agent is implemented in a single Python file (`mini_coding_agent.py`) and is intended to be run via a CLI entry point. It requires an Ollama server running with a compatible model (default: qwen3.5:4b). The code is clean, well-structured, and demonstrates thoughtful agent design patterns, making it a useful reference for building similar agents.
Mistral Vibe is a production-grade terminal-based AI coding agent by Mistral AI. It provides an interactive CLI powered by Textual for multi-turn LLM conversations, file editing, shell execution, and subagent delegation. The system is architected around a core agent loop that orchestrates tool calls with a permission system, session management (resume, rewind, fork), and an event-driven checkpointing mechanism for file state snapshots. It integrates with multiple LLM providers (Mistral, Anthropic, OpenAI, Vertex) via a backend adapter pattern, and supports external integrations through MCP (Model Context Protocol) servers, custom hooks, and a skill system. The codebase heavily employs layered configuration, ports/adapters for audio, notifications, and update checking, and includes an app server for IDE integration via the ACP (Agent Communication Protocol). It also features telemetry via OpenTelemetry and GrowthBook-based experiments. The repository is actively maintained with recent commits, a comprehensive test suite (>1400 tests), and CI/CD pipelines for PyPI releases and PyInstaller binaries.
SIA (Self-Improving AI) is a framework that orchestrates a three-agent loop (meta-agent, target agent, feedback/improvement agent) to iteratively improve an AI system's performance on a benchmark task. In production, it runs as a CLI tool (`sia run`) that sets up a virtual environment, loads task files, and executes generations: the meta-agent reads `task.md` and generates `target_agent.py`, the target agent executes the task and logs results, and the feedback agent reviews the execution and rewrites the target agent for the next generation. A FastAPI web dashboard auto-launches to visualize per-generation accuracy, agent code, trajectories, and improvement plans. The architecture is a CLI-driven monolith with a plugin system for agent backends (Claude, OpenHands, PydanticAI), JSON-based provider/profile configuration, and Docker sandboxing for secure target-agent execution. It relies on external LLM APIs (Anthropic, OpenAI, Gemini, Nebius, OpenRouter, Together) and optionally on Modal/SandboxFusion for the RL-based weights mode. The repository also bundles four benchmark tasks (GPQA, lawbench, longcot-chess, spaceship-titanic) with public/private data splits and evaluators, giving it a hybrid character that includes data/research components. However, its primary function is as an application for running and visualizing self-improvement experiments.
OpenAgentd is a local-first, coding-focused AI agent platform that provides a desktop application (Tauri), mobile companion (iOS), browser-based UI (React/Vite), CLI, and an API server. The system orchestrates an agent loop with pluggable LLM providers (Anthropic, OpenAI, DeepSeek, Grok, Ollama, etc.), MCP tool integration, filesystem tools, code analysis, and a scheduler for recurring tasks. It features a dual backend architecture: a legacy Python/FastAPI implementation and a newer Rust v3 rewrite (axum-based) that is progressively replacing the Python version. The web frontend is a rich SPA with real-time streaming, workspace management, telemetry/observability dashboards, and extensive settings panels. The repository also includes code health analysis scripts, document chunking for RAG, and a comprehensive test suite across all layers.