16 repositorios
Using LLMs to generate precise diffs for modifying existing source code rather than generating new snippets.
Distinct from LLM-Driven: Focuses on the application of precise edits to existing files via diffs, not just snippet generation.
Explore 16 awesome GitHub repositories matching development tools & productivity · Diff-Based Edit Application. Refine with filters or upvote what's useful.
Aider is a terminal-based AI coding assistant and pair programmer that uses large language models to write, edit, and refactor source code across multiple files and programming languages. It functions as a command line interface for automating programming tasks and managing codebase modifications. The tool distinguishes itself by creating structural maps of entire codebases to provide language models with the necessary context for navigating and modifying large repositories. It further expands input capabilities through a speech-to-text pipeline for voice-driven development and multi-modal in
Translates natural language instructions into precise source code modifications using model-generated diffs.
This project is an AI code review tool and asynchronous task orchestrator designed to analyze uncommitted code changes and architectural decisions. It functions as an LLM agent integration plugin and cross-model workflow bridge, connecting different large language model agents to delegate engineering tasks and synchronize session context. The system enables multi-model orchestration to cross-reference design decisions and pressure-test architectural assumptions. It provides mechanisms to export session threads and transfer engineering context between separate AI coding environments, allowing
Identifies technical issues by isolating specific modified code blocks within diffs for targeted AI analysis.
PR-Agent is an AI-powered code review tool and developer assistant designed to automate pull request workflows. It functions as an automated reviewer and git workflow automation tool that uses language models to analyze code diffs and provide technical feedback. The project distinguishes itself through the ability to generate automated pull request descriptions and project changelogs based on code changes. It also enables contextual querying of a codebase, allowing users to ask questions about specific lines of code or change sets within a pull request. The system includes capabilities for A
Processes raw pull request differences to isolate modified code blocks for targeted AI evaluation.
This project is an AI-powered IDE extension and LLM coding assistant that provides a conversational interface for generating, refactoring, and debugging code. It functions as an AI agent framework and a Model Context Protocol client, connecting AI models to external data sources and tools to automate complex development tasks. The system is distinguished by its use of autonomous AI agents capable of multi-step task execution, including the ability to read files, modify code, and run terminal commands iteratively. It supports recursive agent orchestration through subagent delegation and employ
Streams AI-generated modifications into the editor as interactive diffs for review and acceptance.
Reviewdog is an automated review bot and CI code review orchestrator that converts the output of static analysis tools into automated pull request comments. It functions as a linter output parser and static analysis commenter, transforming unstructured logs from compilers or linters into structured diagnostics. The project distinguishes itself by using pattern-based output parsing and a platform-agnostic plugin architecture to unify multi-language linting workflows. It employs diff-based result filtering to isolate issues introduced in a specific commit and provides the ability to post action
Compares analysis findings against the version control commit diff to isolate issues introduced in the current change.
Reviewdog is a linter result posting tool and a diff-aware static analysis filter. It parses the output of various linters and posts findings as comments on pull requests within code hosting platforms. It also functions as a quality gate for CI pipelines, failing builds when findings exceed specified severity thresholds. The project distinguishes itself by isolating static analysis issues to only those introduced within the current git code diff, preventing the reporting of legacy errors. It unifies tool findings by processing industry-standard SARIF and XML diagnostic formats alongside custo
Isolates static analysis issues to only those introduced in the current git diff to avoid reporting legacy errors.
aicommits is a command line tool and AI code summarizer that generates descriptive git commit messages by analyzing staged code changes. It functions as an LLM git commit generator, transforming technical diffs into human-readable summaries based on standardized formats. The project features a multi-provider AI interface that connects to either cloud-based or local artificial intelligence models. Users can customize generation logic through specific language locales, length constraints, and custom prompts to ensure consistent version control documentation. The tool integrates directly into v
Extracts pending git changes via shell commands to provide the primary context for AI analysis.
CodeMirror is a browser-based code editor framework and modular extension system used to embed full-featured text editors into web pages. It functions as a syntax tree parsing engine and a language server protocol client, enabling structural language analysis and deep integration with external programming language services. The project is distinguished by its modular architecture, which uses a system of extensions and compartments for dynamic reconfiguration at runtime. It supports real-time collaborative editing and state synchronization through an operational transformation framework, allow
Highlights differences between editor content and an original document by showing deleted and new text.
Opencommit is a command-line tool and automation suite that uses large language models to analyze staged changes and generate descriptive git commit messages. It functions as an AI-driven commit generator that can be integrated directly into the version control lifecycle. The project distinguishes itself through support for both cloud-based AI providers and locally hosted models to ensure data privacy. It provides specialized automation via git hooks for real-time suggestions and GitHub Actions for refining commit messages during continuous integration workflows. The tool includes capabiliti
Extracts raw git diffs of staged modifications to provide the necessary context for AI commit message generation.
CodeCompanion is a Neovim plugin that brings large language model capabilities directly into the editor, enabling turn-based conversations with AI models in a dedicated chat buffer. It provides a comprehensive interface for interacting with LLMs, supporting multiple providers through a flexible adapter system that can route requests to various hosted or local language model services. The plugin distinguishes itself through its extensive context-sharing capabilities, allowing users to send buffer contents, visual selections, git diffs, LSP diagnostics, terminal output, quickfix lists, and view
Tags a file in the chat so the LLM can directly modify code without manual copy-paste.
Claude Squad is a terminal-based orchestrator for running multiple AI coding assistants in parallel. It manages the lifecycle of AI agent sessions from a single keyboard-driven interface, allowing users to launch, monitor, pause, resume, and terminate agents without leaving the command line. The tool isolates each agent's work in separate git worktrees, so changes remain on independent branches and never interfere with each other. Before any modifications are committed or pushed, users can review a diff preview of what each agent produced and approve or reject the changes. This diff-based app
Shows a diff preview of AI-proposed changes and requires user approval before committing modifications.
1code is an AI-assisted development environment that provides a unified interface for switching between multiple AI coding agents. It toggles between a read-only analysis mode and a full execution mode, asking clarifying questions, building structured plans with previews, and requiring user approval before making code changes. The environment integrates with external services and tools through the Model Context Protocol (MCP), enabling connections to databases, project management systems, and code repositories. Agent sessions can run either locally or in persistent cloud sandboxes that stay al
Shows real-time visual diffs of every file edit, command, and search result before changes are applied.
Este proyecto es una herramienta de análisis estático impulsada por IA y escáner de vulnerabilidades automatizado diseñado para detectar fallos de seguridad como inyecciones y elusiones de autenticación. Utiliza modelos de lenguaje grandes para realizar razonamiento semántico a través de múltiples lenguajes de programación, identificando vulnerabilidades dentro de los cambios de código. La herramienta opera como una GitHub Action que se integra en pipelines de integración continua para analizar diffs de pull requests. Se centra en las líneas de código modificadas para apuntar a nuevos riesgos y reporta los hallazgos publicando comentarios automatizados directamente en el pull request. El análisis está dirigido por políticas de seguridad personalizables e inyección de reglas externas, permitiendo instrucciones específicas para el proyecto. Estas reglas y filtros personalizados se utilizan para reducir el ruido y descartar hallazgos de bajo impacto para priorizar los riesgos de seguridad de alta confianza.
Analyzes only modified code blocks within a diff to target new vulnerabilities for AI evaluation.
Star-vector es un conjunto de sistemas de visión-lenguaje diseñados para generar gráficos vectoriales escalables a partir de texto o imágenes. Utiliza un modelo base de visión-lenguaje para tratar la creación de elementos visuales como una tarea de generación de código estructurado. El sistema emplea una arquitectura multimodal que mapea patrones y formas visuales a elementos estructurales correspondientes en una secuencia de código vectorial. Incorpora un mecanismo de retroalimentación de renderizado y aprendizaje por refuerzo para refinar iterativamente la fidelidad de los gráficos generados, comparando las salidas renderizadas con las imágenes objetivo. El proyecto cubre un amplio rango de capacidades de generación y optimización, incluyendo vectorización de imagen a SVG, síntesis de texto a SVG y la producción de diagramas vectoriales semánticamente ricos. Se enfoca específicamente en el reconocimiento de formas estructurales y la optimización de código vectorial para mejorar la precisión visual.
Refines vector graphics by comparing rendered output against a target image to correct visual errors.
Micro-agent es un framework para agentes impulsados por IA centrado en el desarrollo guiado por pruebas automatizado, conversión de diseño a código y orquestación de herramientas externas. Utiliza agentes que escriben, prueban y refinan iterativamente el código fuente basándose en prompts de lenguaje natural y archivos de diseño. El sistema transforma tokens de diseño visual y componentes en código con tipos seguros y linted comparando URLs en vivo contra capturas de pantalla de referencia para garantizar la paridad visual. También proporciona un protocolo para vincular agentes a servicios externos de comercio, búsqueda y gestión de activos para sincronizar datos y expandir las capacidades funcionales. El proyecto cubre capacidades para la generación iterativa de código y pruebas automatizadas, donde los archivos fuente se modifican repetidamente hasta que pasan scripts de prueba específicos. Incluye soporte para conectores del Protocolo de Contexto de Modelo (Model Context Protocol) y una interfaz de línea de comandos para gestionar ajustes operativos y selección de modelos.
Refines generated code by comparing live URL renders against reference screenshots for visual parity.
Claudian is a framework that combines AI coding agents, knowledge base integration, and a multi-provider orchestrator for managed interactions with large language models. It functions as a browser extension that connects users to AI services through a sidebar and inline editing interface, providing a system for integrating agents into local directories to perform file operations, bash commands, and workspace searches. The project distinguishes itself with a multi-provider orchestrator that allows switching between different AI backends while maintaining separate conversation states and config
Modifies selected content at the cursor position using a word-level diff preview to refine notes.