Aider-AI/aider
Aider is an AI-powered pair programming tool that integrates with your terminal and codebase to assist developers by leveraging large language models for coding, testing, and project management.
Awesome AI Agents › Code Search Engines
adobe-research/dynasaur
DynaSaur is an advanced framework for large language model (LLM) based agents that transcends the limitations of predefined actions by using a programming language, specifically Python, as a universal medium to represent and execute actions. Unlike traditional LLM agents that rely on a fixed set of predefined actions, DynaSaur dynamically generates Python code snippets at each step to either invoke existing actions or create new ones when the current action set is insufficient. This capability allows the system to adapt and expand its functionality by developing new actions from scratch or by composing existing ones, thereby building a reusable and extensible library of actions for future tasks. The framework demonstrates remarkable versatility and robustness, as it can automatically recover from situations where no relevant actions are available or when existing actions fail due to unforeseen edge cases. This dynamic adaptability makes DynaSaur highly effective in handling a wide range of tasks and challenges. Empirically, DynaSaur has achieved top performance on the GAIA benchmark, a notable evaluation platform for language agents, and remains the leading non-ensemble method to date. The repository provides detailed instructions for installation and setup, including environment configuration, API key management, and data preparation using the GAIA benchmark dataset. Users are guided through creating a .env file for necessary API keys, downloading GAIA files from Hugging Face, and setting up a Python environment with required dependencies. The project also includes plans to add support for the OpenAI API in the future. DynaSaur represents a significant advancement in the development of intelligent agents by leveraging the flexibility of programming languages to extend the capabilities of LLMs beyond static action sets. This approach opens new possibilities for creating more adaptive, resilient, and capable AI agents that can evolve their behavior dynamically based on task requirements and environmental feedback.
https://github.com/adobe-research/dynasaur
Also in Code Search Engines
zilliztech/claude-context
Claude Context is an MCP plugin that provides semantic code search to AI coding assistants, enabling them to use the entire codebase as context for improved code understanding and generation.
BloopAI/bloop
Bloop is a fast, AI-powered code search engine that enables natural language queries, code understanding, and generation to boost developer productivity using existing codebases as context.
AbanteAI/archive-old-cli-mentat
This repository contains an archived and unsupported version of an old command-line tool named Mentat, now replaced by an AI-powered GitHub bot available at mentat.ai.
dot-agent/nextpy
Nextpy is a cutting-edge framework for building self-modifying AI software with advanced prompt engineering, modularity, and significant performance optimizations.
melih-unsal/DemoGPT
DemoGPT is an open-source all-in-one library that automates the creation of interactive LLM agent applications by generating LangChain code and transforming it into Streamlit apps, leveraging GPT-3.5-turbo and supporting future autonomous API integration.
chunkhound/chunkhound
Local-first codebase intelligence CLI that indexes a repository for regex and semantic search and answers architecture questions with citations from code, git history and technical web research.
instructa/ai-prompts
Instructa AI Prompts is an open-source repository providing curated AI prompts and rules to enhance AI-assisted coding workflows across multiple popular AI coding tools.