hustnn / devin_family

collection of high quality open source agents and code generation framework powered by LLM

devin_family

I am using llm to generate the application according to my requirements, this process highly depends on the code generation ability and agent ability to finish some complex tasks. I did a lot of research on it and make a collection for the reference. Hope it can also help you.

open source implementaiton

• MetaGPT https://github.com/geekan/MetaGPT/tree/main/examples/di
• Devika https://github.com/stitionai/devika
• OpenDevin https://github.com/OpenDevin/OpenDevin
• Maestro https://twitter.com/skirano/status/1770221447606104154
• gpt-engineer https://github.com/gpt-engineer-org/gpt-engineer
• SWE-agent https://github.com/princeton-nlp/SWE-agent
• auto-coder https://github.com/allwefantasy/auto-coder
• machine learning agent https://github.com/WecoAI/aideml
• CodeHalu https://github.com/yuchen814/CodeHalu
• trove https://github.com/zorazrw/trove
• ChatDev https://github.com/OpenBMB/ChatDev
• Devon https://github.com/entropy-research/Devon
• r2e https://github.com/r2e-project/r2e

agent

• crewAI https://github.com/joaomdmoura/crewAI
• autogen https://www.microsoft.com/en-us/research/project/autogen/
• AgentLite https://github.com/SalesforceAIResearch/AgentLite
• AgentVerse https://github.com/OpenBMB/AgentVerse
• OneTwo https://github.com/google-deepmind/onetwo
• agentkit https://github.com/holmeswww/AgentKit
• Xagent https://github.com/OpenBMB/XAgent
• llm-reasoners https://github.com/maitrix-org/llm-reasoners/
• OSWorld https://github.com/xlang-ai/OSWorld
• agent design pattern https://twitter.com/shao__meng/status/1784474864113570027
• task-decompose https://github.com/zgimszhd61/task-decompose-quickstart
• JARVIS https://github.com/microsoft/JARVIS
• agent search tree https://github.com/andyz245/LanguageAgentTreeSearch
• stream of search https://github.com/kanishkg/stream-of-search
• ambigdocs https://ambigdocs.github.io/
• USACO https://github.com/princeton-nlp/USACO
• Logic-LLM https://github.com/teacherpeterpan/Logic-LLM
• chain of abstraction https://twitter.com/jerryjliu0/status/1778855608202277156
• graph cot https://github.com/PeterGriffinJin/Graph-CoT
• Multiplex https://github.com/LCS2-IIITD/Language_Model_Multiplex
• RoT https://github.com/huiwy/reflection-on-trees
• LanguageModelsasCompilers https://github.com/kyle8581/LanguageModelsasCompilers
• lumos https://github.com/allenai/lumos
• ReaLMistake https://github.com/psunlpgroup/ReaLMistake
• eclair https://github.com/HazyResearch/eclair-agents
• agentscope https://github.com/modelscope/agentscope

frameworks/tools/application

• Dspy https://github.com/stanfordnlp/dspy
• Tool use https://twitter.com/omarsar0/status/1770497515898433896
• Strong types https://github.com/bananaml/fructose
• Phidata https://github.com/phidatahq/phidata/tree/main/cookbook/claude
• RepoToTextForLLMs https://github.com/Doriandarko/RepoToTextForLLMs
• storm https://github.com/stanford-oval/storm
• storm claude https://github.com/angelina-yang/Claude_API_Contest/tree/main
• MathVC https://github.com/MurongYue/MathVC
• AIOS https://github.com/agiresearch/AIOS
• cohere-terrarium https://github.com/cohere-ai/cohere-terrarium
• data labeling https://github.com/HumanSignal/Adala
• data to paper https://github.com/Technion-Kishony-lab/data-to-paper

Prompt

• gpt-prompt-engineer https://github.com/mshumer/gpt-prompt-engineer
• Claude-ception https://twitter.com/LangChainAI/status/1770124528322318522
• Enhancing Few-shot Text-to-SQL Capabilities https://twitter.com/emelia_seles/status/1767797228398342244
• prompt optimization https://twitter.com/hi_pondai/status/1769445700977221769
• advprompter https://github.com/facebookresearch/advprompter
• PromptReps https://github.com/ielab/PromptReps
• PromptAgent https://github.com/XinyuanWangCS/PromptAgent
• promptbench https://github.com/microsoft/promptbench
• coding prompt https://twitter.com/seclink/status/1776408191557075052
• Prompt Engineer https://twitter.com/JagersbergKnut/status/1776230873525788903

planning and reasoning

• ADaPT https://allenai.github.io/adaptllm/
• Step-by-Step Comparisons https://twitter.com/omarsar0/status/1770492690129359135
• Flow Adhering Planning https://twitter.com/ShamiikRoy/status/1768007927342969281
• CoT improves performance for code generation https://twitter.com/ellev3n11/status/1769574342206038201
• Implicit CoT https://twitter.com/Euclaise_/status/1769424716568072641
• RAT https://twitter.com/AndyLin2001/status/1767075865127719101
• Self discover https://twitter.com/hi_pondai/status/1769445700977221769
• Self-Correct Their Reasoning https://twitter.com/jefffhj/status/1709634179594703187
• Dspy optimization https://twitter.com/CShorten30/status/1768780338715623907
• fillerTokens https://github.com/JacobPfau/fillerTokens

LLM coding ability

• Claude's responses superior https://twitter.com/IntuitMachine/status/1770513600689410534
• Claude 3 Opus https://twitter.com/alexalbert__/status/1767258557039378511

paper/article

• Microfot AutoDev https://twitter.com/Marktechpost/status/1770294548629189071
• Coding preference https://twitter.com/MWeyssow/status/1768623185107550295
• Livecodebench https://twitter.com/StringChaos/status/1768264053879873668
• AgentCoder https://twitter.com/debarghya_das/status/1768468420058902536
• 10 agent paper https://twitter.com/GptMaestro/status/1780605438092013797
• OpenChatML https://twitter.com/erhartford/status/1780027791641391489