Skip to main content

Production-gradeLLM tools

programmatically

Due to the powerful understanding and generation capabilities of this wave of LLMs, command-style programming that pays attention to details seems to be no longer needed, and Statement programming that focuses on process or pipeline capabilities of business logic orchestration has become the mainstream "programming" method. Therefore, we have seen that rookies including LCEL, DSPy, and Semantic-kernel have begun to pay attention "upwards" and focus their main energy on business logic orchestration capabilities. Perhaps this indicates a DAG approach similar to Airflow. Will it become a new generation of programming normal form?

totality

Prompt tool

LangSmith or Azure Prompt Flow

Programming language

Python or Javascript

Development framework/orchestration tool

Langchain or Semitic Kernel

API framework/service box frame work

FastAPI or Next.js

Cloud hosting vendor

Render (AWS Lamada) or Vercel

BuildLLM applications

classificationNameStar QuantityremarkLearning
frameLangChain77[Required]*
LlamaIndex29【必学】专注于RAG,Python,JS SDK比较差*
Semantic-kernel17【不确定?】C#,Python,Java?
haystack13有了LangChain可能就不需要这个了
7Too academic?太学术了?The framework for programming-not prompting-foundation models
AgentSuperAgent4This feels very professional, and I will Learning again when I have time?
Rift3专门针对程序员
AutoGen22
Camel4
GPTsOpenGPTs6Produced by LangChain
low-codeLangFlowLangChain visualization
LiteLLM
Flowise21
Langflow14
PromptChainer商业感觉做的很专业
RAGvectaraclosed sourceVery professional RAG commercial soft ware; there are also Open-Source projects; users can upload PDF, Word, PPT, RTF and other file Data to the Vectara platform to build a Data Searching engine.

Archiving and Observation

classificationNameStar Quantityremark
observeObsidian CopilotAn interesting approach to how to use Semantic Searching and the BM25 implementation of OpenSearch
Tanuki一个使用装饰器进行数据验证的 LLM 框架
griptape1.5一个具有稍微更好的内部编码标准的 langchain 替代品
16Controlling modern Language Model more Valid Earth than traditional prompts or links比传统的提示或链接更有效地控制现代语言模型
intructor3.6
0.5Another alternative to langchain, but has an interesting approach to defining workflowslangchain 的另一个替代品,但在定义工作流程方面有一个有趣的方法
6.5There are derivative products for chat, medical/scientific paper workflow, developer semantic Searching, and title and story text semantic Searching有用于聊天、医学/科学论文工作流程、开发者语义搜索以及标题和故事文本语义搜索的衍生产品
5.7A simple, lightweightLLM package一个简单、轻量级的 LLM 封装器
pigeonholeChidori1.2
AutoChains已经3个月没更新
MiniChain1.2已经3个月没更新
SimpleAIChat3.3已经7个月没更新

monitor and analyze

LLM services at the production level require:

  1. Debugging Prompt

  2. Prompt version management

  3. Metric related to the dependence of the test/verification system

  4. Data Set Management

  5. Various Metric monitoring and statistics: visit volume, response time, Token fee, etc.

classificationNameStar Quantityremark
SaaSLangSmithN/A
Open-SourceLangFuse2
Open-SourcePrompt Flow7
Helicone1.2

Production-levelLLM App maintenance platform

  1. LangSmith: LangChain's official platform, SaaS service,not Open-Source
  2. LangFuse:Open-Source+ SaaS, LangSmith replacement, Ensemble LangChain integration, docking with OpenAI API(2.1K)
  3. Prompt Flow: MicrosoftCorp Open-Source + Azure AI cloud service, which can Ensemble Semantic Kernel (7.4K)

Prompt Flow

Learning

https://microsoft.github.io/promptflow/how-to-guides/quick-start.html

UI component

LangUI(1.6K)

agentlabs(0.5K)

other

classificationNameStar Quantityremark
LCEL
Python template languageJinja2

reference

classificationNameStar Quantityremark
frameObsidian CopilotAn interesting approach to how to use Semantic Searching and the BM25 implementation of OpenSearch
Tanuki个使用装饰器进行数据验证的 LLM 框架
LiteLLM5.7一个简单、轻量级的 LLM 封装器
AutoChain1.7langchain 的另一个替代品
griptape1.5一个具有稍微更好的内部编码标准的 langchain 替代品
txtai6.5has spinoffs for chat, workflows for medical/scientific papers, semantic search for developers and semantic search for headlines and story text
有用于聊天、医学/科学论文工作流程、开发者语义搜索以及标题和故事文本语义搜索的衍生产品
Llmflows0.5langchain 的另一个替代品,但在定义工作流程方面有一个有趣的方法
guidance16比传统的提示或链接更有效地控制现代语言模型*
libraryintructor3.6*

E2B:Frameworks and tools for AI products

CategoryTools/Platforms
Monitoring, Observability, AnalyticsLangsmith, Helicone, langfuse, context
FrontendAgentLabs
Runtime for LLMsE2B
Building frameworks & platformsLangchain, Haystack by deepset, Superagent, Poe, GenWorlds, LiteLLM, Rift by Morph, AutoGen, Vercel AI SDK, OpenGPTs, Huggging Face Agents, Fixie
Data integration, memory managementvectara, LlamaIndex, ABACUS.AI, cadea, SID Memory
API and routers for LLMsGenoss GPT, Martian, OpenRouter
Libraries for building AI productsCAMEL, ChatterBot (hasn't been updated in 5 annum)
OrchestrationAGIXT
Building & deploying LLMsBANANA

frameworks and tools for AI products