背景
继上次调研后，很久没更新AI相关技术了，这次一并更新。
应用
全本地部署的ai https://github.com/mudler/LocalAI
应用框架
数据索引框架 data index framework
llamaindex
https://docs.llamaindex.ai/en/stable/ https://github.com/run-llama/llama_index
https://docs.llamaindex.ai/en/stable/use_cases/agents/ Property Graph Index可以用neo4j
https://docs.llamaindex.ai/en/stable/module_guides/storing
Vector Stores：Elasticsearch、Lantern、OpenSearch、Postgres、Qdrant
Document Stores：Mongo、Redis
Index Stores：Mongo、Redis
Chat Stores：Redis
Key-Value Stores：Mongo
Note: At the moment, these storage abstractions are not externally facing. https://docs.llamaindex.ai/en/stable/module_guides/storing/customization/
ai agents framework
自动化框架 restack
agents框架 eko
https://github.com/FellouAI/eko
pydantic
https://github.com/pydantic/pydantic-ai
composio
https://github.com/ComposioHQ/composio/tree/master/python/swe?utm_source=website
https://docs.composio.dev/framework/llamaindex
https://docs.composio.dev/framework/langchain
比langchain底层的工作流agent编排框架 LangGraph
prompt框架
https://github.com/langchain4j/langchain4j
https://docs.spring.io/spring-ai/reference/index.html
https://github.com/spring-projects/spring-ai
AI上下文协议框架MCP
https://modelcontextprotocol.io/introduction
https://modelcontextprotocol.io/quickstart
https://github.com/modelcontextprotocol
https://github.com/modelcontextprotocol/servers
知识库
上层
https://github.com/getzep/graphiti
向量数据库
https://lancedb.github.io/lancedb/concepts/storage/ embeddedable|localfile|s3
https://www.trychroma.com/ embeddedable|localfile
https://milvus.io/ need etcd、Pulsar embeddedable|s3
es dense_vector
0 Comments