內容描述
https://www.patreon.com/kfsoft 基於大型語言模型 (LLM) 的應用程式，涵蓋 Ollama API、語義搜索、vector store、LangChain、LangGraph 等。包括使用 Ollama 進行本地 LLM 程式開發，結合 RAG (檢索增強生成) 提升生成質量，接入 Telegram 機器人等。此外還會探討 faiss 搜尋索引，如 IndexFlatL2、IndexFlatIP、IndexHNSWFlat、IndexIVFFlat，並會探討 LangChain 和 LangGraph 架構。
00:00 Introduction
00:45 Downloadable materials (Patreon)
04:37 Roadmap
07:01 What can LLM offer
08:06 Ollama models and API
10:24 Local and remote connection
12:06 Limitation of LLM
13:30 RAG concept
14:46 Semantic search
17:23 Vector Store
19:01 Available APIs
22:03 Download ollama models
27:42 Ollama API
32:37 ollama.generate()
40:57 ollama.chat()
44:49 Function calling (ollama)
55:57 Chatbot
01:01:31 OpenAI compatible
01:03:41 client.chat.completion()
01:07:05 client.completion()
01:09:31 Rate limits
01:11:25 Function calling (allibaba)
01:14:26 Telegram bot integration
01:42:11 Vector search
01:45:41 Sentence transformer - text to vector
01:49:06 IndexFlatL2
02:00:51 IndexFlatIP
02:04:29 IndexHNSWFlat
02:09:39 IndexIVFFlat
02:18:45 Semantic search
02:25:14 Nomic embed model encoding
02:29:41 Document level indexing
02:35:38 Paragraph level indexing
02:39:07 Sentence level indexing
02:41:41 Paragraph level RAG
02:53:11 LangChain
02:55:19 OllamaLLM
02:59:28 ChatOllama
03:02:37 ChatOpenAI
03:03:49 Response cache
03:07:18 PromptTemplate
03:17:33 Chain examples
03:25:34 Output parsers
03:38:35 Chat messages formats
03:41:40 ChatPromptTemplate
03:48:29 Chat loop
03:51:57 Structured Output
03:57:10 Callabe
04:02:33 Runnable
04:06:20 RunnableLambda
04:13:12 Routing
04:16:58 RunnableParallel
04:23:02 RunnableParallel and routing
04:32:21 ainvoke()
04:34:35 astream()
04:38:21 LangGraph
04:56:14 Summarize history with LangGraph
05:10:48 Graph streaming modes
05:25:59 Slogan generation and review
05:30:46 LLM chat with LLM
05:41:53 ToolNode and bind_tools
05:56:05 create_react_agent()
06:03:13 Document Loaders
06:07:17 Text Loader
06:12:45 FAISS Vector Store
06:22:19 Retriever
06:26:25 create_retreival
