/ Home
LangChain Basics and Intermediate
Note: LC HandsOn
CH01 LangChain Getting Started
- Follow the installation video report
- OpenAI API key issuance and testing
- LangSmith tracking settings
- Using the OpenAI API (GPT-4o multimodal)
- LangChain Expression Language (LCEL)
- LCEL interface
- Runnable
CH02 Prompt
- Prompt
- FewShotPromptTemplate
- LangChain Hub
- Personalized p rompt (uploaded to Hub)
CH03 Output Parsers
- Pydantic output parser (PydanticOutputParser)
- Comma SeparatedListOutputParser
- StructuredOuputParser
- JSON output parser (JsonOutputParser)
- Data frame output parser (PandasDataFrameOutputParser)
- Date format output parser (DatetimeOutputParser)
- EnumOutputParser
- OutputFixingParser
CH04 model (Model)
- LLM models available
- Caching (Cache)
- Model Serialization-Save and Call
- Check token usage
- Google Generative AI
- HuggingFace Endpoints
- HuggingFace Local
- HuggingFace Pipeline
- Olama
- GPT4ALL
- Video (Video) Quality Response LLM (Gemini)
CH05 memory (Memory)
- ConversationBufferMemory
- ConversationBufferWindowMemory
- ConversationTokenBufferMemory
- ConversationEntityMemory
- ConversationKGMemory
- Conversation SummaryMemory
- Vector Storage Search Memory (VectorStoreRetrieverMemory)
- Add memory to LCEL Chain
- Save chat to SQLite
- Add ChatMessageHistory to RunnableWithMessageHistory
CH06 Document Loader
- Structure of Document
- Hangeul (HWP)
- CSV
- Excel
- Word
- PowerPoint
- Web document (WebBaseLoader)
- Text (TextLoader)
- JSON
- Arxiv
- UpstageLayoutAnalysisLoader
- LlamaParser
CH07 text split (Text Splitter)
- Character text split (CharacterTextSplitter)
- Securitic text text split (RecursiveCharacterTextSplitter)
- Token Text Split (TokenTextSplitter)
- Semantic chunker
- Code splitting (Python, Markdown, JAVA, C++, C#, GO, JS, Latex, etc.)
- Markdownheader Text Split (MarkdownheaderTextSplitter)
- Split HTML header text (HTMLHeaderTextSplitter)
- Regressive JSON split (RecursiveJsonSplitter)
CH08 Embedding
- OpenAIEmbeddings
- CacheBackedEmbeddings
- HuggingFace Embeddings
- UpstageEmbeddings
- OllamaEmbeddings
- GPT4ALL embedding
- Llama CPP embedding
CH09 VectorStore
- Chroma
- FAISS
- Pinecone CH10 finder (Retriever)
- VectorStore-backed Retriever
- Contextual CompressionRetriever
- EnsembleRetriever
- Long context rearrangement (LongContextReorder)
- ParentDocumentRetriever
- MultiQueryRetriever
- MultiVectorRetriever
- Self Query Retriever
- Time Weighted Vector StoreRetriever
- Hangeulocyte analyzer (Kiwi, Kkma, Okt) + BM25 finder
- EnsembleRetriever with Convex Combination (CC)
CH11 Reranker
- Cross Encoder Reranker
- Cohere Reranker
- Jina Reranker
- FlashRank Reranker
CH12 Retrieval Augmented Generation (RAG)
- PDF document based QA (Question-Answer)
- Naver News Agency QA (Question-Answer)
- Various module utilizers by function of RAG
- RAPTOR: Long Context Summary
- RAG chain to remember interactive
CH13 LangChain Expression Language (LCEL)
- RunnablePasstrough
- Runnable structure (graph) review
- RunnableLambda
- LLM chain routing (RunnableLambda, RunnableBranch)
- RunnableParallel
- Dynamic attribute assignment (configurable_fields, configurable_alternatives)
- Runnable configuration with @chain decorator
- RunnableWithMessageHistory
- Custom generator
- Runtime Arguments Binding
- Fallback model designation
CH14 chain (Chains)
- Document summary
- SQL
- Structured output chain (with_structed_output)
CH15 Evaluation (Evaluations)
- Synthetic test dataset generation (RAGAS)
- Evaluation using RAGAS
- Upload data set for evaluation generated (HuggingFace Dataset)
- LangSmith dataset generation
- LLM-as-Judge
- Embedding-based evaluation (embedding_distance)
- Custom LLM evaluation
- Rouge, BLEU, METEOR, SemScore based heuristic evaluation
- Experimental (Experiment) evaluation comparison
- Assessment of the summary method
- Groundedness (Halucination) Assessment
- Experimental comparison (Pairwise Evaluation)
- Repeat evaluation
- Automating evaluation using online evaluation
CH16 agent (Agent)
- Tools
- Tool Binding (Binding Tools)
- Agent
- Agent utilizing Claude, Gemini, Ollama, Together.ai
- Iteration function and human intervention (Human-in-the-loop)
- Agentic RAG
- CSVExcel Data Analysis Agent
- Toolkits Use Agent
- RAG + Image Generator Agent (Reporting)
- Discussion agent with tools (Two Agent Debates with Tools)
CH17 LangGraph
01. Core function
- Python grammar often appeared in LangGraph
- Build chatbots using LangGraph
- Building an Agent using LangGraph
- Add memory to Agent
- Node’s step-by-step streaming output
- Human-in-the-loop (human intervention)
- State correction and replay through reverting of intermediate intervention
- Add a node to ask a person
- Delete message (RemoveMessage)
- How to call a tool using ToolNode
- How to generate branches for parallel node execution
- How to add a conversation history summary
- How to add and use subgraphs
- How to convert input and output of subgraph
- Everything in LangGraph streaming mode
02. Structure design
- Default graph generation
- Naive RAG
- Relevance Checker module added
- Add web search module
- Add query rewrite module
- Agentic RAG
- Adaptive RAG
03. Use Cases
- Agent conversation simulation (customer response scenario)
- User-required port-based meta prompt generation agent
- CRAG (Corrective RAG)
- Self-RAG
- Plan-and-Execute
- Multi-Agent Collaboration Network
- Multi-Agent Supervisor
- Hierarchical Multi-Agent Teams
- Agent interacting with SQL database
- Multi-agent for research introducing the STORM concept
CH18 Other Information
- StreamEvent type theorem
Ref:
https://wikidocs.net/233343