Tutorials

Pinecone is a fully managed vector database. Learn to use Pinecone for semantic search and RAG. Getting Started 1. Create Pinecone account 2. Get API key 3. Create an index 4. Start adding vectors Installation pip install pinecone-client Example import pinecone pinecone.init(api_key=”your-key”, environment=”us-west1″) pinecone.create_index(“my-index”, dimension=1536) index = pinecone.Index(“my-index”) index.upsert([(“id1”, embedding1, {“text”: “hello”})]) results = index.query(vector=query_embedding, … Read more

LangChain integrates with multiple LLM providers. Switch between models easily. Supported Providers ✅ OpenAI ✅ DeepSeek ✅ Anthropic ✅ HuggingFace ✅ Local models Using DeepSeek from langchain_openai import OpenAI llm = OpenAI( openai_api_key=”your-deepseek-key”, openai_api_base=”https://api.deepseek.com” ) Model Switching Change the model without rewriting code. Conclusion LangChain provides unified API for all LLMs!

Streaming provides real-time LLM responses. Implement streaming for better user experience. Streaming with Callbacks from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler llm = OpenAI(streaming=True, callbacks=[StreamingStdOutCallbackHandler()]) llm.invoke(“Tell me a story”) Async Streaming Use async callbacks for web applications. Benefits ✅ Faster perceived response ✅ Better UX ✅ Progressive display Conclusion Streaming improves user experience!

Callbacks monitor and log LangChain operations. Track tokens, costs, and performance. Callback Types ✅ StdOutCallbackHandler: Print to console ✅ FileCallbackHandler: Log to file ✅ Custom handlers Example from langchain.callbacks import StdOutCallbackHandler handler = StdOutCallbackHandler() chain = LLMChain(llm=llm, prompt=prompt, callbacks=[handler]) Custom Callback class MyCallback(BaseCallbackHandler): def on_llm_start(self, serialized, prompts): print(f”LLM started with {len(prompts)} prompts”) Conclusion Callbacks provide … Read more

Embeddings convert text into numerical vectors. Understand and use embeddings for semantic search. Embedding Models ✅ OpenAI Embeddings ✅ HuggingFace Embeddings ✅ Cohere Embeddings ✅ Local models Example from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() vector = embeddings.embed_query(“Hello world”) vectors = embeddings.embed_documents([“doc1”, “doc2”]) Similarity Calculation Use cosine similarity to compare embeddings. Conclusion Embeddings enable semantic … Read more

Vector stores enable efficient similarity search. Store and retrieve document embeddings. Popular Vector Stores ✅ Pinecone ✅ Chroma ✅ Weaviate ✅ FAISS ✅ Milvus Example with Chroma from langchain.vectorstores import Chroma from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() vectorstore = Chroma.from_texts(texts, embeddings) results = vectorstore.similarity_search(“query”, k=3) Benefits ✅ Fast similarity search ✅ Scalable ✅ Persistent … Read more

Text splitters break documents into manageable chunks. Split text for optimal LLM processing. Splitter Types CharacterTextSplitter: By characters RecursiveCharacterTextSplitter: Smart splitting TokenTextSplitter: By tokens SentenceTextSplitter: By sentences Example from langchain.text_splitter import RecursiveCharacterTextSplitter splitter = RecursiveCharacterTextSplitter( chunk_size=1000, chunk_overlap=200 ) chunks = splitter.split_text(long_text) Best Practices ✅ Use overlap for context ✅ Match chunk size to model limits … Read more

Document loaders ingest data from various sources. Load documents from files, APIs, and databases. Supported Formats ✅ PDF ✅ Word documents ✅ CSV ✅ JSON ✅ HTML ✅ Databases Example from langchain.document_loaders import PyPDFLoader, TextLoader pdf_loader = PyPDFLoader(“document.pdf”) pdf_docs = pdf_loader.load() text_loader = TextLoader(“file.txt”) text_docs = text_loader.load() Custom Loaders Create loaders for custom data sources. … Read more

Output parsers convert LLM text into structured data. Get reliable, parseable outputs from LLMs. Parser Types JsonParser: JSON output PydanticParser: Pydantic models ListParser: List output DatetimeParser: Date/time parsing Example from langchain.output_parsers import PydanticOutputParser from pydantic import BaseModel class Person(BaseModel): name: str age: int parser = PydanticOutputParser(pydantic_object=Person) prompt = PromptTemplate(template=”{query}\n{format_instructions}”) Benefits ✅ Type-safe outputs ✅ Validation … Read more

Prompt templates create reusable, dynamic prompts. Master prompt engineering with LangChain templates. Template Types PromptTemplate: String templates ChatPromptTemplate: Chat-specific FewShotPromptTemplate: With examples Example from langchain.prompts import PromptTemplate template = “”” You are a {role}. Task: {task} Context: {context} “”” prompt = PromptTemplate( template=template, input_variables=[“role”, “task”, “context”] ) Partial Variables Pre-fill some variables for reuse. Conclusion … Read more