langchain-development

# LangChain Development ## When to Use This Skill | Use this skill when... | Use a sibling skill instead when... | |---|---| | Building LCEL chains (prompt → model → parser) or RAG pipelines | You need stateful graph workflows — use `langgraph-agents` | | Working with chat models, prompt templates, or tool binding | You need hierarchical multi-agent orchestration — use `deep-agents` | | Adding LangChain to an existing TypeScript project | You are scaffolding a brand-new project — use `langchain-init` (`/langchain:init`) | | Implementing document loaders and vector stores | You only need a one-off SDK call without LangChain — use the provider SDK directly | ## Core Expertise LangChain JS/TS is a framework for building LLM applications: - Unified interface across model providers (OpenAI, Anthropic, Google, etc.) - Composable chains and agents - Built-in tool integration - RAG (Retrieval-Augmented Generation) support - LangSmith observability integration ## Installation ### Package Manager Setup ```bash # Core package npm install langchain # or pnpm add langchain # or bun add langchain # Model provider packages (install what you need) npm install @langchain/openai npm install @langchain/anthropic npm install @langchain/google-genai # Common integrations npm install @langchain/community # Community integrations npm install @langchain/textsplitters # Document splitting ``` ## Chat Models ### Basic Usage ```typescript import { ChatOpenAI } from "@langchain/openai"; import { ChatAnthropic } from "@langchain/anthropic"; import { HumanMessage, SystemMessage } from "@langchain/core/messages"; // OpenAI const openai = new ChatOpenAI({ model: "gpt-4o", temperature: 0, }); // Anthropic // Use a real, current model id (never an unversioned alias like "claude-haiku"), // and omit sampling params — Fable-generation models reject temperature/top_p/top_k. const anthropic = new ChatAnthropic({ model: "claude-haiku-4-5", }); // Invoke with messages const response = await openai.invoke([ new SystemMessage("You are a helpful assistant."), new HumanMessage("Hello!"), ]); ``` ### Streaming ```typescript const stream = await openai.stream([new HumanMessage("Tell me a story")]); for await (const chunk of stream) { process.stdout.write(chunk.content as string); } ``` ### Structured Output ```typescript import { z } from "zod"; const schema = z.object({ name: z.string().describe("The name"), age: z.number().describe("The age"), }); const structuredLlm = openai.withStructuredOutput(schema); const result = await structuredLlm.invoke("John is 30 years old"); // { name: "John", age: 30 } ``` ## Prompt Templates ### Basic Templates ```typescript import { ChatPromptTemplate } from "@langchain/core/prompts"; const prompt = ChatPromptTemplate.fromMessages([ ["system", "You are a {role}."], ["human", "{input}"], ]); const formatted = await prompt.invoke({ role: "helpful assistant", input: "Hello!", }); ``` ### Few-Shot Prompts ```typescript import { FewShotChatMessagePromptTemplate } from "@langchain/core/prompts"; const examples = [ { input: "2+2", output: "4" }, { input: "3+3", output: "6" }, ]; const fewShotPrompt = new FewShotChatMessagePromptTemplate({ examplePrompt: ChatPromptTemplate.fromMessages([ ["human", "{input}"], ["ai", "{output}"], ]), examples, inputVariables: ["input"], }); ``` ## Chains (LCEL) ### Basic Chain ```typescript import { ChatOpenAI } from "@langchain/openai"; import { ChatPromptTemplate } from "@langchain/core/prompts"; import { StringOutputParser } from "@langchain/core/output_parsers"; const prompt = ChatPromptTemplate.fromTemplate("Tell me a joke about {topic}"); const model = new ChatOpenAI(); const parser = new StringOutputParser(); // Chain with pipe operator const chain = prompt.pipe(model).pipe(parser); const result = await chain.invoke({ topic: "programming" }); ``` ### Parallel Chains ```typescript import { RunnableParallel } from "@langchain/core/runnables"; const parallel = RunnableParallel.from({ joke: jokeChain, poem: poemChain, }); const results = await parallel.invoke({ topic: "cats" }); // { joke: "...", poem: "..." } ``` ### Branching ```typescript import { RunnableBranch } from "@langchain/core/runnables"; const branch = RunnableBranch.from([ [(x) => x.type === "math", mathChain], [(x) => x.type === "code", codeChain], defaultChain, // Fallback ]); ``` ## Agentic Optimizations | Context | Command/Pattern | | --------------- | -------------------------------------- | | Quick test | `npx tsx --test src/**/*.test.ts` | | Type check | `npx tsc --noEmit` | | Debug traces | Set `LANGCHAIN_TRACING_V2=true` | | Reduce tokens | Use `StringOutputParser` for text-only | | Stream output | Use `.stream()` instead of `.invoke()` | | Batch requests | Use `.batch([inputs])` for parallel | | Cache responses | Use `InMemoryCache` for repeated calls | ## Quick Reference ### Environment Variables | Variable | Description | | ---------------------- | ------------------------ | | `OPENAI_API_KEY` | OpenAI API key | | `ANTHROPIC_API_KEY` | Anthropic API key | | `LANGCHAIN_TRACING_V2` | Enable LangSmith tracing | | `LANGCHAIN_API_KEY` | LangSmith API key | | `LANGCHAIN_PROJECT` | LangSmith project name | ### Common Imports | Import | Package | | -------------------- | -------------------------------- | | `ChatOpenAI` | `@langchain/openai` | | `ChatAnthropic` | `@langchain/anthropic` | | `ChatPromptTemplate` | `@langchain/core/prompts` | | `StringOutputParser` | `@langchain/core/output_parsers` | | `tool` | `@langchain/core/tools` | | `RunnableSequence` | `@langchain/core/runnables` | ### Key Packages | Package | Purpose | | ---------------------- | ---------------------- | | `langchain` | Core framework | | `@langchain/core` | Base abstractions | | `@langchain/openai` | OpenAI integration | | `@langchain/anthropic` | Anthropic integration | | `@langchain/community` | Community integrations | | `@langchain/langgraph` | Graph-based agents | For TypeScript configuration, tool definition and binding, RAG pipelines, and ReAct agents, see [REFERENCE.md](REFERENCE.md).