langchain-oss-primer（LangChain OSS primer 搜索）

It answers three questions every project must resolve upfront:

1. Which framework? — LangChain, LangGraph, or Deep Agents
2. Which agent archetype? — maps your use case to the right API and patterns
3. What to install and which skills to load next

Load this skill first. Once you've decided on a framework and agent type, follow the "Next Skills" section at the bottom — it tells you exactly which skills to invoke next based on your choices.

Step 1 — Pick Your Framework

The three frameworks are layered, not competing. Each builds on the one below:

┌─────────────────────────────────────────┐
│              Deep Agents                │  ← batteries included
│   (planning, memory, skills, files)     │
├─────────────────────────────────────────┤
│               LangGraph                 │  ← custom orchestration
│    (nodes, edges, state, persistence)   │
├─────────────────────────────────────────┤
│               LangChain                 │  ← foundation
│      (models, tools, prompts, RAG)      │
└─────────────────────────────────────────┘

Answer these questions in order:

Higher layers depend on lower ones only when necessary — you can mix them. A LangGraph graph can be a subagent inside Deep Agents; LangChain tools work inside both.

Middleware is a concept specific to Deep Agents (explicit middleware layer). LangGraph has no middleware — behavior is wired directly into nodes and edges. If a user asks for built-in hooks or automatic middleware, route to Deep Agents.

Step 2 — Pick Your Agent Archetype

Once you've chosen a framework, match your use case to the right API and pattern.

LangChain — use create_agent()

Best for single-purpose agents in a ReACT style with a fixed tool set. No built-in planning, memory management, or delegation.

All LangChain agents use create_agent(model, tools=[...]). Next skill: langchain-fundamentals.

LangGraph — use StateGraph

Best when you need explicit, deterministic control flow.

LangGraph agents use StateGraph(State) with explicit add_node, add_edge, add_conditional_edges. Next skill: langgraph-fundamentals.

Deep Agents — use create_deep_agent()

Best when the agent needs to manage its own work: planning tasks, remembering users across sessions, delegating to specialists, or managing files autonomously.

All Deep Agents use create_deep_agent(model, tools=[...],...). Next skill: deep-agents-core — load it immediately after deciding on Deep Agents.

Deep Agents built-in middleware

Six components pre-wired out of the box. First three are always active; the rest are opt-in:

You configure middleware — you don't implement it.

Step 3 — Set Up Your Dependencies

Environment requirements

Always use LangChain 1.0+. LangChain 0.3 is maintenance-only until December 2026 — do not start new projects on it.

Core packages — always required

Orchestration — add based on your framework choice

Model providers — pick the one(s) you use

Common tools & retrieval — add as needed

Prefer dedicated integration packages over langchain-community when one exists — they are independently versioned and more stable.

Dependency templates

Add your model provider:

langchain-openai | langchain-anthropic | langchain-google-genai |...

Add tools/retrieval as needed:

langchain-tavily | langchain-chroma | langchain-text-splitters |...

</python>
</ex-langchain-python>

<ex-langgraph-python>
<python>
LangGraph project — provider-agnostic starting point.

requirements.txt

langchain>=1.0,<2.0 langchain-core>=1.0,<2.0 langgraph>=1.0,<2.0 langsmith>=0.3.0

Add your model provider:

langchain-openai | langchain-anthropic | langchain-google-genai |...

</python>
</ex-langgraph-python>

<ex-langgraph-typescript>
<typescript>
LangGraph project — provider-agnostic starting point.

{ "dependencies": { "@langchain/core": "^1.0.0", "langchain": "^1.0.0", "@langchain/langgraph": "^1.0.0", "langsmith": "^0.3.0" } }

# Add your model provider:

# langchain-openai | langchain-anthropic | langchain-google-genai |...

</python> </ex-deepagents-python>

<ex-deepagents-typescript> <typescript> Deep Agents project — provider-agnostic starting point.

{
  "dependencies": {
    "deepagents": "latest",
    "@langchain/core": "^1.0.0",
    "langchain": "^1.0.0",
    "langsmith": "^0.3.0"
  }
}

Step 4 — Set Your Environment Variables

Model provider — set the one(s) you use

OPENAI_API_KEY= ANTHROPIC_API_KEY= GOOGLE_API_KEY= MISTRAL_API_KEY= GROQ_API_KEY= COHERE_API_KEY= FIREWORKS_API_KEY= TOGETHER_API_KEY= HUGGINGFACEHUB_API_TOKEN=

Common tool/retrieval services

TAVILY_API_KEY= PINECONE_API_KEY=

</environment-variables>

---

## Step 5 — Load the Right Skill Next

Based on the framework and archetype you chose above, invoke these skills **now** before writing any code:

<next-skills>

### If you chose LangChain

| Your archetype | Load next |
|----------------|-----------|
| Any LangChain agent (QA bot, SQL, search, RAG, tool-calling) | **`langchain-fundamentals`** — always |
| Adding external tools/packages (Tavily, Pinecone, etc.) | **`langchain-dependencies`** — package patterns and version guidance |
| Need streaming or async responses | **`langchain-fundamentals`** then `langgraph-fundamentals` |

### If you chose LangGraph

| Your archetype | Load next |
|----------------|-----------|
| Any LangGraph graph | **`langgraph-fundamentals`** — always |
| Approval pipeline, HITL, or pause/resume | **`langgraph-fundamentals`** + `langgraph-human-in-the-loop` |
| State that must survive restarts or cross-thread memory | **`langgraph-persistence`** |
| Streaming output token by token | **`langgraph-fundamentals`** |

### If you chose Deep Agents

**Always load `deep-agents-core` first — it is the mandatory starting point for any Deep Agents project.**

| Your archetype | Load next (after `deep-agents-core`) |
|----------------|--------------------------------------|
| Research Assistant — delegates to specialist subagents | **`deep-agents-orchestration`** — subagent config, TodoList, HITL |
| Personal Assistant — remembers users across sessions | **`deep-agents-memory`** — MemoryMiddleware, Store backends |
| Coding Assistant — reads/writes files, plans refactors | `deep-agents-core` is sufficient; add `deep-agents-orchestration` if using HITL |
| Orchestrator — routes work across multiple named subagents | **`deep-agents-orchestration`** — SubAgentMiddleware patterns |
| Long-running task agent — survives restarts | **`deep-agents-memory`** + `deep-agents-orchestration` |
| On-demand skills agent | `deep-agents-core` covers SkillsMiddleware setup |
</next-skills>