donbr · July 8, 2026 03:45
diff --git a/session9-agent-servers-cheatsheet.md b/session9-agent-servers-cheatsheet.md
diff --git a/session9-agent-servers-learning-journey.md b/session9-agent-servers-learning-journey.md
diff --git a/smoke_test_sdk.py b/smoke_test_sdk.py
 """SDK smoke test against a running agent server (README Part 1, step 4).

 Streams a run through the LangGraph SDK the same way a production client would —
 this is the server-based counterpart to ``main.py`` (which invokes the graph
 in-process). Start the server first:

    uv run langgraph dev            # serves http://localhost:2024

 then run this against it:

    uv run python smoke_test_sdk.py
    uv run python smoke_test_sdk.py agent_with_helpfulness "What vaccines does a kitten need?"

 Usage:
    python smoke_test_sdk.py [ASSISTANT_ID] [QUESTION]

 Notes:
    * ASSISTANT_ID is the GRAPH ID from langgraph.json — ``simple_agent`` (default)
      or ``agent_with_helpfulness``. It is NOT ``"agent"``: the ``assistants`` block
      in langgraph.json is not materialized on the local server, so ``"agent"`` 422s
      with "Invalid assistant". (See LEARNING_JOURNEY.md, Part 1 gotcha #1.)
    * Uses ``get_sync_client`` so the README's plain ``for chunk in ...`` loop works
      without asyncio boilerplate.
 """

 from __future__ import annotations

 import os
 import sys

 from langgraph_sdk import get_sync_client

 URL = os.environ.get("AGENT_URL", "http://localhost:2024")


 def main() -> None:
    assistant_id = sys.argv[1] if len(sys.argv) > 1 else "simple_agent"
    question = (
        sys.argv[2]
        if len(sys.argv) > 2
        else "How often should I deworm my cat?"
    )

    client = get_sync_client(url=URL)
    print(f"→ {URL}  assistant={assistant_id!r}")
    print(f"? {question}\n")

    for chunk in client.runs.stream(
        None,  # threadless run
        assistant_id,
        input={"messages": [{"role": "human", "content": question}]},
        stream_mode="updates",
    ):
        if not isinstance(chunk.data, dict):
            continue
        for node, payload in chunk.data.items():
            messages = payload.get("messages") if isinstance(payload, dict) else None
            if not messages:
                continue
            last = messages[-1]
            content = last.get("content") if isinstance(last, dict) else ""
            tool_calls = last.get("tool_calls") if isinstance(last, dict) else None
            if tool_calls:
                names = ", ".join(tc.get("name", "?") for tc in tool_calls)
                print(f"[{node}] → tool call: {names}")
            elif content:
                print(f"[{node}] {str(content)[:400]}")


 if __name__ == "__main__":
    main()
You want to…	Reach for	One-liner
Tell LangGraph what graphs exist	`langgraph.json` `graphs`	`"simple_agent": "app.graphs.simple_agent:graph"` (id → `module:attribute`)
Export a runnable graph	a module-level `graph`	`graph = create_agent(model=..., tools=..., system_prompt=...)`
Run the agent API locally	`langgraph dev`	serves `http://localhost:2024` + opens LangGraph Studio
Self-host a prod container	`langgraph up`	local Docker; you own scaling/uptime/auth
Deploy to LangSmith cloud	`langgraph deploy`	managed build+host (needs LangSmith Plus) → Deployment URL
Talk to the agent in Python	`langgraph_sdk.get_client`	`client.runs.stream(None, "agent", input={...}, stream_mode="updates")`
Stream the agent into a UI	`@langchain/react` `useStream`	`useStream({ apiUrl: "/api", assistantId: "simple_agent" })`
Hide the API key from the browser	`langgraph-nextjs-api-passthrough`	`initApiPassthrough({ apiUrl, apiKey, runtime: "edge" })` in `route.ts`
Mark a var public in Next.js	`NEXT_PUBLIC_` prefix	only `NEXT_PUBLIC_*` reaches the browser; everything else is server-only
Deploy the UI	`vercel` / `vercel --prod`	set `LANGGRAPH_API_URL`, `LANGSMITH_API_KEY`, `NEXT_PUBLIC_API_URL`
See what the agent actually did	LangSmith traces / Studio	every run is traced; Studio steps through nodes + tool calls
Component	Role
`app/models.py`	model factory `get_chat_model()` (default `gpt-5.4-mini`, `temperature=0`)
`app/tools.py`	tool belt: `TavilySearch`, `ArxivQueryRun`, `retrieve_information` (RAG)
`app/rag.py`	loads `data/*.pdf` → split 750/0 → `text-embedding-3-small` → in-memory Qdrant
`app/graphs/simple_agent.py`	`create_agent(...)` → exports compiled `graph`
`langgraph.json`	manifest: `graphs` + `assistants` the server exposes
`frontend/app/api/[...path]/route.ts`	secure passthrough — injects the key server-side
`frontend/components/chat.tsx`	`useStream` chat UI pointed at `/api`
Plan	Price	Free traces/mo	Cloud deploy?
Developer	$0/seat, 1 seat	5,000	❌
Plus	$39/seat/mo	10,000	✅ 1 free Dev deployment
Enterprise	custom	custom	✅ + self-hosted/hybrid
Topic	Source	Why it matters
`langgraph.json` structure & fields	LangChain docs — Application structure (`/langsmith/application-structure`)	The real manifest keys: `dependencies`, `graphs`, `env`, `python_version`.
Local server	LangChain docs — Run a local server (`/oss/python/langgraph/local-server`)	`langgraph dev` needs `langgraph-cli[inmem]`; needs a LangSmith API key.
Assistants vs graphs	LangChain docs — Default assistants (`/langsmith/assistants`), Configuration (`/langsmith/configuration-cloud`)	The run identifier is the graph ID (key under `graphs`); each graph gets a default assistant. Drives the `assistantId` gotcha below.
Plan	Price	Free traces/mo	Agent deployments?
Developer	$0/seat · 1 seat · then pay-as-you-go	5,000	❌ no deployment access
Plus	$39/seat/mo · then PAYG	10,000	✅ 1 free Dev deployment, unlimited runs
Enterprise	custom	custom	✅ + self-hosted / hybrid
Part	Status	How to prove it
1. Package + `langgraph dev` + SDK	✅ verifiable	server up on :2024, both graphs, SDK stream, helpfulness loop
2. Deploy on LangSmith	📄 needs paid account	grounded CLI/UI/self-host steps
3. Next.js frontend + proxy + `useStream`	✅ verifiable	browser run, full stream through `/api`
4. Deploy on Vercel	📄 needs account	grounded steps + env wiring
Q&A + Activities	🧠 your turn	answer in `README.md`; demo in the Loom
	"""SDK smoke test against a running agent server (README Part 1, step 4).

	Streams a run through the LangGraph SDK the same way a production client would —
	this is the server-based counterpart to ``main.py`` (which invokes the graph
	in-process). Start the server first:

	uv run langgraph dev # serves http://localhost:2024

	then run this against it:

	uv run python smoke_test_sdk.py
	uv run python smoke_test_sdk.py agent_with_helpfulness "What vaccines does a kitten need?"

	Usage:
	python smoke_test_sdk.py [ASSISTANT_ID] [QUESTION]

	Notes:
	* ASSISTANT_ID is the GRAPH ID from langgraph.json — ``simple_agent`` (default)
	or ``agent_with_helpfulness``. It is NOT ``"agent"``: the ``assistants`` block
	in langgraph.json is not materialized on the local server, so ``"agent"`` 422s
	with "Invalid assistant". (See LEARNING_JOURNEY.md, Part 1 gotcha #1.)
	* Uses ``get_sync_client`` so the README's plain ``for chunk in ...`` loop works
	without asyncio boilerplate.
	"""

	from __future__ import annotations

	import os
	import sys

	from langgraph_sdk import get_sync_client

	URL = os.environ.get("AGENT_URL", "http://localhost:2024")


	def main() -> None:
	assistant_id = sys.argv[1] if len(sys.argv) > 1 else "simple_agent"
	question = (
	sys.argv[2]
	if len(sys.argv) > 2
	else "How often should I deworm my cat?"
	)

	client = get_sync_client(url=URL)
	print(f"→ {URL} assistant={assistant_id!r}")
	print(f"? {question}\n")

	for chunk in client.runs.stream(
	None, # threadless run
	assistant_id,
	input={"messages": [{"role": "human", "content": question}]},
	stream_mode="updates",
	):
	if not isinstance(chunk.data, dict):
	continue
	for node, payload in chunk.data.items():
	messages = payload.get("messages") if isinstance(payload, dict) else None
	if not messages:
	continue
	last = messages[-1]
	content = last.get("content") if isinstance(last, dict) else ""
	tool_calls = last.get("tool_calls") if isinstance(last, dict) else None
	if tool_calls:
	names = ", ".join(tc.get("name", "?") for tc in tool_calls)
	print(f"[{node}] → tool call: {names}")
	elif content:
	print(f"[{node}] {str(content)[:400]}")


	if __name__ == "__main__":
	main()