Practical notes: Learning LangGraph : A Journey Through Agents, Blackboards

The following notes reconstruct a practical path around “Learning LangGraph : A Journey Through Agents, Blackboards, and Bottlenecks”. Emphasis stays on contracts, checks, and drop-in code placeholders rather than motivational framing. When working through Overview, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments.

Step 1: The minimum agent

Step 1: The minimum agent works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph. Keep graph state flat and typed. Nested blobs hide which node wrote which field and break resume after interrupts.

from langchain.agents import create_agent          # langchain 1.x (NOT langgraph.prebuilt)
from langchain.tools import tool
from langchain_anthropic import ChatAnthropic

# 1. Model
llm = ChatAnthropic(model="claude-haiku-4-5", max_tokens=1024)

# 2. Tool — a plain function + @tool.
#    - type hints (drug_name: str) become the input schema
#    - the docstring is what the LLM reads to decide WHEN to call it
#    The model only "sees" the signature + docstring, so write the docstring
#    like an API description.
@tool
def get_drug_targets(drug_name: str) -> str:
    """Return the protein targets for a given drug name."""
    return "EGFR, BRAF"   # real data-source call goes here

# 3. System prompt — a plain string
SYSTEM_PROMPT = "You are a research assistant. Use the tools to gather evidence."

# 4. Build the agent
agent = create_agent(model=llm, tools=[get_drug_targets], system_prompt=SYSTEM_PROMPT)

# 5. Run it — pass in {"messages": [...]}, get the full message list back
result = agent.invoke({"messages": [{"role": "user", "content": "What does metformin target?"}]})
print(result["messages"][-1].content)   # last message = final answer

@tool
def calculator(expression: str) -> str:
    """Evaluate a basic arithmetic expression like '2 + 2 * 3'."""
    return str(eval(expression))   # fine for learning; never eval untrusted input in prod

Step 2: Getting two agents to talk

Step 2: Getting two agents to talk works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Keep graph state flat and typed. Nested blobs hide which node wrote which field and break resume after interrupts.

Option 1: Messages

Option 1: Messages works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Keep graph state flat and typed. Nested blobs hide which node wrote which field and break resume after interrupts. Option 1: Messages works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments.

Option 2: Blackboard

For Option 2: Blackboard, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph. Put human approval on edges that spend money or change production data. Compile-time wiring does not equal business completeness.

Step 3: Getting typed data out of one agent

For Step 3: Getting typed data out of one agent, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Put human approval on edges that spend money or change production data. Compile-time wiring does not equal business completeness.

from pydantic import BaseModel, Field

class ResearchOutput(BaseModel):
    attraction: str = Field(description="2 sentences about the attraction")
    city: str = Field(description="1 sentence about the city it is in")

agent = create_agent(
    model=llm,
    tools=[attraction_info, city_info],
    system_prompt=SYSTEM_PROMPT,
    response_format=ResearchOutput,   # <- agent emits this typed object
)

result = agent.invoke({"messages": [...]})
out = result["structured_response"]   # a ResearchOutput instance
out.attraction   # str
out.city         # str

def research_node(state):
    out = research(state["topic"])           # ResearchOutput
    return {"research": out.attraction, "city": out.city}

The coupling this creates

For The coupling this creates, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Put human approval on edges that spend money or change production data. Compile-time wiring does not equal business completeness. For The coupling this creates, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments.

SYSTEM_PROMPT = "...\n" + "\n".join(f"- {t.name}: {t.description}" for t in TOOLS)

Four variants, one dial each

When working through Four variants, one dial each, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph. Checkpoint after expensive steps. Resume should not re-bill the same LLM call when an operator retries a later node.

Step 4: Why is this so slow — and what actually fixes it

When working through Step 4: Why is this so slow — and what actually fixes it, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Checkpoint after expensive steps. Resume should not re-bill the same LLM call when an operator retries a later node.

def _crowd_score_sync(attraction: str) -> str:   # plain blocking function
    total = 0
    for i in range(60_000_000):
        total += i
    return f"... {total % 100} ..."

@tool
async def crowd_score(attraction: str) -> str:
    return await asyncio.to_thread(_crowd_score_sync, attraction)

Why the ReAct agent stays slow regardless

When working through Why the ReAct agent stays slow regardless, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Checkpoint after expensive steps. Resume should not re-bill the same LLM call when an operator retries a later node. When working through Why the ReAct agent stays slow regardless, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments.

Step 5: Running whole pipelines concurrently

Step 5: Running whole pipelines concurrently works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph. Keep graph state flat and typed. Nested blobs hide which node wrote which field and break resume after interrupts.

Step 6: Making the caching actually work

Step 6: Making the caching actually work works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Keep graph state flat and typed. Nested blobs hide which node wrote which field and break resume after interrupts.

system_prompt = "You are a travel research agent..."   # string: no keys

content = [{                                            # dict: has keys
    "type": "text",
    "text": "You are a travel research agent...",
    "cache_control": {"type": "ephemeral"},            # <- attaches here
}]

from langchain_anthropic.middleware import AnthropicPromptCachingMiddleware

agent = create_agent(
    model=llm, tools=[...], system_prompt=SYSTEM_PROMPT,  # plain string
    middleware=[AnthropicPromptCachingMiddleware()],
)

Step 7: Making the blackboard durable

Step 7: Making the blackboard durable works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Keep graph state flat and typed. Nested blobs hide which node wrote which field and break resume after interrupts. Step 7: Making the blackboard durable works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments.

Operational checklist

For Operational checklist, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state.

Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion.

Put human approval on edges that spend money or change production data. Compile-time wiring does not equal business completeness.

Write a short runbook: how to rotate keys, how to drain the queue, how to roll back the last ingest.

Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments.

Put human approval on edges that spend money or change production data. Compile-time wiring does not equal business completeness.

Before promoting the stack, freeze versions, capture a golden transcript for the critical path, and confirm rollback steps. Shared environments need rate limits, tenancy checks, and a clear owner for secret rotation. Prefer boring reliability over clever one-off demos.

Batch note for 7a480fca3eb9: keep provider keys out of the repo, set a per-session token ceiling, and store transcripts next to the eval fixtures so later model swaps stay comparable.

When working through hardening note 0, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish.

Hardening detail 0/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

hardening note 1 works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion.

Hardening detail 1/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

For hardening note 2, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

Hardening detail 2/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

When working through hardening note 3, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline.

Hardening detail 3/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

hardening note 4 works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments.

Hardening detail 4/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

For hardening note 5, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish.

Hardening detail 5/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

When working through hardening note 6, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion.

Hardening detail 6/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

hardening note 7 works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

Hardening detail 7/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

For hardening note 8, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline.

Hardening detail 8/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

When working through hardening note 9, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments.

Hardening detail 9/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

hardening note 10 works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish.

Hardening detail 10/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

For hardening note 11, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion.

Hardening detail 11/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

When working through hardening note 12, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

Hardening detail 12/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

hardening note 13 works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline.

Hardening detail 13/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

For hardening note 14, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments.

Hardening detail 14/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

When working through hardening note 15, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish.

Hardening detail 15/913: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.