Thread ID ignored when deploying via langgraph dev

[muelerma]

I have a simple agent I want to test locally. I start the agent with langgraph dev, then send curl requests to it. In LangSmith UI, I can see that it uses a different Thread for each Run, even though I set the thread_id in configurable (see below). That means some request like "My name is John Snow" followed by "What is my name?" is unanswered by the agent.

How do I correctly set the thread_id?

curl --request POST \
    --url http://localhost:2024/runs/stream \
    --header 'Content-Type: application/json' \
    --data '{
    "assistant_id": "agent",
    "input": {
        "messages": [
            {
                "role": "user",
                "content": "My name is John Snow"
            }
        ]
    },
    "metadata": {},
    "config": {
        "configurable": {"thread_id": 99}
    },
    "multitask_strategy": "reject",
    "stream_mode": [
        "values"
    ]
}'

Agent is just a single tool with an LLM, as presented in the Langgraph basics: https://langchain-ai.github.io/langgraph/tutorials/get-started/2-add-tools/

I deploy it using langgraph dev, so maybe the only other relevant file is langgraph.json

{
    "dependencies": ["."],
    "graphs": {
        "agent": "./eg_agent.py:graph"
    },
    "env": ".env"
}

[hinthornw]

Could check the docs : https://langchain-ai.github.io/langgraph/cloud/reference/api/api_ref.html#tag/thread-runs

http://localhost:2024/threads/{thread_id}/runs/stream

And use
"if_not_exists": "create"

In the payload to have that thread created automatically

[muelerma]

Thank you. I had seen this part of the docs, however it wasn't clear to me that I have to use the endpoint to achieve consistency.