Failed to send compressed multipart ingest: Connection error caused failure to POST https://api.smith.langchain.com/runs/multipart in LangSmith API

[erwan-png]

Issue you'd like to raise.

I'm running into issues using Langsmith with LLM calls. It used to work a few months ago. I'm working with the Python Langchain and Langgraph, here's what I tried:

• working with and without async mode
• local (Ollama) and remote (Groq) LLMs
• with and without a checkpointer

Every time there's an '(a)invoke' called on an LLM, I get the following Langsmith error message:
Failed to send compressed multipart ingest: Connection error caused failure to POST https://api.smith.langchain.com/runs/multipart in LangSmith API. Please confirm your internet connection. ConnectionError(MaxRetryError("HTTPSConnectionPool(host='api.smith.langchain.com', port=443): Max retries exceeded with url: /runs/multipart (Caused by ProtocolError('Connection aborted.', RemoteDisconnected('Remote end closed connection without response')))"))
Content-Length: None
API Key: lsv2********************************************datrace= ....._

and the corresponding line in Langsmith dashboard remains in a "pending" state:

All previous nodes of my Langgraph app that do not invoke an LLM are correctly registered in the corresponding Langsmith tracing project dashboard :

The LLMs do generate a correct (non error) output, it's just the langsmith registration that is failing. All the settings above used to work a few months ago at least.

• Langchain version: 0.3.25
• Langsmith version: 0.3.42
• Langgraph version: 0.4.7
  Python version: 3.12.3

I'd also like to add that, after investigation, my firewall didn't block these multipart requests. I am not using a VPN.

As precise before, the agent execution proceeds normally, it is just langsmith registration that holds.
On the same settings, with the same API key, on the same computer, network, etc., but for a different project B (different langgraph agent and different venv), langsmith registration works fine. After checking, on this other project, langsmith version used was 0.2.11 while it is 0.3.42 on the "failing project A". After upgrading langsmith version to the latest, project B also fails with the same error.

Suggestion:

No response

[angus-langchain]

Hi @erwan-png,

Thanks for the detailed report. I'm not exactly sure what's going on here but eager to debug. Are you able to provide a minimal reproducible example that demonstrates this failure, or is it dependent on your internet? Additionally, can you try setting this environment variable to see if it has any effect on tracing. This would be the only major change between the two SDK versions you provided LANGSMITH_DISABLE_RUN_COMPRESSION

[erwan-png]

Hi @angus-langchain ,

Thanks for your very quick reply.

Below a reproducible example:

import os
from pathlib import Path
from langchain.chat_models import init_chat_model
from langgraph.graph import MessagesState
from typing import Any
from langgraph.graph import END, START, StateGraph

base_path = Path(__file__).parent.parent
path_file = base_path / "secret_groq.txt" # this file must exist
with open(path_file) as f:
    groq_api_key = f.readline()


path_file = base_path / "secret_langsmith.txt" # this file must exist
with open(path_file) as f:
    langsmith_api_key = f.readline()


llama = init_chat_model(
    "llama-3.3-70b-versatile", model_provider="groq", api_key=groq_api_key
)

class GraphState(MessagesState):
    i: int


def update_i(_: GraphState) -> dict[str, Any]:
    return {"i": 2}


def call_llm(state: GraphState) -> dict[str, Any]:
    return {"messages": llama.invoke("What is the color of the sky")}


workflow = (
    StateGraph(GraphState)
    .add_node("update_i", update_i)
    .add_node("call_llm", call_llm)
    .add_edge(START, "update_i")
    .add_edge("update_i", "call_llm")
    .add_edge("call_llm", END)
).compile()


os.environ["LANGSMITH_PROJECT"] = "mre"
os.environ["LANGSMITH_TRACING"] = "true"
os.environ["LANGSMITH_ENDPOINT"] = "https://api.smith.langchain.com"
os.environ["LANGSMITH_API_KEY"] = langsmith_api_key
# os.environ["LANGCHAIN_TRACING_V2"] = "true"
# os.environ["LANGCHAIN_API_KEY"] = langsmith_api_key
# os.environ["LANGCHAIN_PROJECT"] = "mre"

os.environ["LANGSMITH_DISABLE_RUN_COMPRESSION"] = "false" # set to "true" to make it work

for step in workflow.stream({}):
    print(step)

I used a file secret_langsmith.txt to set langsmith key, and secret_groq.txt to use LLMs from Groq.
If I set LANGSMITH_DISABLE_RUN_COMPRESSION to "true" as suggested, then the registration on langsmith works correctly. Otherwise it fails as described in previous messages. I couldn't find any information in the doc about this environment variable.

[angus-langchain]

Thanks @erwan-png, I'll take a look today. The environment variable isn't documented as its an internal optimization for batch ingestion of runs and generally should not be disabled.

[angus-langchain]

Hi @erwan-png,

I was unable to reproduce this issue with the code you provided. See public trace: https://smith.langchain.com/public/3e3da92a-1f11-44ac-bf10-744adf374fed/r

I assume the issue has to do with your own internet connectivity, as parts may be dropped during transport, resulting in the connectivity error and missing patch requests (which carry the llm output). Sounds like disabling run compression is working for you. I will close this issue for now but feel free to provide your workspace ID and I can debug further and report back.

[baswenneker]

I got the exact same issue when using "openai:o4-mini-2025-04-16":

2025-06-13 06:13:41,240 - langsmith.client - WARNING - Failed to multipart ingest runs: langsmith.utils.LangSmithError: Failed to POST https://api.smith.langchain.com/runs/multipart in LangSmith API. HTTPError('403 Client Error: Forbidden for url: https://api.smith.langchain.com/runs/multipart', '{"error":"Forbidden"}\n')

After switching to gpt-4.1-mini it disappears. Seems to be something with reasoning models?

[jewells77]

Got the same error
[WARNING] Failed to send compressed multipart ingest: langsmith.utils.LangSmithError: Failed to POST https://api.smith.langchain.com/runs/multipart in LangSmith API. HTTPError('422 Client Error: unknown for url: https://api.smith.langchain.com/runs/multipart', '{"error":"Unprocessable entity: invalid Content-Length or length parameter for post.b243c7c6-9cb7-42e0-b285-6e4b21e6632c.inputs: field size 583304247 exceeds maximum allowed size of 209715200 bytes"}\n')

[angus-langchain]

@jewells77 @baswenneker Feel free to open new issues as these are all different errors. One is complaining about a missing API key and the other is related to the maximum run inputs size.

[mahimairaja]

I got the same error today

[FurqanBhat]

For those who selected "West Europe" servers, the correct LANGCHAIN_ENDPOINT environment variable is "https://eu.api.smith.langchain.com" and not "https://eu.smith.langchain.com".

[joselpart]

For those who selected "West Europe" servers, the correct LANGCHAIN_ENDPOINT environment variable is "https://eu.api.smith.langchain.com" and not "https://eu.smith.langchain.com".

Thanks @FurqanBhat! It's like you wrote that for me! I was just setting up LangSmith and encountered this error.

[EndoTheDev]

ehm... hey hey i am sorry but i might have the same issue issue or similar...
what am i supposed to do? create a new issue?
its giving me...

2025-10-20 11:27:51,768 - langsmith.client - WARNING - Failed to multipart ingest runs: langsmith.utils.LangSmithError: Failed to POST https://api.smith.langchain.com/runs/multipart in LangSmith API. HTTPError('403 Client Error: Forbidden for url: https://api.smith.langchain.com/runs/multipart', '{"error":"Forbidden"}\n')
2025-10-20 11:27:53,230 - langsmith.client - WARNING - Failed to send compressed multipart ingest: langsmith.utils.LangSmithError: Failed to POST https://api.smith.langchain.com/runs/multipart in LangSmith API. HTTPError('403 Client Error: Forbidden for url: https://api.smith.langchain.com/runs/multipart', '{"error":"Forbidden"}\n')
2025-10-20 11:27:54,500 - langsmith.client - WARNING - Failed to send compressed multipart ingest: langsmith.utils.LangSmithError: Failed to POST https://api.smith.langchain.com/runs/multipart in LangSmith API. HTTPError('403 Client Error: Forbidden for url: https://api.smith.langchain.com/runs/multipart', '{"error":"Forbidden"}\n')
2025-10-20 11:27:56,348 - langsmith.client - WARNING - Failed to send compressed multipart ingest: langsmith.utils.LangSmithError: Failed to POST https://api.smith.langchain.com/runs/multipart in LangSmith API. HTTPError('403 Client Error: Forbidden for url: https://api.smith.langchain.com/runs/multipart', '{"error":"Forbidden"}\n')
2025-10-20 11:27:58,703 - langsmith.client - WARNING - Failed to send compressed multipart ingest: langsmith.utils.LangSmithError: Failed to POST https://api.smith.langchain.com/runs/multipart in LangSmith API. HTTPError('403 Client Error: Forbidden for url: https://api.smith.langchain.com/runs/multipart', '{"error":"Forbidden"}\n')

but... i am not from europe, but from asia (thailand)
now i wonder what this issue could be that i havent faced before yet. last time i worked with langsmith is a few month back already. i just started to add langchain/langgraph to this new project yesterday and wanted to observe and trace today. 🙈

[joselpart]

but... i am not from europe, but from asia (thailand)

@EndoTheDev, your actual location doesn't really matter. When you sign up to LangSmith, you are given the option to choose US Central and EU West for the data region. The endpoint depends on which one you chose when you signed up. That said, it may be a different problem if you were able to use it before.