LangChain 404 at /responses.

If LangChain returns a 404 at POST /v1/responses, inspect the URL before rotating keys. A Chat Completions-only server will not recognize that separate endpoint. A custom base_url sets the host and API prefix; it does not guarantee which API path ChatOpenAI selects.

Read the request URL first

• POST …/v1/responses

  If the provider only documents Chat Completions, explicitly select that API with use_responses_api=False. Remove Responses-only settings and features.

• POST …/v1/chat/completions

  The API family matches a Chat Completions endpoint. Check the exact base URL prefix, model ID and key access.

• A duplicated /v1 or endpoint path

  Set base_url to the documented API prefix. Do not include /chat/completions when the client appends it.

Some 404 responses come from a wrong host, deployment path or API version. Confirm the outgoing method and full URL rather than inferring the cause from “Not Found” alone.

Choose Chat Completions explicitly

For a server that supports Chat Completions but not Responses, use LangChain's ChatOpenAI wrapper and set use_responses_api=False. Put the endpoint prefix, exact model ID and key in environment variables on the machine running your app:

import os
from langchain_openai import ChatOpenAI

chat = ChatOpenAI(
    model=os.environ["MODEL_ID"],
    base_url=os.environ["MODEL_API_BASE_URL"],
    api_key=os.environ["MODEL_API_KEY"],
    use_responses_api=False,
)
answer = chat.invoke("Reply with one short sentence.")
print(answer.content)

The option is available in langchain-openai 0.3.9 and later. If your installed version rejects it, check the package version and its matching API reference before changing the endpoint.

Why a custom URL may still reach /responses

LangChain documents that ChatOpenAI can route to Responses when Responses-specific features are used. A public reproduction against langchain-openai 1.4.1 also showed a model-name heuristic selecting /v1/responses on a custom Chat Completions-only server; setting use_responses_api=False sent that request to /v1/chat/completions. This is a version-bounded issue report, not evidence that every model name or current package version behaves the same way.

If you explicitly use output_version="responses/v1", built-in Responses tools or other Responses-only inputs, do not force the Chat API and expect those features to work. Choose an endpoint that implements the feature you need, or remove it for a Chat Completions request. Standard Chat Completions function tools are a separate API feature from OpenAI's built-in Responses tools.

Use the chat wrapper for a chat endpoint

ChatOpenAI is the LangChain chat-model wrapper. OpenAI documents the legacy text-completion interface; an OpenAI-compatible Chat Completions route expects chat messages and the /chat/completions path. For provider-specific response fields or features, LangChain advises using that provider's integration.

ChinaGenAI documents POST /v1/chat/completions; it does not document /v1/responses. Check current site-key availability and limits, then use the documented connection settings if they fit your task. This guide does not supply a key or test a live request.

Check current documentation and version

• LangChain ChatOpenAI: base URLs and Responses API selection
• LangChain OpenAI text-completion integration
• Public langchain-openai 1.4.1 reproduction and workaround (one reported case)
• Earlier ChatOpenAI 404 report for a Responses-only output setting

Documentation and issue reports checked 23 October 2026. This guide compares documented configuration and public reproductions; no provider account, API key or model request was used.

Free API access and limits

Check the site free API access and limits before trying to obtain a site key. Read the current status and request limits before using a site key. It is not a source of official provider keys.

ChinaGenAI is not affiliated with the providers named here. This release contains setup guidance and explicitly labeled fictional examples, not claimed live model benchmarks.