Persistent 404 Not Found for ALL Assistants API v1/beta Endpoints - Systemic Issue? (No API Logs)

API
1

Hello everyone,

I’m developing a Next.js App Router application that integrates with the OpenAI Assistants API (v2) for RAG (Retrieval Augmented Generation) and file management. I’m encountering a very persistent and unusual 404 Not Found error for all v1/beta endpoints related to Assistants and Vector Stores. I’ve exhausted all client-side and configuration troubleshooting, and my direct API logs show no activity.

I’m hoping someone from the community or OpenAI staff might have encountered this or can offer insights.

My Setup & Goal:

  • App: Next.js (App Router, v15.3.1), React, Supabase (Auth & DB).
  • Goal: Create Assistants, link/create Vector Stores, upload files to Vector Stores, and use these for RAG during chat.
  • API Key: sk-proj- type key, generated within my single OpenAI project ([YOUR_PROJECT_ID]).

Symptoms & Extensive Troubleshooting (What Works & What Fails Consistently):

A. General API Access & Core Assistant Functionality (Works Perfectly):

  • API Key Validity: curl -v https://api.openai.com/v1/models returns HTTP/2 200 OK with a full list of models.
  • File Upload (General Storage): My app successfully uploads files to POST https://api.openai.com/v1/files, returning HTTP/2 200 OK (e.g., file-Xmn5QwkAG43zYpHm1JABCf).
  • Assistant Creation: My app successfully creates Assistants via POST https://api.openai.com/v1/assistants, returning HTTP/2 200 OK.
  • Assistant Visibility: Newly created Assistants are visible and editable in the OpenAI UI dashboard.
  • Assistant Chat: I can successfully chat with these Assistants in the UI Playground and via my app’s API (threads API calls), and they respond as expected.
  • Manual RAG (UI Flow): I can manually upload files to an Assistant’s vector store via the OpenAI UI dashboard, and then the Assistant can successfully answer questions about those documents via my app’s chat interface.

B. Critical Failure Point: ALL v1/beta Assistants API Endpoints (Consistently 404 Not Found):

  • GET Assistant Details: curl -v https://api.openai.com/v1/beta/assistants/[YOUR_ASSISTANT_ID]
  • POST Attach File to Assistant (Directly): curl -v -X POST https://api.openai.com/v1/beta/assistants/[YOUR_ASSISTANT_ID]/files -d '{ "file_id": "[YOUR_FILE_ID]" }'
  • POST Create Vector Store: curl -v -X POST https://api.openai.com/v1/beta/vector_stores -d '{ "name": "Test Vector Store" }'
  • Result for ALL v1/beta attempts (from curl and my app): HTTP/2 404 Not Found with an Nginx HTML error page. No JSON error response.
    • Example cf-ray for a recent failing call: 947fa4f91e073c7f-LHR

C. API Logs Are Missing Entirely:

  • Despite successful calls (like /v1/models), absolutely no API requests appear in my API logs dashboard (https://platform.openai.com/logs?api=responses) whatsoever.

Troubleshooting Already Exhausted:

  • Ensured OpenAI-Beta: assistants=v2 header is sent.
  • Used fresh Assistants, new sk-proj- API keys, and even a completely new OpenAI project.
  • Confirmed Assistant IDs and API keys match the project context (and I only have one project).
  • Performed aggressive node_modules, package-lock.json, .next cleanup and rebuilds.
  • Verified Supabase client setup (client-side cookies, server-side createServerClient with JWT_SECRET), which is now fully functional.
  • Tried different Node.js versions (e.g., v20 LTS).
  • Checked for environment proxy/firewall (none found).

Conclusion & Request for Help:

The evidence strongly points to a systemic provisioning, routing, or feature-flag misconfiguration on OpenAI’s infrastructure, affecting the entire https://api.openai.com/v1/beta namespace for my account/project. Requests are consistently being met with a 404 Not Found at the Nginx layer, preventing them from even reaching your backend API system to be logged.

Has anyone encountered this specific issue where v1/beta endpoints consistently 404 despite working v1 endpoints, UI functionality, and missing API logs? Could this be an account-level provisioning issue? Any insights or suggestions would be greatly appreciated, as standard troubleshooting has yielded no resolution.

2

You seem to have inferred the wrong base URL for the Assistants endpoint, which is:

POST https://api.openai.com/v1/assistants)

Where a direct call with the correct setup parameters will return an object with your new assistant ID.

It can then be modified by a query string referencing the ID:

POST https://api.openai.com/v1/assistants/{assistant_id}

What you also do not show is sending headers, which besides needing your API key, also would need the beta: v2 header.

The API reference demonstrates such a call by CURL (although you’ll want to immediately transition to a programming language for the complexity needed):

curl "https://api.openai.com/v1/assistants" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "OpenAI-Beta: assistants=v2" \
  -d '{
    "instructions": "You are a personal math tutor. When asked a question, write and run Python code to answer the question.",
    "name": "Math Tutor",
    "tools": [{"type": "code_interpreter"}],
    "model": "gpt-4o"
  }'
3

Hi Jay,

Thanks for your quick reply and for providing that curl example! I appreciate you taking the time to help.

However, I think there might be a slight misunderstanding of the specific issue I’m facing. The POST https://api.openai.com/v1/assistants endpoint for creating a new Assistant, as you showed, is actually already working correctly for me. My app successfully creates Assistants, and they appear in my OpenAI dashboard and Supabase.

My problem is specifically with the v1/beta endpoints for the Assistants API, which all consistently return 404 Not Found from Nginx. These are the endpoints used for more advanced features like:

  • Getting Assistant details: GET https://api.openai.com/v1/beta/assistants/{assistant_id}
  • Attaching files to an Assistant’s knowledge base: POST https://api.openai.com/v1/beta/assistants/{assistant_id}/files
  • Creating Vector Stores: POST https://api.openai.com/v1/beta/vector_stores

For all these v1/beta calls, I am consistently sending the Authorization: Bearer [API_KEY] and OpenAI-Beta: assistants=v2 headers.

The 404 Not Found persists even with newly created Assistants, new API keys, and new projects. Crucially, I’m also seeing no API logs whatsoever in my platform.openai.com/logs dashboard for any calls (even successful ones like /v1/models).

My detailed curl outputs (which I can re-paste) confirm these v1/beta endpoints are returning 404 even when using the correct v2 headers.

Given that my basic Assistant creation and v1/models calls work, but all v1/beta calls fail with 404 and no logs appear, it seems to point to a systemic provisioning or routing issue on OpenAI’s backend for my account, rather than a basic API usage error.

Do you or anyone else have insights into why v1/beta endpoints might consistently return 404 Not Found and not appear in API logs for an account, despite other API calls working?

Thanks again for any help!

4

Who’s telling you that such a thing even exists?

Follow along in the API reference, see every URL for assistants and assistants threads at the top of methods:

https://platform.openai.com/docs/api-reference/assistants

2 Likes
6

Let me have a guess … :thinking: :robot: :wink:

7

Btw you should probably consider building on responses instead at this stage because assistants don’t have a long term future.

https://platform.openai.com/docs/api-reference/responses

1 Like
8

Thanks everyone. Yes, sorry looks like we were working off an older version of the info. Went via the File/Upload file + Vector Store route and have it working now.

Re Assistance long term future - yes I am aware but I understand OpenAI will be working on a defined upgrade path in time. Looks like we’ve got until second half of 2026 so will be keeping an eye on this.

Thanks again,
Richard

2 Likes