Issue uploading files to vector store via uploadAndPoll

[EliasC]

Hi,

I’ve consistently run into an issue with uploading files to a vector store via uploadAndPoll:

try {
  openaiFile = await openai.vectorStores.files.uploadAndPoll(
    target.vectorStoreId,
    file
  )
} catch (err) {
  console.log(err)
}
  

I get this error message:

[Error: 404 No file found with id 'file-RXt3Me3fAzy3ZEj9xy6ZLw' in vector store 'vs_68fe949dc6008191ade830d6f9e5935c'.]

Some extra error details:
request_id: 'req_0a41b55322de4f5fbf4b7305b808ede6', error: { message: 'No file found with id \'file-RXt3Me3fAzy3ZEj9xy6ZLw\' in vector store \'vs_68fe949dc6008191ade830d6f9e5935c\'.', type: 'invalid_request_error', param: null, code: null }, code: null, param: null, type: 'invalid_request_error' }

However, when I check the vector store via the OpenAI API dashboard, the file exists and is successfully attached to the vector store.

I tried an alternate method of uploading a file to the vector store by doing this:

const uploaded = await openai.files.create({ file, purpose: ‘assistants’ });

const attached = await openai.vectorStores.files.create(
  targetCase.vectorStoreId,
  { file_id: uploaded.id }
);

const ready = await openai.vectorStores.files.poll(
  targetCase.vectorStoreId,
  attached.id
);

and everything seemed to have worked except the poll.

Anyone else experiencing this?

[_j]

I can confirm, independent of the polling method the API SDK offers, that the vector store service is broken…again.

This script that is basically a “guaranteed to run” with its own newly-created vector store (and attaching itself as a file name):

Vector_stores.files.list() should also return the file name

Yes. Store the file name yourself as a metadata attribute of the vector store file. Demonstration that will run on your filename to upload: from openai import OpenAI; cl = OpenAI() filename = "vs_metadata_demo.py" vs_id = cl.vector_stores.create(name="vstest").id def upload_attach_poll(vs, file, suffix=""): file_id = cl.files.create( file=open(file, "rb"), purpose="user_data", expires_after={"anchor": "created_at", "seconds": 3600}, ).id vs_file_id = cl.vector_sto…

openai.NotFoundError: Error code: 404 - {'error': {'message': "No file found with id 'file-DcBL8Kq6ZcRnGdnyDFYirr' in vector store 'vs_68fedd528f60819191cd84d21a5a2a59'.", 'type': 'invalid_request_error', 'param': None, 'code': None}}

I had to add a 5 second sleep before vector_stores.files.retrieve() for success. The API SDK that checks immediately for polling will obviously not be adaptable.

[Dani_Cajigal]

We’re experiencing the same issue. We’re also seeing that the GET endpoint
https://api.openai.com/v1/vector_stores/{vector_store_id}/files/{file_id}/content
is failing and consistently returning “Not found”. Meanwhile, the search endpoint is still working.

[kexin1]

Error creating files: Error code: 404 - {'error': {'message': "No file found with id 'file-Uzx4KBALYwxHUwbLV8hoGJ' in vector store 'vs_68ff70740380819199a336c074652ac9'.", 'type': 'invalid_request_error', 'param': None, 'code': None}}

same pb

[nurivolkancekuc]

Demonstration that will run on your filename to upload:from openai import OpenAI; cl = OpenAI() filename = “vs_meradate_demo.py” vs_id = cl.vector_stores.create(name=”vstest”).id def

upload_attach_poll(vs, file, suffix=””): file_id = cl.files.create(file=open(file, “rb”), purpose=”user_data”, expires_after= {“anchor”:”created_at”,”seconds”: 3600},).id vs_file_id = cl.vector_stornhust.

Yes. Store the file name yourself as a metadata attribute of the vector store file. Demonstration that will run on your filename to upload.

[Roilan_Roman]

Same issue here @OpenAI_Support

[AndiL]

We are looking into this! For now the workaround is to add the 5 second sleep

[EricZim]

@AndiL A note on this issue….I think a side affect of this is that the initial query of a vector store after adding files comes back with a file count of ZERO and a status of “completed” even though the vector creation process returned a file count of X (in my case 1) and “in_progress”.

If I poll the vector store until I actually get a file count, the status pops back to “in_progress” and I continue to pull until it now CORRECTLY shows a status of “completed”.

I would expect the initial retrieval of the vector store to have a correct file count and correct status of “in_progress”.

Would be great if this issue is fixed, but that is a workaround instead of waiting a random time and hoping the file is available.

[mohanlalranvir]

Also experiencing this issue

[Sunil_Chopra]

i think we are also experiencing this issue

[rahsuri]

Just noticed this issue as well.

[mohanlalranvir]

I added a 10s interval on the poll as below - still get the same issue.

r = openai_client.vector_stores.files.upload_and_poll(
    vector_store_id="vs-xxxx",
    file=file,
    poll_interval_ms=10000
)

[David_Hempy]

We resolved this by adding a short delay between create and poll. This seemed to fix it, but out of paranoia we ALSO added a retry mechanism to give create as much time as it needs to resolve.

This solved our problem, for the first time since it mysteriously appeared a few days ago.

const createFileStore = async (file) => {
  const store = await openAIAgent.beta.vectorStores.create({
    name: 'Elevator speech outline',
    expires_after: { anchor: 'last_active_at', days: 1 },
  });

  const uploadedFile = await openAIAgent.files.create({
    file,
    purpose: 'assistants',
  });

  const attached = await openAIAgent.beta.vectorStores.files.create(
    store.id,
    { file_id: uploadedFile.id },
  );

  await new Promise(resolve => setTimeout(resolve, 5000));

  const pollWithRetries = async (minTimeout = 3 * 1000, maxTimeout = 60 * 1000, retries = 5) => {
    const operation = retry.operation({ retries, factor: 2, minTimeout, maxTimeout });

    return new Promise((resolve, reject) => {
      operation.attempt(async function(currentAttempt) {
        try {
          await openAIAgent.beta.vectorStores.files.poll(store.id, attached.id);
          resolve();
        } catch (e) {
          if (operation.retry(e)) {
            logger.error(e);
            logger.info(`Attempt ${currentAttempt} failed. Retrying...`);
          } else {
            logger.error(e);
            logger.info(`Attempt ${currentAttempt} failed. Giving up.`);
            reject(operation.mainError());
          }
        }
      });
    });
  };

  await pollWithRetries();

  return store;
};

[AndiL]

upload_and_poll should work now!

We are also aware of the file count @EricZim mentioned, but we don’t have an eta for the fix for that

[mohanlalranvir]

I tested this now and it’s still intermittently returning the same error.

[_j]

One can imagine the scenario happening here is a delay in database propagation in making the URL path of the network request for object retrieval by vector store container file id ready. It hits:

GET https://api.openai.com/v1/vector_stores/{vector_store_id}/files/{file_id}

A status 404 should be an expected or anticipated part of the lifecycle, along with knowing the futility of immediate polling.

Quite frankly, the SDK polling method is not in an appropriate place in code. Taking only an ID and a vector store, it doesn’t have knowledge about the size of the upload and expected latency of document extraction and embeddings of the chunks, nor does the method offer such intelligence parameters to be passed.

Giving OpenAI the benefit, they don’t describe using their simple polling function in file search documentation. The steps for file search are upload, add/attach, and check. However, the idea of listing all files with .vector_stores.files.list() as “check” for a single file attachment there is also poor, because the list method cannot return anything over 10000 files nor has a limit, and polling a huge list per-file would be silly.

However the “retrieval” (semantic search endpoint) documentation is indeed offering .vector_stores.files.upload_and_poll() for the same slot, now also proven poor. The name is also terrible, as the method doesn’t upload anything over the network.

The method that should be used is .vectorStores.files.retrieve()

The polling should have application-level consideration:

• tolerate initial 404
• have a file-size minimum delay expectation
• have delay tuning by file type (such as PDF being more complex, but perhaps less tokens extracted)
• have an aggressive polling during the expectation for user experience
• have a back-off after failing to meet expectation
• have a failure state timeout also extrapolated from the file type and size.

This post, from this issue going on a week, I whipped up a mod of my “real uploader” to tolerate an initial 404. You can port and add some brains:

API Problem with file processing, file uploaded but not found in vector store

This is not a guess. This is a solution. The behavior and performance of the vector store endpoint has changed, and the URL containing a file path is not immediately ready. Thus, The API SDK immediately polls and fails. Mine won’t: '''error-tolerant file uploader and vector store attachment''' from openai import OpenAI; cl = OpenAI() from time import sleep from openai import NotFoundError def upload_attach_poll(vs: str, file: str, suffix: str = "") -> str: # step 1: upload the file to O…

Better would be:

• No OpenAI SDK bloat for simple API methods
• No OpenAI vector stores if you don’t want a pattern of downtimes and the cheapest “small” embeddings now
• No file search tool if you don’t want injections saying “user uploaded files”
• No Responses API