All Issues
24,993 verified issues
Late interaction embedding support
Hi, I wanted to ask a question: does pgvector currently support working with late interaction text embeddings, like the ones that come from a ColBERT model for example? This is an example of a vector that I would be referring to: [code block] Currently when inserting such a vector I seem to get `ValueError: expected ndim to be 1`, with this usage: [code block]
How to obtain a Google Identity Service (GIS) ID Token?
Dear team, I got an email telling me to migrate from the old Google Sign In library to the new Google Identity Services. I'm having a hard time with it. I posted this same question on Stack Overflow. In chat, a Google Developer Expert for Identity Platform recommended me to ask my question here. So here it is: Previously, I did (simplified for clarity): [code block] (I know what you're thinking. Why is this guy putting an id token in a variable named access token? It's because I didn't know any better at the time I was building this code. With Facebook's sign in lib, I get an access token, which I use to retrieve the user's email and name. I built Google sign in, thinking it would work the exact same way. I thought the id token was the access token, at the time of development. So bear with me, please.) Now, I'm trying (simplified for clarity): [code block] With the old google sign in library, I validated the access (id) token server side as such: `Payload payloadย =ย awaitย GoogleJsonWebSignature.ValidateAsync(accessToken);` This also returned the user's email and name in the payload. The access/id token I am getting back from GIS, is much shorter than the old one from GAPI. An online token debugger tells me it's not a valid JWT token. The ValidateAsync method throws an exception: `JWT must consist of Header, Payload, and Signature` No surprise, considering it's not a valid JWT token. I also tried the following call: `Payload payload = await JsonWebSignature.Ver
Support for Redis native TLS
I couldn't find an existing issue about this so asking here: Redis 6 (currently in rc phase) supports TLS natively. Details here: https://redis.io/topics/encryption I built Redis 6 with TLS support and created certs as instructed in Redis `TLS.md` file. I then tried to connect to it using ioredis: [code block] Should this work? My redis instance responds `17266:M 12 Mar 2020 15:12:14.455 Error accepting a client connection: error:140760FC:SSL routines:SSL23_GET_CLIENT_HELLO:unknown protocol`
Requests not Throwing Errors in NestJS App or AWS Lambda Node 20/18.x
Describe the bug Within NestJS, when performing calls such as `StripeClient.invoices.create` or other operations with bad data, requests hang indefinitely and don't throw errors. To Reproduce 1. Instantiate Stripe Client with `new Stripe(<API_KEY>)` 2. Wrap API call in `try/catch` block 3. Call `invoices.create` method with an invalid payload (Such as the one in the snippet) 4. Call will hang indefinitely neither erroring or succeeding Expected behavior When an API call fails on any non `2xx` status, it should throw an exception that can be caught by a `try/catch` block or a `.catch`. Code snippets [code block] OS Running on Docker `linux/amd64 node:20-alpine` Node version Node v20.18.0 Library version stripe node version 17.3.0 API version Default (Whatever you get when you don't specify any) Additional context Noticed running in a NestJS Application running in docker. I've tried: - Running requests through an Injectable Service that instantiates Stripe - Running the client directly in the Controller, or Service - Attempting to pass `httpClient` property in `new Stripe` with Axios and NestJS default HTTP Clients, to no avail. I also tried writing a script with the exact same code on the exact same version of node on a loose `.ts` file and running it with `ts-node` and it worked. I'll keep plowing through as it is critical for me to know when Stripe operations fail. If I find a solution, I will post it here.
Connection Pooling
I noticed you had that on the roadmap. I don't see any public work going on with it currently.. Have you put any thought in how to accomplish this yet or began work? I can at least help with spitballing ideas.
Non-ASCII tokens are corrupted sometimes when using the streaming API
Confirm this is a Node library issue and not an underlying OpenAI API issue - [x] This is an issue with the Node library Describe the bug When using the streaming API, sometimes tokens get corrupted. Characters get replaced by two or more `\uFFFD`. For example: [code block] when the token received is actually supposed to be `' ะธะทะฒะตััะฝะธ'`. The issue occurs because `LineDecoder` does not deal with multi-byte characters on chunk boundaries. Instead of using a separate `TextDecoder` instance per buffer, perhaps it should use a single `TextDecoderStream` for the entire stream. To Reproduce 1. Send a streaming completion request that will get non-ASCII tokens as a response. 2. Observe the output. With some probability, some of the tokens will be corrupted. Code snippets _No response_ OS Linux Node version Node v18.19.1 Library version openai v4.14.2
allow ressource filters/callbacks
as discussed on twitter, it would be very interesting to be able to hack proxied ressources before sending them back to the client. something like ressource filtering where we'll be able to change the response body/headers wonder how to integrate this into node http proxy
Neither apiKey nor config.authenticator provided
Describe the bug When installing either the version 17.0.0 or 17.1.0 in a brand new next.js 14 from scratch after initializing stripe [code block] I get the following error: Neither apiKey nor config.authenticator provided This issue is not seen after downgrading to 16.12.0 To Reproduce Create a next.js project adding stripe version 17.1.0 add a lib folder with a new file stripe.ts with the following: [code block] run the project getting this error: Neither apiKey nor config.authenticator provided Expected behavior No error should be seen in either version 17.1.0 and 17.0.0 process.env.STRIPE_SECRET_KEY is well-defined, and I can console.log it value. Code snippets _No response_ OS macos Node version v20.11.0 Library version v17.1.0 API version 2024-09-30.acacia Additional context _No response_
[BUG]: Unable to cast uuid to a field
What version of `drizzle-orm` are you using? 0.29.3 What version of `drizzle-kit` are you using? 0.20.14 Describe the Bug Hello, I'm new to drizzle but I have a error during migration, there is a problem with casting uuid, I expected uuid to be casted automatically in a field, I'm using PostgreSQL as my database of choice: [code block] schema.ts: [code block] Expected behavior _No response_ Environment & setup _No response_
Examples for streaming tools calls need fixing
Confirm this is a Node library issue and not an underlying OpenAI API issue - [X] This is an issue with the Node library Describe the bug I have used the `messageReducer` from the examples. After streaming the chunks finished, and I had the assembled message, I saw that the `arguments` property was always empty. I am testing the new GPT-4 model which supports multiple function calls. After hours of debugging, I finally realized that the problem was the reducer function, which didn't support arrays. This is how my reducer function currently looks, but I am not sure if that is correct. It would never add additional items from the delta: [code block] To Reproduce Use the new `gpt-4-1106-preview` model which returns a `tool_calls` array in combination with streaming, while accumulating the message using the reducer function from the examples. Code snippets _No response_ OS macOS Node version Node v20.7.0 Library version openai 4.17.4
Secure Website Problem
I am trying to create a simple reverse proxy. Firefox Browser -> Node Proxy -> Proxy 1 (LimeProxies.com) -> Target Firefox Browser -> Node Proxy -> Proxy 2 (LImeProxies.com, or some other provider) -> Target Firefox Browser -> Node Proxy -> Proxy 3 (LImeProxies.com, or some other provider) -> Target Very straight forward. The problem comes in with requesting a secure website, "https://bing.com" for example as compared to "http://bing.com", which has no problems. [code block] When I request a secure website, jp.process doesn't even fire. Through research, I found that server.on('connect', function(){}); was required. By adding this, I was able to see the request for a secure website "https://bing.com". [code block] I've tried both the ws and web methods, with no luck. It attempts to proxy the request, but it fails for some reason through a timeout without any helpful error message. I've tried dozens of proxies, even SOCKS proxies for the WS method, but I cannot get the request to process successfully. If I put something like this in the connect method, it works, but the request obviously comes from the nodejs server instead of a 3rd party proxy. [code block] I am just trying to create a simple reverse proxy, but have been unable to do so for weeks now. Any help is greatly appreciated. Thanks!
Only constructor options are passed to the `passes` functions
All of this options parsing that is found here is not being utilized. We are extending the global `this.options` but not passing them into any of the `passes`. I will see if the fix is as trivial as it seems otherwise I'll let you handle it @yawnt.
Default disableClientInfo: false causes connection failures with Redis servers < 7.2.x
Description Summary Starting from recent versions of ioredis, the default behaviour changed to send client information to Redis servers by setting `disableClientInfo: false` by default. This breaks compatibility with Redis servers older than version 7.2.x, as they don't support the `CLIENT SETINFO` command that ioredis attempts to use. Environment - ioredis version: 5.8.1 - Redis server version: < 5.0 - Node.js version: 20.17.x - Platform: linux Steps to Reproduce 1. Install the latest ioredis version 2. Connect to a Redis server with version < 7.2.x 3. Attempt to perform any Redis operation 4. Connection fails or commands are rejected Expected Behaviour The Redis client should successfully connect to and operate with Redis servers running versions before 7.2.x, maintaining backward compatibility. Actual Behaviour The connection fails because ioredis attempts to execute the `CLIENT SETINFO` command, which is only available in Redis 7.2.0+. This results in errors and broken functionality. Error Messages [code block] Root Cause The `disableClientInfo` option defaults to `false` in newer ioredis versions, causing the client to automatically send client metadata using `CLIENT SETINFO`, which was introduced in Redis 7.2.0. Older Redis servers don't recognise this command and reject it. Workaround Explicitly set `disableClientInfo: true` when creating the Redis client: [code block] Or pin ioredis to a version before this breaking change was introduced. Proposed S
why doesn't this use .pipe()?
this would be way simpler if we used pipe something like this: [code block] the code we currently have just duplicates functionality in pipe
Invalid types for nested fields
Some of the types exported by this package aren't valid. There's a mix of variable casing and it doesn't seem like results get normalized correctly For example in this program: [code block] [code block] `result` is of type `NumberInstance` and it sets `outboundCallPrices` but the actual type of that field says a property `currentPrice` should exist but it's never converted from `current_price` into `currentPrice`. `inboundCallPrice` can also be null, but that's not reflected in the field type: https://github.com/twilio/twilio-node/blob/7d9dd1fe9d2bc7e68f90a66e05785586ab20b270/src/rest/pricing/v2/voice/number.ts#L208 The source type descriptions in the openapi spec correctly mentions that both of these are nullable: https://github.com/twilio/twilio-oai/blob/bcf1d20ba0710f272748c3d23312498e28a99315/spec/yaml/twilio_pricing_v2.yaml#L304-L341
Support for HTTP CONNECT?
Does node-http-proxy support HTTP CONNECT requests, like: [code block] The proxy would then maintain the two connections, and forward the SSL traffic between the client and the target server. If so, how do I set this up? Thanks
Error when running example
The Proxy requests within another http server example (from README.md) raises the following error: [code block]
Unable to proxy multi-part form
Multiple attempts to proxy multi-part using both the browser and curl. Might be something really small but cannot seem to figure it out. The curl command sends the expect header and the browser does not. However, both do not proxy the request. Might be easier to see the first issue with the curl command and then narrow the issue down for the browser. Curl command curl -v --upload-file ./cam_espn.jpg http://127.0.0.1:3000/todkap%40us.ibm.com/cam_espn5.jpg Output todkapmacbookpro:bluemix_objectstorage_bug todd$ curl -v --upload-file ./cam_espn.jpg http://127.0.0.1:3000/todkap%40us.ibm.com/cam_espn5.jpg - Hostname was NOT found in DNS cache - Trying 127.0.0.1... - Connected to 127.0.0.1 (127.0.0.1) port 3000 (#0) > PUT /todkap%40us.ibm.com/cam_espn5.jpg HTTP/1.1 > User-Agent: curl/7.37.1 > Host: 127.0.0.1:3000 > Accept: _/_ > Content-Length: 197616 > Expect: 100-continue > < HTTP/1.1 100 Continue < HTTP/1.1 100 Continue - We are completely uploaded and fine < HTTP/1.1 404 Not Found < X-Powered-By: Express < SkipProxyWeb: true < X-Content-Type-Options: nosniff < Content-Type: text/html; charset=utf-8 < Content-Length: 46 < Date: Fri, 02 Jan 2015 14:55:39 GMT < Connection: keep-alive < Cannot PUT /todkap%40us.ibm.com/cam_espn5.jpg - Connection #0 to host 127.0.0.1 left intact Node script todkapmacbookpro:bluemix_objectstorage_bug todd$ cat app.js [code block] Server side log Logging inbound request PUT /todkap%40us.ibm.com/c
5.10.0: crashes at startup when bundled with esbuild
I think the new dual CJS/ESM export is causing an error. 5.9 worked fine, but bundling my node project with esbuild in 5.10 is causing this error. [code block]
Limit HNSW build's shared memory size for small tables
It occurs to me that it's silly that we allocate an area based on `maintenance_work_mem`, even if the table is small. For example, if maintenance_work_mem is set to 10 GB, and the table has 1000 rows, clearly we don't need to allocate 10 GB. There's some precedence for this in parallel VACUUM. It allocates an array to hold TIDs of dead tuples based on `maintenance_work_mem`, but it uses an upper limit on the worst case assumption that every page is full of tiny tuples, all deleted. See `dead_items_max_items` It'd be nice t do something similar in HNSW build.