TypeError: Invalid initialization vector

[alensiljak]

Getting the following error when typing a question (with local Ollama as the model, but it seems never to get to contact Ollama):

app-1            |  ⨯ TypeError: Invalid initialization vector
app-1            |     at c (.next/server/app/api/chat-rooms/[id]/messages/route.js:11:7689)
app-1            |     at rg (.next/server/app/api/chat-rooms/[id]/messages/route.js:6:12896) {
app-1            |   code: 'ERR_CRYPTO_INVALID_IV'
app-1            | }

[damajor]

Same here

[sfc-gh-pkommini]

Same issue.

[sfc-gh-pkommini]

@pu-re ? Do you know what might be happening?

I'm running on a macOS.

[neo4u]

@pu-re Is the project still active? Just curious if there's a recommended way to run it. Because it doesn't work on a mac with your current set of instructions.

[damajor]

It doesn't work on Linux either !!!!

[neo4u]

I'm happy to contribute a fix if you could point me in the right direction. Is it expecting some kind of initialization vector?

[neo4u]

@Macleykun @pu-re @wesleykirkland @GwonHyeok Can any of you help point me as what I can do to fix this? I'm happy to spend some time troubleshooting it but I do need a starting point and a direction to look into.

[Macleykun]

@Macleykun @pu-re @wesleykirkland @GwonHyeok Can any of you help point me as what I can do to fix this? I'm happy to spend some time troubleshooting it but I do need a starting point and a direction to look into.

Heya
Sorry I don’t use this project, I only helped with optimising the image and suggesting to make a public image (but bc of the encryption key requirement for the build, it wasn’t implemented)

[wesleykirkland]

@Macleykun I submitted an initial PR #78 to do a random encryption key upon each build which was merged to main and @GwonHyeok undid it the next day for good reason in 7aacfc3 and c9332be BTW. This allows for a public image.

@neo4u I do not have any suggestions for you, I can not get v0.2.0 or main to load. I can get a successful build and the schema builds the tables in SQL but no data is loaded. If you're getting this far I'd love to work together and see if maybe there's a path forward.

[alensiljak]

Guys, if you are really keen on getting something working why not either look at the source code and fix it or simply use another solution, like Open WebUI for Ollama? Constantly pinging people who volunteer their free time is not going to fix anything and may backfire.
What is great about this project is that it shows how easy it is to create a good medical assistant today.
From the technological standpoint, I think this may as well have been just a list of instructions how to get the solution by using existing tools.
Ollama + Open WebUI + a couple of models can do pretty much everything that is the functional goal of the project. It may not be as neatly packed but, if you're in a need of a quick solution then you can use that. Otherwise, the code is right here. Try not running in Docker and debug?

[neo4u]

@wesleykirkland Thank you for your response! This gives me a good starting point to start troubleshooting :)

@alensiljak All I wanted was a starting point so that I can better direct my efforts and I got that from @wesleykirkland's response. I'm also a volunteer who is willing to help but volunteers come in different shapes and sizes and not everyone has the same context to immediately start fixing things :)

[wesleykirkland]

@alensiljak Atleast for me I've been debugging this for weeks, last week I debugged it for 2 hours and built the database data by hand and still can't get anywhere. Unfortunately the project seems dead which is a shame. You can see one of my issues here #81 where I've outlined a lot of debugging. I've done further debugging unfortunately I'm not a Typescript developer and I'm personally in over my head and doing my best as are several people in this thread.

[alensiljak]

What I want to say is that it can be more effective to spend that (limited) time on building something using existing solutions. Recently I tried Open WebUI and it is a working UI for Ollama. There are also tutorials on how to use RAG with it, to feed it your documents.
For us, who are interested in this topic, it would be great if we could test and discuss different pieces and share that knowledge. LLMs, runners, MCPs, vectors and so on.
This project is great in that it initiated quite some interest. But there is no need to reinvent the wheel. Unless they want to create a commercial product.
This is a fairly fast-moving field.

[mfbloom]

I was able to get it to work by removing the encrypt and decrypt routines from the two route.ts files. It seems like they were trying to provide some basic encryption by wrapping the provider calls since there is no TLS setup by default in the application. However, I'd rather a working application locally... I'll probably submit a PR this week to "fix" it by removing those crypt functions.

[mfbloom]

Created PR #90