File uploads default to using VectorDB strategy #6921
-
What happened?File uploads don't work without at least a Vector DB created. I'm running LibreChat latest on Railway and anytime I try to upload a file in chat, it fails. Images upload fine, however. I don't know if this is intentional or an oversight. Version InformationI'm running the image on Railway: librechat-dev 16aa5ed Steps to Reproduce
What browsers are you seeing the problem on?No response Relevant log output2025-04-16 19:34:35 error: Error uploading vectors No response received for POST http://host.docker.internal:8000/embed: connect ETIMEDOUT 192.168.0.99:8000
2025-04-16 19:34:35 error: [/files] Error processing file: connect ETIMEDOUT 192.168.0.99:8000ScreenshotsNo response Code of Conduct
|
Beta Was this translation helpful? Give feedback.
Replies: 4 comments 1 reply
-
|
Not a bug, this is the default behavior when you are uploading non-image files from a regular endpoint. Try using an agent, or enabled Code Interpreter from the chatarea, then you will have more file options
|
Beta Was this translation helpful? Give feedback.
-
|
Images work as they are not sent to the VectorDB and are instead uploaded to a folder and stationed there and reused as needed. Take a look at your yml for pgvector which includes all the info used to setup pgvector( the VectorDB). You don't need to create the VectorDB. You have a container connectivity issue. Until you solve that and provide error logs(not timeouts), this is most likely an issue with Docker or the type of host OS you're running Docker on and maybe the version of Docker as well. None of this was mentioned to help solve the issue you are having. |
Beta Was this translation helpful? Give feedback.
-
|
My setup is from the deploy to Railway button on GitHub. For whatever
reason, none of that stuff gets setup with this deployment
Jayson
…On Wed, Apr 16, 2025, 3:01 PM Jerry ***@***.***> wrote:
Images work as they are not sent to the VectorDB and are instead uploaded
to a folder and stationed there and reused as needed. Take a look at your
yml for pgvector which includes all the info used to setup pgvector( the
VectorDB). You don't need to create the VectorDB. You have a container
connectivity issue. Until you solve that and provide error logs(not
timeouts), this is most likely an issue with Docker or the type of host OS
you're running Docker on and maybe the version of Docker as well.
None of this was mentioned to help solve the issue you are having.
—
Reply to this email directly, view it on GitHub
<#6921 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ABRK56EFDUKWD5STMJKBFEL2Z2ZIBAVCNFSM6AAAAAB3JE7QPSVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTEOBVHEZTEMY>
.
You are receiving this because you authored the thread.Message ID:
***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
Oh good drop, Railway. Missed that one. Master deploying LC on your own before trying to fire it up on Railway. This is the only way you are going to truly be able to run LC however way you wish and wherever. Else, you'll drive youself in donuts getting nowhere. For starters, I recommend a 2GB DigitalOcean running Ubuntu 24.04/Docker/LC. Get to know each container that is part of the stack intimately. Test manually stopping the container running postgres VectorDB and watch what breaks and what the logs say. Its cool you went the Railway route but you really need to get familiar with the LC project. You may want to look at what it was when it started out in early 2023 to grasp a baseline from which the project started. Then see how it pivoted with changes and models evolved, functionality evolved. I won't lie to you on this, LC is a beast BUT extremely powerful software and worth spending all the time you need to master it. |
Beta Was this translation helpful? Give feedback.
-
|
Yeah, I do enough software engineering for my day job. I don't want to
spend time outside of work doing that. I spent enough time in the past
getting it to work with Heroku. I was really excited when the one click
deployment to Railway was introduced.
I just want it to work (which it does for the most part). I can live
without file uploads
Jayson
…On Wed, Apr 16, 2025, 4:12 PM Jerry ***@***.***> wrote:
Oh good drop, Raiway.
Master deploying LC on your own before trying to fire it up on Railway.
This is the only way you are going to truly be able to run LC however way
you wish and wherever. Else, you'll drive youself in donuts getting nowhere.
For starters, I recommend a 2GB DigitalOcean running Ubuntu
24.04/Docker/LC. Get to know each container that is part of the stack
intimately. Test manually stopping the container running postgres VectorDB
and watch what breaks and what the logs say.
Its cool you went the Railway route but you really need to get familiar
with the LC project. You may want to look at what it was when it started
out in early 2023 to grasp a baseline from which the project started. Then
see how it pivoted with changes and models evolved, functionality evolved.
I won't lie to you on this, LC is a beast BUT extremely powerful software
and worth spending all the time you need to master it.
—
Reply to this email directly, view it on GitHub
<#6921 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ABRK56BJ46WW32CTRA5POET2Z3BTPAVCNFSM6AAAAAB3JE7QPSVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTEOBVHE3TQNY>
.
You are receiving this because you authored the thread.Message ID:
***@***.***>
|
Beta Was this translation helpful? Give feedback.
Not a bug, this is the default behavior when you are uploading non-image files from a regular endpoint. Try using an agent, or enabled Code Interpreter from the chatarea, then you will have more file options