Perplexica icon indicating copy to clipboard operation
Perplexica copied to clipboard

balanced requests do not load

Open realies opened this issue 1 year ago • 10 comments

only speed requests work, nothing in the perplexica-frontend and perplexica-backend logs when attempting a balanced request, it just never ends up loading; how to debug?

realies avatar Jan 05 '25 12:01 realies

What embedding model are you using?

ItzCrazyKns avatar Jan 05 '25 12:01 ItzCrazyKns

nomic-embed-text:latest and mxbai-embed-large:latest

realies avatar Jan 05 '25 12:01 realies

I can confirm this

davidtavarez avatar Feb 19 '25 08:02 davidtavarez

Same here with Gemini embedding. Which model should work? Edit: It indeed does work with text-embedding-3-small from OpenAI

ottsch avatar Mar 28 '25 15:03 ottsch

I confirm this as well. I use BGE Small as embedding model.

mark-kazakov avatar Apr 27 '25 04:04 mark-kazakov

I confirm this as well. I use BGE Small as embedding model.

Strange enough, I am able to reproduce it in my docker environment, but the Balanced mode works just fine when running the dev environment with npm run dev.

mark-kazakov avatar Apr 27 '25 08:04 mark-kazakov

Can you all try using some different embedding model? I suspect the issue to be model related.

ItzCrazyKns avatar Apr 27 '25 09:04 ItzCrazyKns

I also have the problem. When I use the "balanced" mode in the http://localhost:3000/, it doesn't return the result with any the Hugging Face embedding models.

crash-zwt avatar Jul 13 '25 14:07 crash-zwt

Confirm the problem, using LMStudio and tried embedding models like bge small and qwen3-embedding-0.6b, not working in balanced mode.

summereasy avatar Sep 20 '25 02:09 summereasy

In conclusion: people having issues with balanced mode. My scenareo: downloading from the latest main branch. Was using Groq OSS-120b for inference with Google Embedding 001 in speed mode, it works well; when switching to balanced mode, it will not load. However when I switch to Google text embedding 004, balanced works well just like the speed mode.

grempire2 avatar Oct 15 '25 04:10 grempire2

I had luck with qwen3-embedding at first run, the second follow up run in the chat goes south. Is it possible that Ollama unloads the embedding model but the server assumes something something?

perplexica-1  |   code: 'ERR_BAD_REQUEST',
perplexica-1  |   config: [Object],
perplexica-1  |   request: [ClientRequest],
perplexica-1  |   response: [Object],
perplexica-1  |   status: 400,
perplexica-1  |   constructor: [Function],
perplexica-1  |   toJSON: [Function: toJSON]
perplexica-1  | }
perplexica-1  | An error occurred while getting documents from links:  Error [AxiosError]: Request failed with status code 400
perplexica-1  |     at eN (.next/server/chunks/162.js:5:245225)
perplexica-1  |     at IncomingMessage.<anonymous> (.next/server/chunks/162.js:7:9469)
perplexica-1  |     at tU.request (.next/server/chunks/162.js:7:21283)
perplexica-1  |     at async (.next/server/chunks/265.js:1:4561)
perplexica-1  |     at async m (.next/server/chunks/265.js:1:4448)
perplexica-1  |     at async e.func (.next/server/chunks/265.js:9:249)
perplexica-1  |     at async (.next/server/chunks/229.js:118:12849) {
perplexica-1  |   code: 'ERR_BAD_REQUEST',
perplexica-1  |   config: [Object],
perplexica-1  |   request: [ClientRequest],
perplexica-1  |   response: [Object],
perplexica-1  |   status: 400,
perplexica-1  |   constructor: [Function],
perplexica-1  |   toJSON: [Function: toJSON]
perplexica-1  | }
perplexica-1  | An error occurred while getting documents from links:  Error [AxiosError]: Request failed with status code 400
perplexica-1  |     at eN (.next/server/chunks/162.js:5:245225)
perplexica-1  |     at IncomingMessage.<anonymous> (.next/server/chunks/162.js:7:9469)
perplexica-1  |     at tU.request (.next/server/chunks/162.js:7:21283)
perplexica-1  |     at async (.next/server/chunks/265.js:1:4561)
perplexica-1  |     at async m (.next/server/chunks/265.js:1:4448)
perplexica-1  |     at async e.func (.next/server/chunks/265.js:9:249)
perplexica-1  |     at async (.next/server/chunks/229.js:118:12849) {
perplexica-1  |   code: 'ERR_BAD_REQUEST',
perplexica-1  |   config: [Object],
perplexica-1  |   request: [ClientRequest],
perplexica-1  |   response: [Object],
perplexica-1  |   status: 400,
perplexica-1  |   constructor: [Function],
perplexica-1  |   toJSON: [Function: toJSON]
perplexica-1  | }
perplexica-1  | TypeError: Cannot read properties of undefined (reading 'message')
perplexica-1  |     at e.invoke (.next/server/chunks/229.js:82:86517)
perplexica-1  |     at async (.next/server/chunks/265.js:11:87)
perplexica-1  |     at async e.func (.next/server/chunks/265.js:11:47)
perplexica-1  |     at async (.next/server/chunks/229.js:118:12849)
perplexica-1  |  ⨯ unhandledRejection:  TypeError: Cannot read properties of undefined (reading 'message')
perplexica-1  |     at e.invoke (.next/server/chunks/229.js:82:86517)
perplexica-1  |     at async (.next/server/chunks/265.js:11:87)
perplexica-1  |     at async e.func (.next/server/chunks/265.js:11:47)
perplexica-1  |     at async (.next/server/chunks/229.js:118:12849)

gerroon avatar Dec 09 '25 21:12 gerroon