search

mem0ai / mem0

The Memory layer for your AI apps
https://mem0.ai
Apache License 2.0
22.85k stars 2.1k forks source link

when something goes wrong with creating the embeddings after adding a PDF, it doesn't give an error, but no embeddings are added/ #1422

Open jonabert opened 5 months ago

jonabert commented 5 months ago

šŸ› Describe the bug

I used a MTEB huggingface embedding model (Alibaba-NLP / gte-large-en-v1.5) I get this as an erratic situation, without an obvious error. INFO: Started server process [339843] INFO: Waiting for application startup. INFO: Application startup complete. ā—‹ Compiling / ... āœ“ Compiled / in 1629ms (637 modules) āœ“ Compiled in 138ms (313 modules) āœ“ Compiled /admin/data/add in 395ms (743 modules) (node:339731) [DEP0060] DeprecationWarning: The util._extend API is deprecated. Please use Object.assign() instead. (Use node --trace-deprecation ... to show where the warning was created) Failed to proxy http://127.0.0.1:8000/api/v1/admin/data_sources Error: socket hang up at Socket.socketCloseListener (node:_http_client:475:25) at Socket.emit (node:events:532:35) at TCP. (node:net:338:12) at TCP.callbackTrampoline (node:internal/async_hooks:130:17) { code: 'ECONNRESET' } Error: socket hang up at Socket.socketCloseListener (node:_http_client:475:25) at Socket.emit (node:events:532:35) at TCP. (node:net:338:12) at TCP.callbackTrampoline (node:internal/async_hooks:130:17) { code: 'ECONNRESET' } āœ“ Compiled /_error in 320ms (912 modules) Inserting batches in chromadb: 0%| | 0/21 [00:00<?, ?it/s] Inserting batches in chromadb: 0%| | 0/21 [00:00<?, ?it/s] Inserting batches in chromadb: 0%| | 0/21 [00:00<?, ?it/s] Inserting batches in chromadb: 0%| | 0/21 [00:00<?, ?it/s] Inserting batches in chromadb: 0%| | 0/21 [00:00<?, ?it/s] Inserting batches in chromadb: 0%| | 0/21 [00:00<?, ?it/s] Inserting batches in chromadb: 0%| | 0/8 [00:00<?, ?it/s] INFO: 127.0.0.1:37692 - "POST /api/v1/admin/data_sources HTTP/1.1" 201 Created

Dev-Khant commented 5 months ago

@jonabert, can you please check and verify whether you are getting answers related to context by querying the app or not?

jonabert commented 5 months ago

can you be more specific in what I shall test? I'm using the full-stack example

Dev-Khant commented 5 months ago

can you be more specific in what I shall test? I'm using the full-stack example

Sure, here can check that after adding your data, when you query your bot are you getting relevant answers? I mean answers around the context of your data source.