I got a notice from google that its embedding-001 model will be deprecated this month and to switch over to gemini-embedding-001 instead, which I’ve done. However, now I have a bunch of these errors in my Discourse log: Error ERROR: expected 768 dimensions, not 3072 querying embeddings for model …

[Bild] Lee_Ars: am I safe to just configure a new embedding setup wtih gemini that uses 3072-dimension vectors? This! Since we use Binary and Scalar Embedding Quantization for Significantly Faster & Cheaper Retrieval for our vector indexing, using vector with large dimensions is supported an…

Gotcha! I will run with 3072-dimension embeddings and stop caring. After having created a new embedding config and tested it, do I need to go into the rails console and nuke my existing embeddings and regenerate, or should I just leave Discourse alone and let it do its thing?

After you delete the old embeddings configuration we will eventually clean the database up in a background task.

Gemini Einbettungseinstellung übergibt output_dimensionality nicht?

Unterstützung

Falco (Falco) 8. Oktober 2025 um 14:38 5

Nachdem Sie die alte Einbettungskonfiguration gelöscht haben, werden wir die Datenbank in einer Hintergrundaufgabe bereinigen.

3 „Gefällt mir“

Thema		Antworten	Aufrufe
Gemini Embeddings Issue After Discourse Update to 3.6.0 Beta 2 Support ai	4	156	14. Oktober 2025
Gemini API Embedding Configuration Clarification Support ai	4	276	15. November 2025
OpenAI's text-embedding-ada-002 error "Embedding dimensions" Bug embedding , ai	4	267	25. Januar 2025
Discourse Gemini’s embedding-001 Call error Support ai	8	213	8. Februar 2026
Gemini Embeddings are not working Bug ai	4	174	2. Januar 2025

Gemini Einbettungseinstellung übergibt output_dimensionality nicht?

Verwandte Themen