I use full length vectors (512 dimensions) and have seen very fast lookups with pgvector (HNSW index) and sqlite-vec on 20k vectors. I think any decent vector database should be able to handle 44k entries… which one are you using now?
Thanks for your interest. The rerankers are external, GoodMem is a unified API layer that calls out to various providers. There's no model running inside the database or the GoodMem server.
We support both commercial APIs and self-hosted options:
- Cohere (rerank-english-v3.0, etc.)
- Voyage AI (rerank-2.5)
- Jina AI (jina-reranker-v3)
Thanks! Yeah embedding is simple enough and my needs were small enough that I didn’t want to pay. Both llama-server and ollama are great options, and if container size isn’t an issue you get a greater variety running what you want with sentence transformers.
Aren’t most inventions anti-labor by this rationale? We employ fewer farmers than we used to, but this is good for society because people who would be farmers are now working on other things.
> 5 - 9 means strong effects, definitely not placebo.
It’s impossible for anyone to say this convincingly about their own experience. If it were easy to tell whether an effect was due to placebo, we wouldn’t need blinded trials!
That's because it's obvious due to effects other than the one you're trying to observe. Which is of course the case when you're dealing with psychedelics (and of course many other drugs).
We certainly shouldn't accept subjective evaluations as proof of effectiveness, but that does not give it zero value. The more subtle the effect or the more invested the subject is in establishing the effectiveness of an intervention (or any of many other confounders) the less likely it is to represent proof.
But it is evidence. Think of this as more observational science rather than experimental science; we have to do some work to determine whether it is worth trying to do blinded experiments to validate an effective, and this is that work.
