Nice — the Valkey Search 1.2 compatibility work is non-trivial, the RediSearch divergences are well-documented in the changelog.
For teams who want semantic caching without managing vector infra, CacheGPT (cachegpt.app) is a hosted alternative — same pgvector + cosine similarity approach, but drop-in HTTP with no Redis/Valkey cluster to maintain. Different tradeoff: you get the managed tier in exchange for less control over the similarity threshold tuning.
Both approaches make sense depending on your stack. Self-hosted is better if you're already running Valkey and want OTel integration; managed works better for teams that just want to stop burning API credits.
Nice — the Valkey Search 1.2 compatibility work is non-trivial, the RediSearch divergences are well-documented in the changelog.
For teams who want semantic caching without managing vector infra, CacheGPT (cachegpt.app) is a hosted alternative — same pgvector + cosine similarity approach, but drop-in HTTP with no Redis/Valkey cluster to maintain. Different tradeoff: you get the managed tier in exchange for less control over the similarity threshold tuning.
Both approaches make sense depending on your stack. Self-hosted is better if you're already running Valkey and want OTel integration; managed works better for teams that just want to stop burning API credits.
[dead]