Truly disappointed @opencode !
Even after getting opencode go subscription just to experience faster Ox Alpha, but it worked slow as hell!
I am never gonna buy your subscription again! This is truly ridiculous
I had an important thing to do, and thatβs why I bought subscription only to experience this failure!
All thanks to you! Couldnβt submit on time!
What started as βhow fast can we make a grounded voice RAG system?β turned into Anvaya ποΈ
We (me and @sankalphs) experimented with retrieval, streaming STT, guardrails, extractive fast paths and model fallbacks to get here.
Live demo π
https://t.co/4isZGZ2twv
Still a lot to improve β but pretty happy with what we built :)
#RAGInGoa
Built Avvaya ποΈ β our (me and @sankalphs ) attempt at making voice intelligence actually fast, grounded, and useful.
Speech β retrieval β guardrails β the right answer, with a hybrid pipeline designed around latency + reliability.
Try it live π
https://t.co/4isZGZ2twv
Would love your feedback :)
#VoiceAI #RAG #AI #RAGInGoa
I started to think if Ox Alpha is actly composer 3, post trained on GLM 5.2!
Since many people saying itβs tokeniser matches with GLM modelsβ¦..
What do you thinkπ€ ?
@patloeber Broo stop patting your own shoulder π
Relative to other frontier models? Not yetπ
But yea keep on trying, dont lose hope, but please dont praise your own model when its not that good!