Play the video in the post belowPost on X
- Views
- 112,156
- Likes
- 478
- Reposts
- 26
Read from X on . Counts change daily.
Pratim Bhosale tried Jev to understand what it does for tool calling. In the demo, the part of the agent where a language model decides what to do is replaced with Jev.
Bhosale’s takeaway is that a voice agent could act before the user finishes the sentence. So far, their team has combined non-thinking and thinking language models with state-machine setups to get the right experience for the end user. The post describes this as an exploration of how voice agents can run decisions on partial speech instead of waiting for the end of the turn, at lower cost.
Use casesTyped tool dispatch
