Literally just had this idea 2 days ago - each TS agentic project feels like doing this stdlib work over and over again, but with a slightly differently architecture each time.
One question for our use case - what if a standard function requires bi-directional communication (as opposed to a single call to `generateText`)?
For one standard function, our envisioned interaction loop is:
Agent --> SSE to browser --> browser processes --> browser sends results back to server --> server sends to Agent --> Agent returns result
Comments
Literally just had this idea 2 days ago - each TS agentic project feels like doing this stdlib work over and over again, but with a slightly differently architecture each time.
One question for our use case - what if a standard function requires bi-directional communication (as opposed to a single call to `generateText`)?
For one standard function, our envisioned interaction loop is: Agent --> SSE to browser --> browser processes --> browser sends results back to server --> server sends to Agent --> Agent returns result