well done to @Henrique_Cruz1 and his team.
this is EXACTLY what i have been looking for.
i have spent the day testing this feature across all our coda ai agents, from the simplest to the most complicated.
my Agentix mechanism uses JSON objects as the ‘memory’ that coordinates all the agentic workflows.
so getting the coda ai to generate correct and precise JSON is vital to its effeciency.
prior to this i had to implement a range of special trickery to guarantee this, even repeating some prompts if the response was not proper JSON.
today i tested almost 1000 test cases on many different agent types, and the success rate was almost perfect.
(the few test cases that failed were due to badly formed prompts that need to be corrected.)
so this is a VERY SIGNIFICANT improvement to the superdocs ai capabilities.
you can read about the Agentix Framework HERE
Max