API Reliability, Console Navigation, Tools Enhancements & Prompt Versioning - August Update
This release improves platform reliability and usability across several modules. In API Manager, the agent_initiated flag now saves correctly, and data source timestamps update as expected after content changes. The Console streamlines navigation by merging resource channels and integrations, and voice sample fetching for Google voices is enhanced for accuracy. Tools now support precise outbound campaign rate limiting across multiple active batches. Insights LLM evaluations enforce strict JSON schema compliance, reducing errors. Backend support is added for new Salesforce tools, enabling lead management. Additionally, prompt version history is introduced, allowing users to track and manage prompt changes efficiently.
v25.8.14
Aug 14, 2025
API Manager
Fixed a bug where the agent_initiated flag on agent create request did not propagate to the database, hence, the value was always FALSE.
Fixed a defect where a data source's last-modified timestamp was not getting updated when the content was altered.
Console
Combined console resource channels and integrations into one navigation item.
Fixed an issue when fetching Google voices by disabling pitch setting on Google chirp voices when requesting voice samples.
Tools
Outbounds - Fixed an issue with outbound campaigns to accurately support rate limiting with multiple active batches–correctly divide the campaign’s hourly rate by the number of active batches which have pending outbound requests.
Improved Insights LLM evaluations that use OpenAI models by strictly enforcing JSON schema for function calls and formatted results, ensuring guaranteed adherence to given schemas (e.g., enum lists); while unsupported JSON schema keywords will now be rejected during initial configuration to prevent runtime errors.
Added backend integration and support for new Salesforce tools: get_salesforce_leads and create_salesforce_lead.
Added version history to prompts.