Skip to main content

Release 7.10.0 - ConvyAI

What's new​

  • Agentic Orchestrator Dashboard - A new Dashboard page shows how many executions are in progress or waiting and for how long, and analyzes the executions ended in a period compared with the previous one: executions, completion rate, duration, tokens per execution, top flows, executions over time and errors by node type. Every figure opens Monitoring already filtered. See Dashboard for details.

  • Flow Report - A new page, reached from the Dashboard, analyzes a single flow with the same period and comparison rules. It adds the technical outcome, the results emitted by Result nodes and the nodes where errors concentrate. See Flow Report for details.

  • Executed nodes panel with CSV export - In live monitoring and replay, a new panel lists every node execution in chronological order, with path, edge, status, timing and AI Agent tokens. The table can be exported to CSV. See Debug Executions for details.

  • Thinking Level for Gemini - A new Thinking Level field in the advanced parameters of Gemini AI Models configures reasoning on Gemini 3.5 and later. You configure it once on the AI Model and it applies to the AI Agent nodes of FlowDesigner and Agentic Designer. See Google Gemini for details.

  • Thinking Budget for Xference - A new Thinking Budget field in the advanced parameters of Xference AI Models configures the reasoning of the model. An empty field uses the model default, 0 turns reasoning off and any other value sets the token budget. You configure it once on the AI Model and it applies to Knowledge Base nodes and AI Agent nodes of FlowDesigner and Agentic Designer. The reasoning never appears in the response returned to the end user. The reasoning tokens actually consumed are not reported and whether the budget is respected depends on the provider. See Xference for details.

Improvements​

  • Monitoring filters - The Monitoring page now offers combinable filters on time range, flow, status and result value, applied with Save filters and reset with Clear filters. The Period field is preselected on Last 7 days. See Monitoring for details.

  • Filter by end date in Monitoring - A new Filter period by selector applies the time range to the start or to the end date of executions, so you can list the same executions counted by the Dashboard. See Monitoring for details.

  • Provider errors for Gemini and Xference - The AI Model test console shows the errors returned by the Gemini and Xference providers without filtering, using the same parameters the flow nodes send at runtime.

  • Optional parameters on Xference - Max token completion, Temperature, Top P, Frequency penalty and Presence penalty start empty on new Xference AI Models and follow the rule of the other providers: an empty field is not sent, any value is sent as is. Xference AI Models already configured keep their saved values; review Max token completion, which is no longer prefilled with 1000.

Breaking changes​

⚠️ Breaking change: Gemini 2.5 models are no longer supported. AI Models configured on a Gemini 2.5 model stop working and the test console shows the error returned by the provider. Enter a 3.5 or later model in Completion model and, if needed, set Thinking Level. AI Models already on 3.5 or later keep working without reconfiguration.

⚠️ Breaking change: Thinking budget, Temperature, Top P, Frequency penalty and Presence penalty are removed from the configuration of Gemini AI Models, because Gemini 3.5 and later models do not support them. These values are no longer sent to the provider, even if present in AI Models already saved. In new Gemini AI Models, Max token completion is empty by default and Completion model defaults to gemini-3.8-flash.