Stateful Python & AgentKit UX
This release improves AI workflow capabilities by introducing a stateful Python interpreter in the SDK and Daemon.
This allows agents to maintain context across executions, enabling more complex data analysis and iterative tasks. Additionally, the documentation now features a new integration guide for using AgentKit by Inngest with Daytona.
To enhance usability, the dashboard's limits page has been redesigned to provide clearer insight into resource constraints.
Technical stability is improved with configurable runner start timeouts and optimized API performance regarding usage tracking.
The SDK also receives a fix to correctly preserve headers when encountering rate limit errors.
Release Details
Features
daemon,docs,sdk: stateful python interpreter (#2950)
dashboard: limits page redesign (#2856)
docs: AgentKit by Inngest integration (#3065)
sdk: add sandbox refresh activity method (#3022)
Fixes
api: cache lastUsedAt updates (#3060)
api: usage period query for close (#3059)
api: unlock sync runner snapshots cron (#3066)
docs: update data analysis example with agent iterations (#3049)
docs: clarify sandbox language feature (#3054)
sdk: preserve headers on rate limit errors (#3061)
Refactor
runner: make start timeouts configurable (#2928)