List view

• Upgrade of autogen

  * leverage functions in agents (P0) * improve usability of tuning (P1) - OSS integration - more and clear use cases and documentation * better infra for agents (P1) - async - logging * continual learning agents (P1) - use interaction log to improve across sessions - config list - filter - prompt * few-shot example selection * LLM-based prompt revision - functions - summarize large amount of data - personalized chat - long chat * multi-agent collaboration (P1) - planning - goal checking * test case generation (P2) - adversarial example generation - test case using functions * support OSS LLMs (P2)

  No due date

  •17/26 issues closed
  65% complete9 open17 closed

• Release of a new version for an initial autogen.agent package

  * Use a filter to select a response * Use simulated/human feedback to improve in a session * Perform tasks with multiple steps * Leverage external information: the assistant agent should be able to leverage external information (if provided), e.g., a link to a pdf, to help with the given task * Allow customized response generation function or response generation configuration * Examples and use-case documentation

  Due by June 15, 2023

  •11/11 issues closed
  100% complete0 open11 closed