-
Notifications
You must be signed in to change notification settings - Fork 2.2k
Closed
Labels
core[Component] This issue is related to the core interface and implementation[Component] This issue is related to the core interface and implementation
Milestone
Description
Describe the bug
Static Instructions won't work for an entry point agent who has a pipeline of agents as a sub_agent. We have an entry team leader agent that receives the user input and transfers the job to a complicated pipeline of agents. I got to set the App() at the end of this agent in order set the ContextCaching, but no luck getting static_instructions working there.
Expected behavior
I would expect that my entry point agent could have a long monolithic static instruction that could be passed and used as a collective memory to drive the rest of the team's behavior. I used the feature tho, but in a loop_agent with a critic pattern that I have later in the flow.
- Are you using LiteLLM: Yes (I'm using OpenAI through LiteLLM because it's the company model standard)
- Which model is being used: gpt-5 and gpt-4.1-nano
Metadata
Metadata
Assignees
Labels
core[Component] This issue is related to the core interface and implementation[Component] This issue is related to the core interface and implementation