RDRaju Dandigaminrajudandigam.hashnode.dev·1d ago · 5 min readHow to Build a Failure-Mining Loop for Google ADK EvaluationsAn agent evaluation set starts healthy. It contains the obvious intents, a few tool failures, and the happy paths used during development. Six months later, production has changed. New tools exist. Us00
RDRaju Dandigaminrajudandigam.hashnode.dev·2d ago · 5 min readHow to Use Google ADK Callbacks as an AI Agent Policy PlaneAn agent proposes refund_order. The tool exists, its arguments are valid JSON, and the model sounds confident. None of that answers the question that matters: is this action allowed now? Google's Agen00
RDRaju Dandigaminrajudandigam.hashnode.dev·3d ago · 7 min readHow to Validate Gemini Grounding With Evidence ContractsAdding citations to an AI answer feels like the moment the system becomes trustworthy. The response looks researched. Source links appear beside the text. The model is no longer answering only from it12AM
RDRaju Dandigaminrajudandigam.hashnode.dev·4d ago · 7 min readHow to Build Safe Fan-Out and Join Workflows With Google ADK“Let’s split it into agents” has become the AI equivalent of “let’s make it a microservice.” Sometimes the boundary is useful. Sometimes it only creates more state, more coordination, and a harder fai00
RDRaju Dandigaminrajudandigam.hashnode.dev·4d ago · 7 min readHow to Choose Between ADK Sessions, State, Memory, and ArtifactsThe first version of an agent usually has one place for context: the prompt. The user request goes in. Conversation history goes in. Tool results, preferences, account details, retrieved documents, an00
