GEMINI LABJP
API — The Interactions API reaches general availability as the primary API for Gemini models and agents, with a stable schemaAGENTS — With GA, the Interactions API formally supports Managed Agents and background execution, with Gemini Omni coming soonLIVE — The Live API and AI Studio add real-time speech-to-speech translation, auto-detecting 70+ languages while preserving intonationOMNI — Gemini Omni Flash, a natively multimodal model, enters API public preview for building custom video workflowsSPARK — Gemini Spark, Google's personal agent, arrives on macOS to work across local files and Google Workspace on your behalfPRO — Gemini 3.5 Pro is reported to be delayed to July 17, centered on a 2M-token context and a Deep Think Reasoning LayerAPI — The Interactions API reaches general availability as the primary API for Gemini models and agents, with a stable schemaAGENTS — With GA, the Interactions API formally supports Managed Agents and background execution, with Gemini Omni coming soonLIVE — The Live API and AI Studio add real-time speech-to-speech translation, auto-detecting 70+ languages while preserving intonationOMNI — Gemini Omni Flash, a natively multimodal model, enters API public preview for building custom video workflowsSPARK — Gemini Spark, Google's personal agent, arrives on macOS to work across local files and Google Workspace on your behalfPRO — Gemini 3.5 Pro is reported to be delayed to July 17, centered on a 2M-token context and a Deep Think Reasoning Layer
TAG

Eval Harness

1 articles
Back to all tags
Related:
Google ADK1SessionService1Conversation Compaction1Gemini1Long-Term Memory1Python1
Gemini Advanced/2026-07-10Advanced

My ADK Assistant Quietly Forgot a Deadline — Catching Compaction Memory Loss With a Recall Probe

Compacting conversation history in Google ADK with Gemini lowers cost, but it also erodes what your assistant remembers — silently. Here is how I built a recall probe to measure that loss, compared three compaction strategies against the same ledger, and stopped trading memory for tokens.