Google Cloud engineers publish tutorial comparing agent memory options and context fixes
Start here
Google Cloud engineer Laxmi Harikumar and developer Miles Libbey published a tutorial comparing four memory options and three context fixes for Scribant, a TypeScript voice-to-action agent built with Google's Agent Development Kit (ADK). The memory comparison covers flat prompting, vector RAG, knowledge graphs and Memory Bank in the Gemini platform, tested with a user who moves from New York to San Francisco. The context fixes are a 40-word rolling window for the intent classifier, a rolling summary per session, and Memory Bank across sessions. The authors describe these as results from one demo, not benchmarks, and say the 40-word window should be treated as a starting point to tune to how users talk. Token-tracking code is in the repo for measuring other agents.
Generated from reports · updated 16 h ago
Report timeline
Follow the reports to see the event from different sides.
- Google AI · DEVGoogle Cloud engineers compare 4 agent memory options and 3 context fixes in an ADK voice agent
Google Cloud engineer Laxmi Harikumar and developer Miles Libbey demonstrate four memory options and three context fixes for Scribant, a TypeScript voice-to-action agent built with Agent Development Kit (ADK). The memory comparison covers flat prompting, vector RAG, knowledge graphs and Memory Bank in the Gemini platform, tested with a user who moves from New York to San Francisco; the context fixes are a 40-word rolling window for the intent classifier, a rolling summary per session and Memory Bank across sessions.
Heat for this event
- Current, comparable range
- 3
- Peak, comparable range
- 4November 19 12:00
- Change in 24 hours
- –
The trend compares the same participants over hours observed continuously, so its range can be narrower than the current heat count. Hover or tap the chart for hourly heat; the arrow keys move through it.