All Articles
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.
Setting a Token Budget Per Free User: Balancing AdMob Revenue Against AI Feature Cost
Rate limits protect requests per minute. They do nothing for the invoice that arrives at the end of the month. Here is how I derive a per-user token budget from ad revenue, keep the ledger inside a single call wrapper, degrade gracefully at a soft cap, and detect abuse with one concentration ratio.
When Your Knowledge Base Shifts Mid-Run: Pinning File Search to an Execution Epoch for Consistent Agent Grounding
When a File Search store is updated while a Managed Agent is running, a single execution can mix old and new grounding. Borrowing MVCC ideas, pinning an execution epoch keeps one agent run's evidence consistent. Here is the design and implementation.
One File Search Store for Many Apps: Splitting Retrieval With customMetadata and Chunk Config
Put several apps' FAQs in a single Gemini File Search store and metadataFilter can silently return empty grounding, or answers get split across chunk boundaries. Here is the customMetadata design, the AIP-160 filter-syntax trap, and measured chunkingConfig tuning.
Render Structured Output Field by Field as It Streams: Safe Partial JSON Parsing
With responseSchema streaming, the screen stays blank until the JSON closes. This walks through a partial parser that safely completes unclosed JSON, plus anti-flicker fencing that never lets a field move backward, and shows how time-to-first-field dropped from about 2.4s to 0.4s in practice.
Designing So the Next Shutdown Notice Doesn't Cost You an Afternoon: Isolating Gemini Behind a Single Port
The morning an image model shutdown notice landed, I couldn't say where my app touched that model. This is the design I use now: collapse Gemini dependencies into one port, with fallback and a CI deadline guard, shown as working code.
Being Able to Say Later Where a User's Data Was Processed — Region-Pinned Gemini (Vertex AI) Clients and a Residency Policy Gate
The default global endpoint is convenient, but it leaves you unable to explain where an EU user's data was processed. Here is a design built from three parts—a region-pinned Vertex AI client, a policy gate that forbids implicit global fallback, and a call ledger—with working code and measured latency.
Gemini 3 Multi-Tool Agents: Function Calling + Built-in Tools + Context Circulation in Production
A hands-on look at Gemini 3 multi-tool agents: combining Built-in Tools with Function Calling, Context Circulation, and parallel tool IDs, with measured latency numbers and the pitfalls I hit in production.
Your Tool Results Are Quietly Eating the Conversation — Handle Passing to Keep Gemini Function Calling Contexts Lean
Tool results linger in Function Calling history and compound your input tokens every turn. Two implementations — a token-budgeted compactor and handle passing — cut my measured input by roughly 8x, with the pitfalls I hit along the way.
Your Night Batch Is Causing the Morning 429s — Priority Admission Control for a Shared Gemini Quota
When bulk jobs and interactive features share one project's RPM/TPM, the bulk lane wins by default. A priority token bucket design with measurements: 429 rate 3.2% down to 0.03%.
When the Gemini Review Bot in Your CI Quietly Stops Earning Its Keep — Rebuilding Trust with Coverage and Actioned-Rate Metrics
A Gemini-powered PR review bot in GitHub Actions degrades without ever throwing an error. Field notes on catching diff truncation, model alias drift, and swallowed parse failures with one-line JSON logs and an actioned-rate metric.
After You Improve the Prompt, How Far Back Do You Regenerate? — Designing a Budget-Bounded Backfill
A prompt improvement only helps future output — thousands of old artifacts stay on the previous generation. This piece covers a budget-bounded backfill: selection scoring, edit-detection hashes, a pre-replacement gate, and a resumable cursor, with working code.