When Gemini API Leaks Japanese Into Your English Output Once in a While — Field Notes on Measuring the Contamination Rate and Tightening It in Stages
You told Gemini to answer in English, and 3 out of 100 runs slip a Japanese sentence into the tail. Here is why you cannot stop that 'once in a while', and a production pattern that measures the contamination rate as an SLO and tightens it with graded recovery, with working code.
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%.
After the One-Click Deploy — Hardening an AI Studio Gemini App on Cloud Run for Real Production Use
AI Studio's one-click deploy to Cloud Run gives you a working URL in minutes — but not a production service. A practical checklist for API key storage, authentication, cost ceilings, and observability, with copy-paste gcloud commands.
Routing Between Local Gemma 4 and the Gemini API Cut My Bill from ¥32,000 to ¥9,000 — A Production Hybrid Router Design
How I cut a ¥32,000/month Gemini API bill to the ¥9,000 range with hybrid inference: routing design, a full Python router, production pitfalls, and how Gemma 4 arriving on the Gemini API in July 2026 changes the decision.
Tracing Which Prompt Revision Moved Your Quality — Prompt Versioning for a Gemini Pipeline
Editing prompts in place erases the trail: when quality shifts you can't tell whether the model moved or your wording did. Here's a small system that pins prompts by content hash, stamps every generation with the model ID and revision, and bisects a quality drop down to the exact revision boundary, with copy-paste Python.
A Finished Gemini Job Flipped Back to 'Running' — Stopping Out-of-Order Webhooks with Monotonic State Apply
When you receive Gemini long-running operations over webhooks, a stale 'running' event can arrive after completion and roll your state backward. Here is a monotonic-apply reducer that safely drops regressing updates.
When Gemini × Qdrant Hybrid Search Was Quietly Losing Recall — Field Notes on Instrumenting RRF Weights and Sparse-Vector Drift
Run Gemini embeddings with Qdrant hybrid search in production and your dashboards stay green while recall quietly slips. These field notes show how to catch it with measurement — RRF weights, sparse-vector drift, missing payload indexes — and protect it with a quality budget.
Mixing Text and Images in One File Search Skewed My Results Toward Images — Rebalancing by Modality After Retrieval
When you put text and images in a single File Search store with gemini-embedding-2, results can quietly skew toward one modality. Here is how to measure that skew and even it out after retrieval, using per-modality normalization and quota-based merging — with working code.
A Promotion Gate So gemini-flash-latest Flipping to 3.5 Flash Doesn't Break Your Pipeline at 3 AM
Floating aliases like gemini-flash-latest swap their target on every GA, quietly shifting the assumptions your unattended automation depends on. Here is a role-to-pinned-ID indirection layer, an acceptance harness that measures four metrics against your own golden set, and threshold-driven promotion and automatic rollback — with working code.
Your Gemini Completion Event Will Arrive Twice — An Idempotent Sink That Makes Webhook + Reconciliation Effectively Once-Only
Once you receive Gemini long-running operations over a Webhook and back it up with a reconciliation poller, the same completion arrives twice and publishing or billing runs twice. Build an idempotent sink with a normalized key and a claim-run-commit pattern that keeps side effects effectively once-only.
Don't Retry Every Gemini 429 — Telling Rate Limits Apart From Spend Cap Exhaustion
A 429 RESOURCE_EXHAUSTED can mean 'wait a second and it clears' or 'you're out of budget for the month.' Now that Project Spend Caps is generally available, the second case is real in production. Here's how to classify the two and build a retry layer plus a circuit breaker around them.
When Gemini Computer Use Acts on a Stale Screen and Fails Quietly — Field Notes on Guarding the Loop
A Computer Use agent will click based on a screenshot taken moments ago, miss the real target, and throw no error. These are field notes on measuring those silent misclicks and stopping them with an observe-act-verify loop.