Failure‑Drill Your AI Platform Trial: Three Injections, One Scorecard

Description and use

A test protocol for assessing resilience of free AI tiers using a local proxy that injects latency, block, and quota failures.

Problem

Teams using free AI platform tiers often fail to detect and recover from latency, blocking, or quota‑exhaustion failures, leading to hangs or crashes during trials.

Outcome

Teams can quantify detection and recovery latency for free‑tier failures, enabling better resilience planning.

Workflow

• 1. Save the provided drill_proxy.py script.
• 2. Configure the proxy with target endpoint, mode (latency/block/quota), and quota limit if needed.
• 3. Point your AI agent to the proxy instead of the real endpoint.
• 4. Run each injection mode for 15 minutes, observing agent behavior.
• 5. Record how long it takes to notice the failure and how the agent recovers.
• 6. Use the results to adjust monitoring, retry logic, or fallback strategies.

Evidence

The drill simulates the three failure modes that kill free‑tier trials: latency, block, quota.

Source