AIR Blackbox Gateway — Test Suite Demo

🧪 Golden Fixture Tests (10 scenarios)

✅ happy_path PASS

GPT-4o-mini • 22 tokens • Standard chat completion

✅ tool_call_chain PASS

GPT-4o • 107 tokens • Multi-turn tool usage

✅ sensitive_payload PASS

GPT-4o-mini • 54 tokens • PII content detected

✅ huge_payload PASS

GPT-4o-mini • 5003 tokens • Large payload handling

✅ mixed_providers Anthropic

claude-3-5-sonnet-20241022 • 14 tokens

✅ deepseek_chat DeepSeek

deepseek-chat • 28 tokens

✅ grok_xai xAI

grok-2 • 18 tokens

✅ qwen_alibaba Alibaba

qwen-turbo • 16 tokens

✅ runaway_loop PASS

GPT-4o • 51 tokens • Loop detection

✅ malformed_request PASS

Error handling • Graceful failure

✅ NoPlaintextInAIRRecords

Verifies sensitive content is never stored in plaintext

✅ AllFixturesNoContentLeak (10)

All 10 fixture scenarios checked for content leaks

✅ AuthHeaderNotStored

API keys never persisted to AIR records

✅ ConcurrentRequests (10)

10 parallel requests with no race conditions

pkg/proxy 0.44s

22 tests • Golden fixtures, security, concurrency, auth

pkg/recorder 0.14s

3 tests • Write, load, directory creation

pkg/replay 0.27s

3 tests • Key extraction, content, token similarity

pkg/vault 0.54s

2 tests • Checksum verification, ref field resolution