Response Headers
CloudVera adds metadata headers to every response. Use these for debugging, monitoring, and cost tracking. All headers are exposed via CORS for browser-based applications.
| Header | Example Value | Description |
|---|---|---|
| X-Request-ID | req_a1b2c3d4 | Unique request ID — use this when contacting support or searching logs |
| X-Latency-Ms | 145 | Total gateway processing time in milliseconds (includes security scans) |
| X-Provider-Latency | 892 | Time spent waiting for the upstream LLM provider (ms) |
| X-Cache | HIT / SEMANTIC-HIT / MISS | Cache status — HIT = exact match, SEMANTIC-HIT = similar query matched |
| X-Cache-Similarity | 0.9542 | Semantic similarity score (0-1) when cache is a semantic hit |
| X-Routed-Model | gpt-4o-2024-08-06 | Actual model ID used (may differ from requested if routed or aliased) |
| X-Routed-Provider | openai | Provider that handled the request |
| X-Routing-Reason | Fallback: anthropic circuit open | Why this provider/model was chosen (only present when routing occurred) |
| X-Budget-Remaining | 42.58 | Remaining budget in USD for this virtual key's current period |
| X-Budget-Percent-Used | 57 | Percentage of budget consumed this period |
| X-Budget-Period | monthly | Budget reset period: daily, weekly, or monthly |
| X-Circuit-State | closed | Circuit breaker state for the provider: closed (healthy), half-open (testing), open (unavailable) |
| Retry-After | 45 | Seconds until rate limit resets (only present on 429 responses) |