OpenAI-compatible AI gateway. One-line URL change → signed receipts + cost optimization on every call.
client = OpenAI(
api_key="sk-...",
base_url="https://gateway.veriops.io/v1",
)
| Vendor | Status | Endpoint |
|---|---|---|
| OpenAI | live | /v1/chat/completions (default) |
| Anthropic | stub | /v1/anthropic/messages |
| Mistral | stub | /v1/mistral/chat/completions |
X-VeriOps-Optimize — comma-separated list of opt-ins.
cache | L1 content-addressed cache (default on) |
receipt | Signed per-call receipt to Firestore (default on) |
routing | Auto-swap gpt-4o → gpt-4o-mini / gpt-5-mini |
nudge | Auto-append “Reply concisely” system instruction |
json_force | Auto-force response_format=json_object on JSON-asking prompts |
hdr | Hallucination Detection Receipt with logprob composite signal |
Performance-based: 20% of measured savings. Customer keeps 80%. No upfront, no platform fee on entry tier. Receipts are an add-on.
Set stream: true on the request body. The gateway forwards the SSE stream live (TTFB unchanged) and emits a Compliance receipt at the trailing usage chunk.
Vendor introspection endpoint: GET /v1/_meta/vendors
VeriOps is built by Clouthier Simulation Labs — veriops.io