Cryptographically private AI inference (Arcium-encrypted, x402-paywalled)

OpenAI-compatible chat completions. The prompt MUST be Arcium-encrypted client-side using `@solrouter/sdk`'s `encryptPrompt()` helper. The Solrouter backend relays the ciphertext without reading it; the enclave decrypts it and calls the model's GPU: an H200 GPU enclave for Qwen3.5 122B (its host reads nothing), or a Nosana GPU node for the smaller models, which sees the prompt and reply in plaintext during inference. The response returns encrypted. Pay-per-call x402 ($0.02 USDC flat per call). A signed-in caller with an active plan spends one plan call instead.

curl -X POST "https://example.com/api/v1/x402/chat/completions" \  -H "Content-Type: application/json" \  -d '{    "encryptedPrompt": "base64-arcium-ciphertext...",    "model": "gpt-oss:20b",    "maxTokens": 2048  }'

{
  "encryptedResponse": "base64-arcium-ciphertext...",
  "usage": {
    "prompt_tokens": 18,
    "completion_tokens": 124,
    "total_tokens": 142
  },
  "cost": 0.005
}

{
  "error": "missing_required_fields"
}