moonshotai
1 endpoint
novita
1 endpoint
venice
1 endpoint
moonshotai/kimi-k2.5Create a key inDashboardAPI Keys, then set it as AI_STATS_API_KEY in your environment variables.
Add "stream": true to receive token events.
# 1) Set your key
export AI_STATS_API_KEY="sk-live-***"
# 2) Send a request
curl -s https://api.phaseo.app/v1/responses \
-H "Authorization: Bearer $AI_STATS_API_KEY" \
-H "Content-Type: application/json" \
--data-raw @- <<'JSON'
{
"model": "moonshotai/kimi-k2.5",
"input": "Give me one fun fact about cURL."
}
JSONMalformed JSON, unknown model, or invalid parameters.
Validate against schema; check model ID and docs; correct formatting.
Missing/invalid API key, or key lacks required scope.
Use the correct key and ensure scopes/permissions include this endpoint.
Insufficient funds/credit for the request.
Top up your wallet or reduce request cost (tokens/options).
Key is valid but not allowed to access this resource/model.
Update access policy or choose an allowed model/endpoint.
Resource or model ID does not exist.
Double-check IDs and the available models list.
Too many requests or token throughput exceeded.
Back off with jitter, respect headers, and retry (see guidance below).
HTTP/1.1 402 Payment Required
{
"error": {
"type": "payment_error",
"code": "insufficient_funds",
"message": "Your wallet balance is too low for this request.",
"generation_id": "G-abc123"
}
}HTTP/1.1 502 Bad Gateway
{
"error": {
"type": "server_error",
"code": "upstream_provider_error",
"message": "The model provider returned an error. Please retry.",
"generation_id": "G-xyz789"
}
}For 429 and transient 5xx, use exponential backoff with jitter (e.g. randomised 100-1500 ms, doubling on repeats). Include an Idempotency-Key header on write-like operations to avoid duplicates.
If you’re consistently seeing 4xx, something in the request likely needs adjusting. If you’re seeing 5xx, it may be our services or an upstream provider. We always include clear messages and a generation_id to help you diagnose fast.