cloud
diapason.engine.cloud
¶
Cloud inference engine.
OpenAI, Anthropic, Google, MiniMax, and DeepSeek API backends.
CloudEngine
¶
Bases: InferenceEngine
Cloud inference via OpenAI, Anthropic, Google, MiniMax, and DeepSeek SDKs.
can_serve(model)
¶
Return True only if the provider client for model exists.
health() is True whenever any provider client is configured,
but a request for, say, a gpt-* model still needs the OpenAI
client specifically. Without this check the cloud engine gets picked
as a fallback (when the local engine is down) for a model it can't
serve, then dies at call time with "
stream_full(messages, *, model, temperature=0.7, max_tokens=1024, **kwargs)
async
¶
Yield StreamChunks with content, tool_calls, and finish_reason.
estimate_cost(model, prompt_tokens, completion_tokens)
¶
Estimate USD cost based on the hardcoded pricing table.
resolve_pricing(model)
¶
The (input, output) per-million rate for model, or None if unknown.
Exact match first, then the longest matching prefix. Longest matters:
the table holds both gpt-4o and gpt-4o-mini, and a dated variant
like gpt-4o-mini-2024-07-18 is a prefix match for both. Taking the
first hit in insertion order billed it at the full gpt-4o rate —
16× the real price — while MiniMax-M2.7-highspeed-v2 fell back to
plain MiniMax-M2.7 and was billed at half.