-
[BE] fix: register deepseek as a canonical provider so DeepSeek model prices load (#7434)
frostbyte_neo 发布于
2026-07-13 11:30:12 +00:00 Following the pattern of #7262 (azure), #7432 (perplexity) and #7433
(xai) —deepseekis not currently in PROVIDERS_MAPPING, so all 12
entries in model_prices_and_context_window.json tagged with
litellm_provider: "deepseek" silently drop at load time. Every one
publishes non-zero input+output rates: deepseek-chat, deepseek-reasoner,
deepseek-v4-flash, deepseek-v4-pro, plus the deepseek/deepseek-coder,
deepseek/deepseek-r1, deepseek/deepseek-v3, deepseek/deepseek-v3.2
variants.9 of the 12 also publish cache_read_input_token_cost. DeepSeek's cost
calc in LiteLLM (litellm/llms/deepseek/cost_calculator.py) is a
one-liner that delegates directly to generic_cost_per_token — the same
helper OpenAI/xAI use — so DeepSeek's usage payload follows OpenAI's
shape (prompt_tokens_details.cached_tokens flattened under
original_usage.*). Route it through the existing
textGenerationWithCacheCostOpenAI calculator; no new cache-calc method
needed.Parameterized test uses deepseek/deepseek-coder (no cache ->
textGenerationCost) and deepseek/deepseek-chat (input 2.8e-7, output
4.2e-7, cache_read 2.8e-8 -> textGenerationWithCacheCostOpenAI). Both
branches exercised in regressions.Co-authored-by: Claude Opus 4.7 (1M context) noreply@anthropic.com
下载附件