fix: disable prompt caching for kimi-k2p5-turbo

Roo Code passes Anthropic-style cache_control blocks in OpenAI-compatible
API calls, which causes validation errors when using Fireworks as an
OpenAI-compatible provider. Disable prompt caching until this is fixed.
This commit is contained in:
Jonathan Evans 2026-04-06 14:16:57 -07:00
parent 917b558663
commit eb51e1117c

View file

@ -77,10 +77,9 @@ export const fireworksModels = {
maxTokens: 16384,
contextWindow: 262144,
supportsImages: true,
supportsPromptCache: true,
supportsPromptCache: false,
inputPrice: 0,
outputPrice: 0,
cacheReadsPrice: 0,
description:
"Kimi K2.5 Turbo router - same configuration as kimi-k2p5 but with improved performance through Fireworks routing (FirePass - https://docs.fireworks.ai/firepass).",
},