@henjicc/ai-sdk 0.2.8 → 0.4.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (244) hide show
  1. package/CHANGELOG.md +157 -141
  2. package/README.md +497 -455
  3. package/dist/capabilities/embedding/bailian.d.ts +23 -0
  4. package/dist/capabilities/embedding/bailian.js +10 -0
  5. package/dist/capabilities/embedding/bigmodel.d.ts +9 -0
  6. package/dist/capabilities/embedding/bigmodel.js +5 -0
  7. package/dist/capabilities/embedding/index.d.ts +1 -0
  8. package/dist/capabilities/embedding/index.js +1 -0
  9. package/dist/capabilities/embedding/ppio.d.ts +7 -0
  10. package/dist/capabilities/embedding/ppio.js +5 -0
  11. package/dist/capabilities/embedding/siliconflow.d.ts +21 -0
  12. package/dist/capabilities/embedding/siliconflow.js +10 -0
  13. package/dist/capabilities/embedding/volcengine.d.ts +10 -0
  14. package/dist/capabilities/embedding/volcengine.js +6 -0
  15. package/dist/capabilities/rerank/bailian.d.ts +21 -0
  16. package/dist/capabilities/rerank/bailian.js +14 -0
  17. package/dist/capabilities/rerank/bigmodel.d.ts +9 -0
  18. package/dist/capabilities/rerank/bigmodel.js +5 -0
  19. package/dist/capabilities/rerank/index.d.ts +1 -0
  20. package/dist/capabilities/rerank/index.js +1 -0
  21. package/dist/capabilities/rerank/ppio.d.ts +7 -0
  22. package/dist/capabilities/rerank/ppio.js +5 -0
  23. package/dist/capabilities/rerank/siliconflow.d.ts +21 -0
  24. package/dist/capabilities/rerank/siliconflow.js +10 -0
  25. package/dist/capabilities/retrieval/http.d.ts +3 -0
  26. package/dist/capabilities/retrieval/http.js +48 -0
  27. package/dist/capabilities/retrieval/module.d.ts +11 -0
  28. package/dist/capabilities/retrieval/module.js +110 -0
  29. package/dist/capabilities/retrieval/types.d.ts +54 -0
  30. package/dist/capabilities/retrieval/types.js +1 -0
  31. package/dist/capabilities/retrieval/validation.d.ts +10 -0
  32. package/dist/capabilities/retrieval/validation.js +65 -0
  33. package/dist/capabilities/types.d.ts +1 -1
  34. package/dist/catalog/apimart/gpt-image-2.5.model.d.ts +2 -0
  35. package/dist/catalog/apimart/gpt-image-2.5.model.js +51 -0
  36. package/dist/catalog/fal/gpt-image-2.5.model.d.ts +2 -0
  37. package/dist/catalog/fal/gpt-image-2.5.model.js +53 -0
  38. package/dist/catalog/fal/ic-light-v2.model.js +9 -13
  39. package/dist/catalog/grsai/gpt-image-2.5.model.d.ts +2 -0
  40. package/dist/catalog/grsai/gpt-image-2.5.model.js +54 -0
  41. package/dist/catalog/grsai/gptImage25Sizes.d.ts +1 -0
  42. package/dist/catalog/grsai/gptImage25Sizes.js +76 -0
  43. package/dist/catalog/index.js +112 -104
  44. package/dist/catalog/kie/gpt-image-2.5.model.d.ts +2 -0
  45. package/dist/catalog/kie/gpt-image-2.5.model.js +38 -0
  46. package/dist/catalog/shared/gptImage25.d.ts +16 -0
  47. package/dist/catalog/shared/gptImage25.js +48 -0
  48. package/dist/catalog/shared/gptImage25Pricing.d.ts +12 -0
  49. package/dist/catalog/shared/gptImage25Pricing.js +508 -0
  50. package/dist/llm/defaults.d.ts +1 -1
  51. package/dist/llm/defaults.js +1 -1
  52. package/dist/llm/discovery.d.ts +2 -0
  53. package/dist/llm/discovery.js +16 -5
  54. package/dist/llm/modelCatalogEntries.js +32 -0
  55. package/dist/llm/providerPresets.d.ts +3 -1
  56. package/dist/llm/providerPresets.js +6 -3
  57. package/dist/llm/providerReasoningRequest.js +14 -1
  58. package/dist/llm/siliconflow/index.d.ts +21 -0
  59. package/dist/llm/siliconflow/index.js +32 -0
  60. package/dist/llm/siliconflow/preset.d.ts +6 -0
  61. package/dist/llm/siliconflow/preset.js +33 -0
  62. package/dist/packs/models/apimart/gpt-image-2.5.d.ts +6 -0
  63. package/dist/packs/models/apimart/gpt-image-2.5.js +7 -0
  64. package/dist/packs/models/fal/gpt-image-2.5.d.ts +6 -0
  65. package/dist/packs/models/fal/gpt-image-2.5.js +7 -0
  66. package/dist/packs/models/grsai/gpt-image-2.5.d.ts +6 -0
  67. package/dist/packs/models/grsai/gpt-image-2.5.js +7 -0
  68. package/dist/packs/models/kie/gpt-image-2.5.d.ts +6 -0
  69. package/dist/packs/models/kie/gpt-image-2.5.js +7 -0
  70. package/dist/packs/provider-packs/apimart.d.ts +1 -1
  71. package/dist/packs/provider-packs/apimart.js +21 -20
  72. package/dist/packs/provider-packs/fal.d.ts +1 -1
  73. package/dist/packs/provider-packs/fal.js +37 -36
  74. package/dist/packs/provider-packs/grsai.d.ts +1 -1
  75. package/dist/packs/provider-packs/grsai.js +6 -5
  76. package/dist/packs/provider-packs/kie.d.ts +1 -1
  77. package/dist/packs/provider-packs/kie.js +28 -27
  78. package/dist/packs/tool-models/fal/{qwen-image-edit-2509-multiple-angles.d.ts → qwen-image-edit-2511-multiple-angles.d.ts} +1 -1
  79. package/dist/packs/tool-models/fal/{qwen-image-edit-2509-multiple-angles.js → qwen-image-edit-2511-multiple-angles.js} +1 -1
  80. package/dist/packs/tool-packs/fal-multi-angle-tools.js +1 -1
  81. package/dist/tool-packs/fal-multi-angle/models/qwen-image-edit-2511-multiple-angles.model.d.ts +2 -0
  82. package/dist/tool-packs/fal-multi-angle/models/{qwen-image-edit-2509-multiple-angles.model.js → qwen-image-edit-2511-multiple-angles.model.js} +20 -17
  83. package/docs/README.md +26 -26
  84. package/docs/consumers.md +56 -48
  85. package/docs/llm-adaptation/README.md +139 -137
  86. package/docs/llm-adaptation//344/276/233/345/272/224/345/225/206/DeepSeek.md +103 -91
  87. package/docs/llm-adaptation//344/276/233/345/272/224/345/225/206/Kimi.md +75 -75
  88. package/docs/llm-adaptation//344/276/233/345/272/224/345/225/206/MiniMax.md +84 -84
  89. package/docs/llm-adaptation//344/276/233/345/272/224/345/225/206//345/260/217/347/261/263MiMo.md +85 -85
  90. package/docs/llm-adaptation//344/276/233/345/272/224/345/225/206//346/231/272/350/260/261GLM.md +232 -232
  91. package/docs/llm-adaptation//344/276/233/345/272/224/345/225/206//347/201/253/345/261/261/345/274/225/346/223/216.md +87 -87
  92. package/docs/llm-adaptation//344/276/233/345/272/224/345/225/206//347/231/276/347/202/274Qwen.md +84 -84
  93. package/docs/llm-adaptation//346/226/207/346/241/243/351/207/207/351/233/206/346/211/213/345/206/214.md +88 -88
  94. package/docs/model-adaptation/Bria-Eraser/Bria-Eraser_Fal.md +32 -32
  95. package/docs/model-adaptation/Bria-/345/210/233/346/204/217/346/224/276/345/244/247/Bria-/345/210/233/346/204/217/346/224/276/345/244/247_Fal.md +31 -31
  96. package/docs/model-adaptation/ControlLight/ControlLight_Fal.md +85 -85
  97. package/docs/model-adaptation/FLUX-2-/345/244/232/350/247/222/345/272/246/FLUX-2-/345/244/232/350/247/222/345/272/246_Fal.md +107 -107
  98. package/docs/model-adaptation/Finegrain-Eraser/Finegrain-Eraser_Fal.md +31 -31
  99. package/docs/model-adaptation/Flux-Pro-Erase/Flux-Pro-Erase_Fal.md +34 -34
  100. package/docs/model-adaptation/Fun-ASR/Fun-ASR_/347/231/276/347/202/274.md +43 -43
  101. package/docs/model-adaptation/Fun-ASR-Flash-2026-06-15/Fun-ASR-Flash-2026-06-15_/347/231/276/347/202/274.md +41 -41
  102. package/docs/model-adaptation/Fun-ASR-Realtime/Fun-ASR-Realtime_/347/231/276/347/202/274.md +37 -37
  103. package/docs/model-adaptation/Fun-ASR-Realtime-2026-02-28/Fun-ASR-Realtime-2026-02-28_/347/231/276/347/202/274.md +39 -39
  104. package/docs/model-adaptation/GLM-5.3-Flash/GLM-5.3-Flash_/346/231/272/350/260/261.md +207 -207
  105. package/docs/model-adaptation/GPT-Image-2/GPT-Image-2_APIMart.md +173 -173
  106. package/docs/model-adaptation/GPT-Image-2/GPT-Image-2_Fal.md +94 -94
  107. package/docs/model-adaptation/GPT-Image-2/GPT-Image-2_Grsai.md +137 -137
  108. package/docs/model-adaptation/GPT-Image-2/GPT-Image-2_KIE.md +93 -93
  109. package/docs/model-adaptation/GPT-Image-2.5/GPT-Image-2.5_APIMart.md +66 -0
  110. package/docs/model-adaptation/GPT-Image-2.5/GPT-Image-2.5_Fal.md +55 -0
  111. package/docs/model-adaptation/GPT-Image-2.5/GPT-Image-2.5_Grsai.md +53 -0
  112. package/docs/model-adaptation/GPT-Image-2.5/GPT-Image-2.5_KIE.md +55 -0
  113. package/docs/model-adaptation/GPT-OSS-20B/GPT-OSS-20B_Groq.md +62 -62
  114. package/docs/model-adaptation/Gemini-Omni-Flash/Gemini-Omni-Flash_APIMart.md +124 -124
  115. package/docs/model-adaptation/Gemini-Omni-Flash/Gemini-Omni-Flash_Fal.md +84 -84
  116. package/docs/model-adaptation/Gemini-Omni-Flash/Gemini-Omni-Flash_KIE.md +129 -129
  117. package/docs/model-adaptation/Grok-Imagine/Grok-Imagine_KIE.md +36 -36
  118. package/docs/model-adaptation/Grok-Imagine-2.0/Grok-Imagine-2.0_APIMart.md +250 -250
  119. package/docs/model-adaptation/Grok-Imagine-2.0/Grok-Imagine-2.0_Fal.md +84 -84
  120. package/docs/model-adaptation/Grok-Imagine-2.0/Grok-Imagine-2.0_KIE.md +105 -105
  121. package/docs/model-adaptation/Hailuo-02/Hailuo-02_Fal.md +93 -93
  122. package/docs/model-adaptation/Hailuo-02/Hailuo-02_KIE.md +56 -56
  123. package/docs/model-adaptation/Hailuo-2.3/Hailuo-2.3_Fal.md +88 -88
  124. package/docs/model-adaptation/Hailuo-2.3/Hailuo-2.3_KIE.md +51 -51
  125. package/docs/model-adaptation/Hailuo-2.3/Hailuo-2.3_/346/264/276/346/254/247/344/272/221.md +70 -70
  126. package/docs/model-adaptation/IC-Light-v2/IC-Light-v2_Fal.md +97 -94
  127. package/docs/model-adaptation/Ideogram-/345/233/276/347/211/207/346/224/276/345/244/247/Ideogram-/345/233/276/347/211/207/346/224/276/345/244/247_Fal.md +41 -41
  128. package/docs/model-adaptation/Image-Apps-v2-/345/225/206/345/223/201/346/221/204/345/275/261/Image-Apps-v2-/345/225/206/345/223/201/346/221/204/345/275/261_Fal.md +72 -72
  129. package/docs/model-adaptation/Image-Apps-v2-/346/211/251/345/233/276/Image-Apps-v2-/346/211/251/345/233/276_Fal.md +93 -93
  130. package/docs/model-adaptation/Image-Apps-v2-/347/205/247/347/211/207/344/277/256/345/244/215/Image-Apps-v2-/347/205/247/347/211/207/344/277/256/345/244/215_Fal.md +86 -86
  131. package/docs/model-adaptation/Image-Apps-v2-/351/207/215/346/211/223/345/205/211/Image-Apps-v2-/351/207/215/346/211/223/345/205/211_Fal.md +84 -84
  132. package/docs/model-adaptation/Kling-3.0/Kling-3.0_APIMart.md +112 -112
  133. package/docs/model-adaptation/Kling-3.0/Kling-3.0_Fal.md +89 -89
  134. package/docs/model-adaptation/Kling-3.0/Kling-3.0_KIE.md +106 -106
  135. package/docs/model-adaptation/Kling-3.0/Kling-3.0_/346/264/276/346/254/247/344/272/221.md +119 -119
  136. package/docs/model-adaptation/Kling-3.0-Omni/Kling-3.0-Omni_APIMart.md +125 -125
  137. package/docs/model-adaptation/Kling-3.0-Omni/Kling-3.0-Omni_Fal.md +107 -107
  138. package/docs/model-adaptation/Kling-3.0-Omni/Kling-3.0-Omni_KIE.md +143 -143
  139. package/docs/model-adaptation/Kling-3.0-Turbo/Kling-3.0-Turbo_APIMart.md +85 -85
  140. package/docs/model-adaptation/Kling-3.0-Turbo/Kling-3.0-Turbo_Fal.md +83 -83
  141. package/docs/model-adaptation/Kling-3.0-Turbo/Kling-3.0-Turbo_KIE.md +84 -84
  142. package/docs/model-adaptation/Midjourney/Midjourney_APIMart.md +396 -396
  143. package/docs/model-adaptation/MiniMax-H3/MiniMax-H3_APIMart.md +234 -234
  144. package/docs/model-adaptation/MiniMax-H3/MiniMax-H3_Fal.md +108 -108
  145. package/docs/model-adaptation/MiniMax-H3/MiniMax-H3_KIE.md +111 -111
  146. package/docs/model-adaptation/MiniMax-Speech/MiniMax-Speech_/346/264/276/346/254/247/344/272/221.md +170 -170
  147. package/docs/model-adaptation/Nano-Banana-2/Nano-Banana-2_APIMart.md +89 -89
  148. package/docs/model-adaptation/Nano-Banana-2/Nano-Banana-2_Fal.md +99 -99
  149. package/docs/model-adaptation/Nano-Banana-2/Nano-Banana-2_Grsai.md +82 -82
  150. package/docs/model-adaptation/Nano-Banana-2/Nano-Banana-2_KIE.md +70 -70
  151. package/docs/model-adaptation/Nano-Banana-2-Lite/Nano-Banana-2-Lite_APIMart.md +79 -79
  152. package/docs/model-adaptation/Nano-Banana-2-Lite/Nano-Banana-2-Lite_Grsai.md +65 -65
  153. package/docs/model-adaptation/Nano-Banana-2-Lite/Nano-Banana-2-Lite_KIE.md +64 -64
  154. package/docs/model-adaptation/Nano-Banana-Pro/Nano-Banana-Pro_APIMart.md +83 -83
  155. package/docs/model-adaptation/Nano-Banana-Pro/Nano-Banana-Pro_Fal.md +89 -89
  156. package/docs/model-adaptation/Nano-Banana-Pro/Nano-Banana-Pro_Grsai.md +82 -82
  157. package/docs/model-adaptation/Nano-Banana-Pro/Nano-Banana-Pro_KIE.md +64 -64
  158. package/docs/model-adaptation/Pixelcut-/350/203/214/346/231/257/347/247/273/351/231/244/Pixelcut-/350/203/214/346/231/257/347/247/273/351/231/244_Fal.md +79 -79
  159. package/docs/model-adaptation/Qwen-Image-3.0/Qwen-Image-3.0_APIMart.md +117 -117
  160. package/docs/model-adaptation/Qwen-Image-3.0/Qwen-Image-3.0_Fal.md +84 -84
  161. package/docs/model-adaptation/Qwen-Image-3.0/Qwen-Image-3.0_KIE.md +103 -103
  162. package/docs/model-adaptation/Qwen-Image-3.0/Qwen-Image-3.0_/347/231/276/347/202/274.md +138 -138
  163. package/docs/model-adaptation/Qwen-Image-Edit-2511-/345/244/232/350/247/222/345/272/246/Qwen-Image-Edit-2511-/345/244/232/350/247/222/345/272/246_Fal.md +57 -0
  164. package/docs/model-adaptation/Qwen-MT-Flash/Qwen-MT-Flash_/347/231/276/347/202/274.md +47 -47
  165. package/docs/model-adaptation/Qwen-MT-Lite/Qwen-MT-Lite_/347/231/276/347/202/274.md +39 -39
  166. package/docs/model-adaptation/Qwen-MT-Plus/Qwen-MT-Plus_/347/231/276/347/202/274.md +39 -39
  167. package/docs/model-adaptation/Qwen3-ASR-Flash/Qwen3-ASR-Flash_/347/231/276/347/202/274.md +40 -40
  168. package/docs/model-adaptation/Qwen3-ASR-Flash-2026-02-10/Qwen3-ASR-Flash-2026-02-10_/347/231/276/347/202/274.md +33 -33
  169. package/docs/model-adaptation/Qwen3-ASR-Flash-Filetrans/Qwen3-ASR-Flash-Filetrans_/347/231/276/347/202/274.md +41 -41
  170. package/docs/model-adaptation/Qwen3-ASR-Flash-Realtime/Qwen3-ASR-Flash-Realtime_/347/231/276/347/202/274.md +35 -35
  171. package/docs/model-adaptation/Qwen3-ASR-Flash-Realtime-2026-02-10/Qwen3-ASR-Flash-Realtime-2026-02-10_/347/231/276/347/202/274.md +37 -37
  172. package/docs/model-adaptation/README.md +313 -295
  173. package/docs/model-adaptation/SeedASR-2.0-File/SeedASR-2.0-File_/347/201/253/345/261/261/345/274/225/346/223/216.md +115 -115
  174. package/docs/model-adaptation/SeedASR-2.0-Realtime/SeedASR-2.0-Realtime_/347/201/253/345/261/261/345/274/225/346/223/216.md +143 -143
  175. package/docs/model-adaptation/SeedVR2-/345/233/276/347/211/207/346/224/276/345/244/247/SeedVR2-/345/233/276/347/211/207/346/224/276/345/244/247_Fal.md +45 -45
  176. package/docs/model-adaptation/Seedance-2.0/Seedance-2.0_APIMart.md +121 -121
  177. package/docs/model-adaptation/Seedance-2.0/Seedance-2.0_Fal.md +97 -97
  178. package/docs/model-adaptation/Seedance-2.0/Seedance-2.0_KIE.md +86 -86
  179. package/docs/model-adaptation/Seedance-2.0-Fast/Seedance-2.0-Fast_APIMart.md +89 -89
  180. package/docs/model-adaptation/Seedance-2.0-Fast/Seedance-2.0-Fast_Fal.md +95 -95
  181. package/docs/model-adaptation/Seedance-2.0-Fast/Seedance-2.0-Fast_KIE.md +78 -78
  182. package/docs/model-adaptation/Seedance-2.0-Mini/Seedance-2.0-Mini_APIMart.md +89 -89
  183. package/docs/model-adaptation/Seedance-2.0-Mini/Seedance-2.0-Mini_Fal.md +96 -96
  184. package/docs/model-adaptation/Seedance-2.0-Mini/Seedance-2.0-Mini_KIE.md +78 -78
  185. package/docs/model-adaptation/Seedance-2.5/Seedance-2.5_APIMart.md +212 -212
  186. package/docs/model-adaptation/Seedance-2.5/Seedance-2.5_Fal.md +94 -94
  187. package/docs/model-adaptation/Seedance-2.5/Seedance-2.5_KIE.md +89 -89
  188. package/docs/model-adaptation/Seedream-5.0-Lite/Seedream-5.0-Lite_APIMart.md +106 -106
  189. package/docs/model-adaptation/Seedream-5.0-Lite/Seedream-5.0-Lite_Fal.md +91 -91
  190. package/docs/model-adaptation/Seedream-5.0-Lite/Seedream-5.0-Lite_KIE.md +89 -89
  191. package/docs/model-adaptation/Seedream-5.0-Lite/Seedream-5.0-Lite_/347/201/253/345/261/261/345/274/225/346/223/216.md +108 -108
  192. package/docs/model-adaptation/Seedream-5.0-Pro/Seedream-5.0-Pro_APIMart.md +194 -194
  193. package/docs/model-adaptation/Seedream-5.0-Pro/Seedream-5.0-Pro_Fal.md +100 -100
  194. package/docs/model-adaptation/Seedream-5.0-Pro/Seedream-5.0-Pro_KIE.md +141 -141
  195. package/docs/model-adaptation/Seedream-5.0-Pro/Seedream-5.0-Pro_/347/201/253/345/261/261/345/274/225/346/223/216.md +137 -137
  196. package/docs/model-adaptation/SenseVoiceSmall/SenseVoiceSmall_/347/241/205/345/237/272/346/265/201/345/212/250.md +69 -69
  197. package/docs/model-adaptation/TeleSpeechASR/TeleSpeechASR_/347/241/205/345/237/272/346/265/201/345/212/250.md +72 -72
  198. package/docs/model-adaptation/Topaz-/345/233/276/347/211/207/346/224/276/345/244/247/Topaz-/345/233/276/347/211/207/346/224/276/345/244/247_Fal.md +87 -87
  199. package/docs/model-adaptation/Topaz-/351/200/217/346/230/216/345/233/276/346/224/276/345/244/247/Topaz-/351/200/217/346/230/216/345/233/276/346/224/276/345/244/247_Fal.md +34 -34
  200. package/docs/model-adaptation/Wan-2.5-Preview/Wan-2.5-Preview_/346/264/276/346/254/247/344/272/221.md +82 -82
  201. package/docs/model-adaptation/Wan-2.6/Wan-2.6_/346/264/276/346/254/247/344/272/221.md +84 -84
  202. package/docs/model-adaptation/Wan-2.7/Wan-2.7_/346/264/276/346/254/247/344/272/221.md +122 -122
  203. package/docs/model-adaptation/Whisper-Large-v3/Whisper-Large-v3_Groq.md +76 -76
  204. package/docs/model-adaptation/Whisper-Large-v3-Turbo/Whisper-Large-v3-Turbo_Groq.md +75 -75
  205. package/docs/model-adaptation/Z-Image-Turbo/Z-Image-Turbo_APIMart.md +88 -88
  206. package/docs/model-adaptation/Z-Image-Turbo/Z-Image-Turbo_Fal.md +95 -95
  207. package/docs/model-adaptation/Z-Image-Turbo/Z-Image-Turbo_KIE.md +71 -71
  208. package/docs/model-adaptation/Z-Image-Turbo/Z-Image-Turbo_/347/231/276/347/202/274.md +122 -122
  209. package/docs/model-adaptation/Z-Image-Turbo/Z-Image-Turbo_/351/255/224/346/220/255.md +97 -97
  210. package/docs/model-adaptation//344/276/233/345/272/224/345/225/206/APIMart.md +238 -238
  211. package/docs/model-adaptation//344/276/233/345/272/224/345/225/206/Fal.md +172 -172
  212. package/docs/model-adaptation//344/276/233/345/272/224/345/225/206/Groq.md +121 -121
  213. package/docs/model-adaptation//344/276/233/345/272/224/345/225/206/Grsai.md +217 -217
  214. package/docs/model-adaptation//344/276/233/345/272/224/345/225/206/KIE.md +131 -131
  215. package/docs/model-adaptation//344/276/233/345/272/224/345/225/206//345/200/231/351/200/211/344/276/233/345/272/224/345/225/206/350/260/203/347/240/224-/345/233/276/347/211/207/345/267/245/345/205/267.md +151 -151
  216. package/docs/model-adaptation//344/276/233/345/272/224/345/225/206//345/277/253/351/200/237/351/200/202/351/205/215/344/276/233/345/272/224/345/225/206.md +65 -65
  217. package/docs/model-adaptation//344/276/233/345/272/224/345/225/206//346/264/276/346/254/247/344/272/221.md +147 -138
  218. package/docs/model-adaptation//344/276/233/345/272/224/345/225/206//347/201/253/345/261/261/345/274/225/346/223/216.md +181 -171
  219. package/docs/model-adaptation//344/276/233/345/272/224/345/225/206//347/231/276/347/202/274.md +237 -227
  220. package/docs/model-adaptation//344/276/233/345/272/224/345/225/206//347/241/205/345/237/272/346/265/201/345/212/250.md +140 -75
  221. package/docs/model-adaptation//344/276/233/345/272/224/345/225/206//351/255/224/346/220/255.md +207 -207
  222. package/docs/model-adaptation//346/226/207/346/241/243/351/207/207/351/233/206/346/211/213/345/206/214.md +338 -311
  223. package/docs/model-adaptation//351/200/217/350/247/206/345/217/230/346/215/242//351/200/217/350/247/206/345/217/230/346/215/242_Fal.md +59 -59
  224. package/docs//346/216/245/345/205/245/346/214/207/345/215/227/Electron.md +77 -77
  225. package/docs//346/216/245/345/205/245/346/214/207/345/215/227/Tauri.md +81 -81
  226. package/docs//346/216/245/345/205/245/346/214/207/345/215/227/UXP.md +169 -169
  227. package/docs//346/216/245/345/205/245/346/214/207/345/215/227//344/276/233/345/272/224/345/225/206/345/237/237/345/220/215.md +83 -83
  228. package/docs//346/216/245/345/205/245/346/214/207/345/215/227//351/224/231/350/257/257/345/244/204/347/220/206.md +72 -72
  229. package/examples/form-renderer/README.md +17 -17
  230. package/examples/form-renderer/cli.ts +32 -32
  231. package/examples/form-renderer/index.ts +144 -144
  232. package/examples/form-renderer/package.json +16 -16
  233. package/examples/form-renderer/tsconfig.json +14 -14
  234. package/examples/llm-chat/README.md +32 -32
  235. package/examples/llm-chat/index.ts +100 -100
  236. package/examples/llm-chat/package.json +19 -19
  237. package/examples/llm-chat/tsconfig.json +14 -14
  238. package/examples/minimal-node/README.md +31 -31
  239. package/examples/minimal-node/index.ts +138 -138
  240. package/examples/minimal-node/package.json +17 -17
  241. package/examples/minimal-node/tsconfig.json +14 -14
  242. package/package.json +255 -195
  243. package/dist/tool-packs/fal-multi-angle/models/qwen-image-edit-2509-multiple-angles.model.d.ts +0 -2
  244. package/docs/model-adaptation/Qwen-Image-Edit-2509-/345/244/232/350/247/222/345/272/246/Qwen-Image-Edit-2509-/345/244/232/350/247/222/345/272/246_Fal.md +0 -116
@@ -1,304 +1,322 @@
1
- # 模型适配清单(总索引)
2
-
3
- 本目录是**痕迹AI 的模型适配唯一资料源**。Henji-AI 同时是 `@henjicc/ai-sdk` 的主开发仓库与首发验证宿主;官方调研、实现和完整验证先在这里闭环,发布后消费项目才精确锁版本接入。适配思路已从「挑一个供应商、把它的模型尽量做全」调整为:
4
- **先定一份常用且优秀的模型清单,再为清单里的每个模型逐一接入所有支持它的供应商。**
5
-
6
- | 项目 | 内容 |
7
- |---|---|
8
- | 最后更新 | 2026-08-31 |
9
- | 模型数量 | 生成主清单 19(图片 10 / 视频 9)+ 主清单外存量 12 + SDK 跨项目能力模型 20(ASR 15 / 翻译 3 / LLM 2) |
10
- | 模型供应商文档数量 | 生成/存量 80 + SDK 跨项目能力 20(另有 Fal 工具 12) |
11
- | 覆盖供应商 | 火山引擎(官方生成/ASR)、百炼(官方)、硅基流动(官方 ASR)、智谱(官方 LLM)、Groq(官方 LLM/ASR)、APIMart、KIE、Fal、派欧云、魔搭、Grsai |
12
-
13
- 另有 12 个按能力按需分发的 Fal 图像工具,不进入默认 105 模型兼容目录:
14
-
15
- - 消除:[Flux Pro Erase](Flux-Pro-Erase/Flux-Pro-Erase_Fal.md)、[Bria Eraser](Bria-Eraser/Bria-Eraser_Fal.md)、[Finegrain Eraser](Finegrain-Eraser/Finegrain-Eraser_Fal.md)。
16
- - 图片实用工具:[Image Apps v2 重打光](Image-Apps-v2-重打光/Image-Apps-v2-重打光_Fal.md)、[ControlLight](ControlLight/ControlLight_Fal.md)、[Image Apps v2 扩图](Image-Apps-v2-扩图/Image-Apps-v2-扩图_Fal.md)、[Image Apps v2 商品摄影](Image-Apps-v2-商品摄影/Image-Apps-v2-商品摄影_Fal.md)、[Image Apps v2 照片修复](Image-Apps-v2-照片修复/Image-Apps-v2-照片修复_Fal.md)、[Pixelcut 背景移除](Pixelcut-背景移除/Pixelcut-背景移除_Fal.md)。
17
- - 多角度:[Qwen Image Edit 2509 多角度](Qwen-Image-Edit-2509-多角度/Qwen-Image-Edit-2509-多角度_Fal.md)、[透视变换](透视变换/透视变换_Fal.md)、[FLUX 2 多角度](FLUX-2-多角度/FLUX-2-多角度_Fal.md)。
18
-
19
- 工具均复用 Fal 队列、上传与结果解析;它们是独立的按需 SDK 入口,不是新的执行内核。Henji-AI 运行客户端显式装载三个 pack,但普通模型选择器仍固定为 105 个默认模型:六个图片实用工具通过图片能力菜单创建固定模型节点,三个多角度工具由专用多角度节点选择,消除工具保留给后续遮罩能力接入。真实质量、延迟和账单仍待获得付费授权后验证。
20
-
1
+ # 模型适配清单(总索引)
2
+
3
+ 本目录是**痕迹AI 的模型适配唯一资料源**。Henji-AI 同时是 `@henjicc/ai-sdk` 的主开发仓库与首发验证宿主;官方调研、实现和完整验证先在这里闭环,发布后消费项目才精确锁版本接入。适配思路已从「挑一个供应商、把它的模型尽量做全」调整为:
4
+ **先定一份常用且优秀的模型清单,再为清单里的每个模型逐一接入所有支持它的供应商。**
5
+
6
+ | 项目 | 内容 |
7
+ |---|---|
8
+ | 最后更新 | 2026-09-10 |
9
+ | 模型数量 | 生成主清单 20(图片 11 / 视频 9)+ 主清单外存量 12 + SDK 跨项目能力模型 20(ASR 15 / 翻译 3 / LLM 2) |
10
+ | 模型供应商文档数量 | 生成/存量 84 + SDK 跨项目能力 20(另有 Fal 工具 12) |
11
+ | 覆盖供应商 | 火山引擎(官方生成/ASR)、百炼(官方)、硅基流动(官方 ASR)、智谱(官方 LLM)、Groq(官方 LLM/ASR)、APIMart、KIE、Fal、派欧云、魔搭、Grsai |
12
+
13
+ 另有 12 个按能力按需分发的 Fal 图像工具,不进入默认 109 模型兼容目录:
14
+
15
+ - 消除:[Flux Pro Erase](Flux-Pro-Erase/Flux-Pro-Erase_Fal.md)、[Bria Eraser](Bria-Eraser/Bria-Eraser_Fal.md)、[Finegrain Eraser](Finegrain-Eraser/Finegrain-Eraser_Fal.md)。
16
+ - 图片实用工具:[Image Apps v2 重打光](Image-Apps-v2-重打光/Image-Apps-v2-重打光_Fal.md)、[ControlLight](ControlLight/ControlLight_Fal.md)、[Image Apps v2 扩图](Image-Apps-v2-扩图/Image-Apps-v2-扩图_Fal.md)、[Image Apps v2 商品摄影](Image-Apps-v2-商品摄影/Image-Apps-v2-商品摄影_Fal.md)、[Image Apps v2 照片修复](Image-Apps-v2-照片修复/Image-Apps-v2-照片修复_Fal.md)、[Pixelcut 背景移除](Pixelcut-背景移除/Pixelcut-背景移除_Fal.md)。
17
+ - 多角度:[Qwen Image Edit 2511 多角度](Qwen-Image-Edit-2511-多角度/Qwen-Image-Edit-2511-多角度_Fal.md)、[透视变换](透视变换/透视变换_Fal.md)、[FLUX 2 多角度](FLUX-2-多角度/FLUX-2-多角度_Fal.md)。
18
+
19
+ 工具均复用 Fal 队列、上传与结果解析;它们是独立的按需 SDK 入口,不是新的执行内核。Henji-AI 运行客户端显式装载三个 pack,但普通模型选择器仍固定为 109 个默认模型:六个图片实用工具通过图片能力菜单创建固定模型节点,三个多角度工具由专用多角度节点选择,消除工具保留给后续遮罩能力接入。真实质量、延迟和账单仍待获得付费授权后验证。
20
+
21
21
  ## 一、目录结构约定
22
22
 
23
- ```
24
- docs/model-adaptation/
25
- ├── README.md # 本文件:总清单 + 索引
26
- ├── 文档采集手册.md # 官方调研、事件契约、fixture 与 SDK 首发验证的唯一详细规范
27
- ├── 供应商/ # 先完成供应商公共协议,再做具体模型
28
- │ ├── 快速适配供应商.md # 新供应商的统一核对与落地清单
29
- │ ├── APIMart.md
30
- │ ├── KIE.md
31
- │ ├── Fal.md
32
- │ ├── Groq.md
33
- │ ├── 硅基流动.md
34
- │ ├── 百炼.md
35
- │ ├── 火山引擎.md
36
- │ ├── 派欧云.md
37
- │ └── 魔搭.md
38
- ├── <模型名>/ # 一个模型一个文件夹
39
- │ └── <模型名>_<供应商名>.md # 支持该模型的每个供应商一个文件
40
- ```
41
-
42
- - **供应商名**使用项目内统一的显示名:
43
- 派欧云 / Fal / 魔搭 / KIE / APIMart / 百炼 / 火山引擎 / 硅基流动 / Grsai / Groq(Grsai/Groq 沿用官方品牌拼写)
44
- - 每个供应商文件都是**自包含**的:接入协议、能力清单、请求参数、响应结构、价格、适配要点、原始链接索引,看完这一份就能完成该供应商的适配
45
- - 每份文件都标注**信息来源链接**以及**该链接是否需要登录**
46
-
47
- ### 供应商基础文档
48
-
49
- 模型适配前先看 [快速适配供应商](供应商/快速适配供应商.md),确认供应商公共层已经覆盖端点、鉴权、上传、任务查询、结果解析和计价。现有十家基础文档:
50
-
51
- - [APIMart](供应商/APIMart.md):含中国大陆备用线路与图片上传协议
52
- - [KIE](供应商/KIE.md):生成任务、文件上传、实际扣费与余额查询
53
- - [Fal](供应商/Fal.md):同步 / 队列协议、Fal CDN、官方计价 API
54
- - [百炼](供应商/百炼.md):地域端点、临时 OSS 上传与同步 / 异步边界
55
- - [火山引擎](供应商/火山引擎.md):方舟图片,以及 SeedASR 2.0 实时/文件两套独立协议与计费
56
- - [硅基流动](供应商/硅基流动.md):multipart 文件转写、免费模型与在线文档/历史公告冲突
57
- - [派欧云](供应商/派欧云.md):异步任务协议、**无官方上传(借用 KIE)**、实名认证前置、在售清单极小
58
- - [魔搭](供应商/魔搭.md):API-Inference 统一接口、**魔粒积分计费**、仅图片、可用性判定方法
59
- - [Grsai](供应商/Grsai.md):聚合中转,**同一模型拆多个「渠道」**(`-cl`/`-vip`/`-vt`/`-lite` 等后缀,价差可达 10 倍以上)、无独立上传接口(图片直接 base64/URL 塞进请求体)、双线路连通性探测
60
- - [Groq](供应商/Groq.md):OpenAI 兼容 LLM、Whisper 文件转写、动态模型发现、价格/免费层与错误边界
61
- - [智谱 GLM](../llm-adaptation/供应商/智谱GLM.md):国内 BigModel / 国际 Z.AI 两个 endpoint profile、OpenAI 兼容 Chat 与供应商级思考约定;模型专属契约仍以本目录模型文件为准
62
-
63
- 尚未适配的图片工具供应商候选统一记录在[图片工具候选供应商调研](供应商/候选供应商调研-图片工具.md):当前优先级为 Segmind / WaveSpeedAI(P0)、PiAPI(P1)、Replicate(P2)、Novita AI(P3);中国内地访问友好性均需真实网络验证,不能仅凭中文站或支付方式下结论。
64
-
65
- ## 二、通用适配规则
66
-
67
- 1. **默认绝对不显示的参数**:随机种子 `seed`、负面提示词。
68
- 即使供应商提供了这两个字段,也**不注册、不下发**。各文件的「适配要点」会逐一标出哪个供应商有这两个字段。
69
- 2. **`output_format` 默认不显示且不请求**,保持供应商默认值。
70
- 3. 价格以**供应商的定价页 / 模型 `llms.txt` / 任务返回的 `cost`** 为准;部分 API 文档正文中的价格已过期,出现不一致时文件内会同时列出两处并说明取舍。
71
- 4. 各供应商同名参数的**默认值、取值集合、字段名经常不一致**(例如 `adaptive` vs `auto`、整数 `5` vs 字符串 `"5"`),必须按供应商分别声明 schema,不要共用常量。
72
- 5. 价格依赖本地媒体像素、时长或文件指标时,由模型在 `pricing.mediaContext` 声明需求,宿主统一读取并注入 calculator;禁止在单个页面或画布节点复制媒体计价逻辑。指标未解析完成时不显示伪造的兜底总价。
73
-
74
- ## 三、模型清单
75
-
76
- ### 图片(10)
77
-
78
- | 模型 | 火山引擎 | 百炼 | APIMart | KIE | Fal | 派欧云 | 魔搭 | Grsai |
79
- |---|---|---|---|---|---|---|---|---|
80
- | [Seedream 5.0 Pro](Seedream-5.0-Pro/) | [✅ 官方](Seedream-5.0-Pro/Seedream-5.0-Pro_火山引擎.md) | — | [✅](Seedream-5.0-Pro/Seedream-5.0-Pro_APIMart.md) | [✅](Seedream-5.0-Pro/Seedream-5.0-Pro_KIE.md) | [✅](Seedream-5.0-Pro/Seedream-5.0-Pro_Fal.md) | ❌ 无 | ❌ 无 | ❌ 无 |
81
- | [Seedream 5.0 Lite](Seedream-5.0-Lite/) | [✅ 官方](Seedream-5.0-Lite/Seedream-5.0-Lite_火山引擎.md) | — | [✅](Seedream-5.0-Lite/Seedream-5.0-Lite_APIMart.md) | [✅](Seedream-5.0-Lite/Seedream-5.0-Lite_KIE.md) | [✅](Seedream-5.0-Lite/Seedream-5.0-Lite_Fal.md) | ❌ 无 | ❌ 无 | ❌ 无 |
82
- | [Qwen Image 3.0](Qwen-Image-3.0/) | — | [✅ 官方](Qwen-Image-3.0/Qwen-Image-3.0_百炼.md) | [✅](Qwen-Image-3.0/Qwen-Image-3.0_APIMart.md) | [✅](Qwen-Image-3.0/Qwen-Image-3.0_KIE.md) | [✅](Qwen-Image-3.0/Qwen-Image-3.0_Fal.md) | ❌ 无 | ❌ 无 | ❌ 无 |
83
- | [Z-Image Turbo](Z-Image-Turbo/) | — | [✅ 官方](Z-Image-Turbo/Z-Image-Turbo_百炼.md) | [✅](Z-Image-Turbo/Z-Image-Turbo_APIMart.md) | [✅](Z-Image-Turbo/Z-Image-Turbo_KIE.md) | [✅](Z-Image-Turbo/Z-Image-Turbo_Fal.md) | ❌ 无 | [✅](Z-Image-Turbo/Z-Image-Turbo_魔搭.md) | ❌ 无 |
84
- | [GPT-Image-2](GPT-Image-2/) | — | — | [✅](GPT-Image-2/GPT-Image-2_APIMart.md) | [✅](GPT-Image-2/GPT-Image-2_KIE.md) | [✅](GPT-Image-2/GPT-Image-2_Fal.md) | ❌ 无 | ❌ 无 | [✅](GPT-Image-2/GPT-Image-2_Grsai.md) |
85
- | [IC-Light v2](IC-Light-v2/) | — | — | ❌ 无 | ❌ 无 | [✅](IC-Light-v2/IC-Light-v2_Fal.md) | ❌ 无 | ❌ 无 | ❌ 无 |
86
- | [Nano Banana 2](Nano-Banana-2/) | — | — | [✅](Nano-Banana-2/Nano-Banana-2_APIMart.md) | [✅](Nano-Banana-2/Nano-Banana-2_KIE.md) | [✅](Nano-Banana-2/Nano-Banana-2_Fal.md) | ❌ 无 | ❌ 无 | [✅](Nano-Banana-2/Nano-Banana-2_Grsai.md) |
87
- | [Nano Banana 2 Lite](Nano-Banana-2-Lite/) | — | — | [✅](Nano-Banana-2-Lite/Nano-Banana-2-Lite_APIMart.md) | [✅](Nano-Banana-2-Lite/Nano-Banana-2-Lite_KIE.md) | ❌ 无(已探测确认,见 [Fal 文档说明](Nano-Banana-2/Nano-Banana-2_Fal.md)) | ❌ 无 | ❌ 无 | [✅](Nano-Banana-2-Lite/Nano-Banana-2-Lite_Grsai.md) |
88
- | [Nano Banana Pro](Nano-Banana-Pro/) | — | — | [✅](Nano-Banana-Pro/Nano-Banana-Pro_APIMart.md) | [✅](Nano-Banana-Pro/Nano-Banana-Pro_KIE.md) | [✅](Nano-Banana-Pro/Nano-Banana-Pro_Fal.md) | ❌ 无 | ❌ 无 | [✅](Nano-Banana-Pro/Nano-Banana-Pro_Grsai.md) |
89
- | [Midjourney](Midjourney/) | — | — | [✅](Midjourney/Midjourney_APIMart.md) | ❌ 无 | ❌ 无 | ❌ 无 | ❌ 无 | ❌ 无 |
90
- | [Grok Imagine 2.0](Grok-Imagine-2.0/) | — | — | [✅](Grok-Imagine-2.0/Grok-Imagine-2.0_APIMart.md) | [✅](Grok-Imagine-2.0/Grok-Imagine-2.0_KIE.md) | [✅](Grok-Imagine-2.0/Grok-Imagine-2.0_Fal.md) | ❌ 无 | ❌ 无 | ❌ 无 |
91
-
92
- **Nano Banana 2 Lite 是独立模型,不是 Nano Banana 2 的分辨率/渠道选项**:代码里两者 `canonicalModelId` 不同(`nano-banana-2` vs `nano-banana-2-lite`),`generationModelDescriptions.ts` 也各自维护一条独立描述,APIMart/KIE 均已用独立 `.model.ts` 文件实现,因此本表把它列为单独一行,不再挂在 Nano Banana 2 文档里。
93
-
94
- ### 视频(9)
95
-
96
- | 模型 | APIMart | KIE | Fal | 派欧云 | 魔搭 |
97
- |---|---|---|---|---|---|
98
- | [Seedance 2.0](Seedance-2.0/) | [✅](Seedance-2.0/Seedance-2.0_APIMart.md) | [✅](Seedance-2.0/Seedance-2.0_KIE.md) | [✅](Seedance-2.0/Seedance-2.0_Fal.md) | ❌ 已下架 | ❌ 无 |
99
- | [Seedance 2.0 Fast](Seedance-2.0-Fast/) | [✅](Seedance-2.0-Fast/Seedance-2.0-Fast_APIMart.md) | [✅](Seedance-2.0-Fast/Seedance-2.0-Fast_KIE.md) | [✅](Seedance-2.0-Fast/Seedance-2.0-Fast_Fal.md) | ❌ 已下架 | ❌ 无 |
100
- | [Seedance 2.0 Mini](Seedance-2.0-Mini/) | [✅](Seedance-2.0-Mini/Seedance-2.0-Mini_APIMart.md) | [✅](Seedance-2.0-Mini/Seedance-2.0-Mini_KIE.md) | [✅](Seedance-2.0-Mini/Seedance-2.0-Mini_Fal.md) | ❌ 已下架 | ❌ 无 |
101
- | [Seedance 2.5](Seedance-2.5/) | [✅](Seedance-2.5/Seedance-2.5_APIMart.md) | [✅](Seedance-2.5/Seedance-2.5_KIE.md) | [✅](Seedance-2.5/Seedance-2.5_Fal.md) | ❌ 已下架 | ❌ 无 |
102
- | [MiniMax H3](MiniMax-H3/) | [✅](MiniMax-H3/MiniMax-H3_APIMart.md) | [✅](MiniMax-H3/MiniMax-H3_KIE.md) | [✅](MiniMax-H3/MiniMax-H3_Fal.md) | ❌ 无 | ❌ 无 |
103
- | [Kling 3.0](Kling-3.0/) | [✅](Kling-3.0/Kling-3.0_APIMart.md) | [✅](Kling-3.0/Kling-3.0_KIE.md) | [✅](Kling-3.0/Kling-3.0_Fal.md) | [✅](Kling-3.0/Kling-3.0_派欧云.md) | ❌ 无 |
104
- | [Kling 3.0 Turbo](Kling-3.0-Turbo/) | [✅](Kling-3.0-Turbo/Kling-3.0-Turbo_APIMart.md) | [✅](Kling-3.0-Turbo/Kling-3.0-Turbo_KIE.md) | [✅](Kling-3.0-Turbo/Kling-3.0-Turbo_Fal.md) | ❌ 无 | ❌ 无 |
105
- | [Kling 3.0 Omni](Kling-3.0-Omni/) | [✅](Kling-3.0-Omni/Kling-3.0-Omni_APIMart.md) | [✅](Kling-3.0-Omni/Kling-3.0-Omni_KIE.md) | [✅](Kling-3.0-Omni/Kling-3.0-Omni_Fal.md) | ❌ 无 | ❌ 无 |
106
- | [Gemini Omni Flash](Gemini-Omni-Flash/) | [✅](Gemini-Omni-Flash/Gemini-Omni-Flash_APIMart.md) | [✅](Gemini-Omni-Flash/Gemini-Omni-Flash_KIE.md) | [✅](Gemini-Omni-Flash/Gemini-Omni-Flash_Fal.md) | ❌ 无 | ❌ 无 |
107
-
108
- 视频类模型在**火山引擎与百炼上未纳入本清单**:Seedance 系列的官方接口不在本次适配范围,Kling 官方接口按用户要求**仅作参考、暂不适配**(能力地图:https://www.klingai.com/document-api/guides/capability-map/video )。
109
-
110
- Grsai 暂未在此表出现:其站内「Veo API」旧版文档给出了 `veo3.1-fast` 视频模型的端点,但该模型**未出现在 dashboard「模型列表」的定价清单中**,可用性与计价未确认,详见 [Grsai 基础文档 §8](供应商/Grsai.md)。
23
+ ### 文本 Embedding / Rerank(2026-09-11)
111
24
 
112
- ### 主清单外存量模型(12)
25
+ SDK 使用独立 `capabilities/embedding/<provider>` 与 `capabilities/rerank/<provider>` 入口,显式注册到能力客户端和发现目录,不混入图片生成目录。
113
26
 
114
- 以下模型**不在主清单里**,但已在代码中适配并对用户开放,文档用于维护现有实现。部分模型已有多个供应商入口,未来是否纳入主清单待定。
115
-
116
- | 模型 | 模态 | 供应商 | 文档 |
117
- |---|---|---|---|
118
- | MiniMax Hailuo 2.3 | 视频 | 派欧云 / KIE / Fal | [派欧云](Hailuo-2.3/Hailuo-2.3_派欧云.md) / [KIE](Hailuo-2.3/Hailuo-2.3_KIE.md) / [Fal](Hailuo-2.3/Hailuo-2.3_Fal.md) |
119
- | MiniMax Hailuo 02 | 视频 | KIE / Fal | [KIE](Hailuo-02/Hailuo-02_KIE.md) / [Fal](Hailuo-02/Hailuo-02_Fal.md) |
120
- | Grok Imagine | 图片 | KIE | [KIE](Grok-Imagine/Grok-Imagine_KIE.md) |
121
- | Wan 2.5 Preview | 视频 | 派欧云 | [Wan-2.5-Preview_派欧云.md](Wan-2.5-Preview/Wan-2.5-Preview_派欧云.md) |
122
- | Wan 2.6 | 视频 | 派欧云 | [Wan-2.6_派欧云.md](Wan-2.6/Wan-2.6_派欧云.md) |
123
- | Wan 2.7 | 视频 | 派欧云 | [Wan-2.7_派欧云.md](Wan-2.7/Wan-2.7_派欧云.md) |
124
- | MiniMax Speech | 音频 | 派欧云 | [MiniMax-Speech_派欧云.md](MiniMax-Speech/MiniMax-Speech_派欧云.md) |
125
- | Topaz 图片放大 | 图片 | Fal | [Topaz-图片放大_Fal.md](Topaz-图片放大/Topaz-图片放大_Fal.md) |
126
- | Topaz 透明图放大 | 图片 | Fal | [Topaz-透明图放大_Fal.md](Topaz-透明图放大/Topaz-透明图放大_Fal.md) |
127
- | SeedVR2 图片放大 | 图片 | Fal | [SeedVR2-图片放大_Fal.md](SeedVR2-图片放大/SeedVR2-图片放大_Fal.md) |
128
- | Bria 创意放大 | 图片 | Fal | [Bria-创意放大_Fal.md](Bria-创意放大/Bria-创意放大_Fal.md) |
129
- | Ideogram 图片放大 | 图片 | Fal | [Ideogram-图片放大_Fal.md](Ideogram-图片放大/Ideogram-图片放大_Fal.md) |
130
-
131
- 魔搭侧另有 5 个已适配的开源图像模型(`Qwen/Qwen-Image`、`Qwen/Qwen-Image-Edit-2509`、`black-forest-labs/FLUX.1-Krea-dev`、`MusePublic/majicMIX_realistic`、`MusePublic/14_ckpt_SD_XL`)。它们共用魔搭 API-Inference 的同一套端点与参数,差异只有 model ID、尺寸上限和魔粒档位,因此不单独建文件,统一记在 [供应商/魔搭.md](供应商/魔搭.md) 第 7 节。
132
-
133
- ### SDK / Say-It 语音识别(15)
134
-
135
- 以下 15 个模型均有独立 SDK capability descriptor 与按需执行模块;消费应用仍需显式导入和注册,升级包版本不会自动增加界面入口。火山文件版只接受供应商可访问的公网 URL,Say-It 在没有受管上传/TOS 之前不得把它伪装成本地文件入口。
136
-
137
- | 模型 | 模式 | 主协议 | 上限 | 文档 |
138
- |---|---|---|---|---|
139
- | `fun-asr-realtime-2026-02-28` | 实时 | WebSocket `run-task` | 时长无限制 | [百炼](Fun-ASR-Realtime-2026-02-28/Fun-ASR-Realtime-2026-02-28_百炼.md) |
140
- | `fun-asr-realtime` | 实时稳定别名 | WebSocket `run-task` | 时长无限制 | [百炼](Fun-ASR-Realtime/Fun-ASR-Realtime_百炼.md) |
141
- | `qwen3-asr-flash-realtime-2026-02-10` | 实时 | WebSocket session | 时长无限制 | [百炼](Qwen3-ASR-Flash-Realtime-2026-02-10/Qwen3-ASR-Flash-Realtime-2026-02-10_百炼.md) |
142
- | `qwen3-asr-flash-realtime` | 实时稳定别名 | WebSocket session | 时长无限制 | [百炼](Qwen3-ASR-Flash-Realtime/Qwen3-ASR-Flash-Realtime_百炼.md) |
143
- | `fun-asr-flash-2026-06-15` | 短音频 | DashScope HTTP/SSE 同步 | 5 分钟 / 2 GB | [百炼](Fun-ASR-Flash-2026-06-15/Fun-ASR-Flash-2026-06-15_百炼.md) |
144
- | `qwen3-asr-flash` | 短音频稳定别名 | OpenAI 兼容 / DashScope 同步 | 5 分钟 / 10 MB | [百炼](Qwen3-ASR-Flash/Qwen3-ASR-Flash_百炼.md) |
145
- | `qwen3-asr-flash-2026-02-10` | 短音频快照 | OpenAI 兼容 / DashScope 同步 | 5 分钟 / 10 MB | [百炼](Qwen3-ASR-Flash-2026-02-10/Qwen3-ASR-Flash-2026-02-10_百炼.md) |
146
- | `fun-asr` | 文件异步 | 提交 + 轮询 + 结果 JSON | 12 小时 / 2 GB | [百炼](Fun-ASR/Fun-ASR_百炼.md) |
147
- | `qwen3-asr-flash-filetrans` | 文件异步 | 提交 + 轮询 | 12 小时 / 2 GB | [百炼](Qwen3-ASR-Flash-Filetrans/Qwen3-ASR-Flash-Filetrans_百炼.md) |
148
- | `seedasr-2.0-realtime` | 实时 | 火山二进制 WebSocket | PCM S16LE / 16kHz / mono | [火山引擎](SeedASR-2.0-Realtime/SeedASR-2.0-Realtime_火山引擎.md) |
149
- | `seedasr-2.0-file` | 文件异步 | HTTP submit + query | 仅公网 URL;本地文件需外部 TOS/受管上传 | [火山引擎](SeedASR-2.0-File/SeedASR-2.0-File_火山引擎.md) |
150
- | `FunAudioLLM/SenseVoiceSmall` | 文件同步 | multipart HTTP | 1 小时 / 50 MB | [硅基流动](SenseVoiceSmall/SenseVoiceSmall_硅基流动.md) |
151
- | `TeleAI/TeleSpeechASR` | 文件同步 | multipart HTTP | 1 小时 / 50 MB | [硅基流动](TeleSpeechASR/TeleSpeechASR_硅基流动.md) |
152
- | `whisper-large-v3-turbo` | 文件同步 | OpenAI-compatible multipart | 免费层 25 MB;Developer 100 MB | [Groq](Whisper-Large-v3-Turbo/Whisper-Large-v3-Turbo_Groq.md) |
153
- | `whisper-large-v3` | 文件同步 | OpenAI-compatible multipart | 免费层 25 MB;Developer 100 MB | [Groq](Whisper-Large-v3/Whisper-Large-v3_Groq.md) |
154
-
155
- ### Say-It 翻译(3)
156
-
157
- | 模型 | 流式语义 | 北京原价(输入/输出,每百万 Token) | 文档 |
158
- |---|---|---|---|
159
- | `qwen-mt-flash` | 增量 | 0.7 / 1.95 元 | [百炼](Qwen-MT-Flash/Qwen-MT-Flash_百炼.md) |
160
- | `qwen-mt-plus` | 累积序列 | 1.8 / 5.4 元 | [百炼](Qwen-MT-Plus/Qwen-MT-Plus_百炼.md) |
161
- | `qwen-mt-lite` | 增量 | 0.6 / 1.6 元 | [百炼](Qwen-MT-Lite/Qwen-MT-Lite_百炼.md) |
162
-
163
- ### SDK LLM(2)
164
-
165
- | 模型 | 供应商 | 备注 | 文档 |
27
+ | 供应商 | Embedding | Rerank | 协议资料 |
166
28
  |---|---|---|---|
167
- | `openai/gpt-oss-20b` | Groq | Say-It 默认;`reasoning_effort=low/medium/high`;不支持 `reasoning_format` | [Groq](GPT-OSS-20B/GPT-OSS-20B_Groq.md) |
168
- | `glm-5.3-flash` | 智谱 | 原生多模态 Chat 模型;1M 上下文 / 128K 输出;完整能力仍有官方冲突项 | [智谱](GLM-5.3-Flash/GLM-5.3-Flash_智谱.md) |
169
-
170
- ## 四、平台模型 ID 速查
29
+ | 硅基流动 | Qwen3 8B/4B/0.6B、BGE-M3 | Qwen3 8B/4B/0.6B、BGE v2 M3 | [供应商资料](供应商/硅基流动.md) |
30
+ | 百炼 | Qwen3.7、Qwen3.7 Flash、v4 | Qwen3.7、Qwen3、GTE v2 | [供应商资料](供应商/百炼.md) |
31
+ | 派欧云 | BGE-M3 | BGE v2 M3 | [供应商资料](供应商/派欧云.md) |
32
+ | 智谱 | embedding-3 | rerank | 见下方官方依据 |
33
+ | 火山方舟 | Doubao Embedding Vision 251215(文本单条) | VikingDB 独立服务暂未适配 | [供应商资料](供应商/火山引擎.md) |
171
34
 
172
- ### 可选工具模型
35
+ 智谱直接依据:[Embedding API](https://docs.bigmodel.cn/api-reference/模型-api/文本嵌入)、[Embedding-3 模型](https://docs.bigmodel.cn/cn/guide/models/embedding/embedding-3)、[Rerank API](https://docs.bigmodel.cn/api-reference/模型-api/文本重排序)、[公开价格](https://bigmodel.cn/pricing)。国内根地址 `https://open.bigmodel.cn`,Bearer。Embedding POST `/api/paas/v4/embeddings`,model=embedding-3,input 字符串数组,最多 64 条,dimensions=256/512/1024/2048;返回 data 数组的 index/embedding。字段表限制每条 3072 tokens,概述 8K 表述存在冲突,不用字符数猜 token 限制。Rerank POST `/api/paas/v4/rerank`,model=rerank,query/documents/top_n;最多 128 条,每条及 query 最多 4096 字符;返回 results 的 index/relevance_score。省略 top_n 返回全部;SDK 显式 topN 必须为正整数,超出候选数裁到候选数。文本从原输入索引回填。Embedding 公开价 0.5 元/百万 tokens,Rerank 搜索可见官方价格 0.8 元/百万 tokens,价格页运行时需要 JavaScript,实际账单以控制台为准。
173
36
 
174
- | 工具 | Fal endpoint ID | SDK 入口 |
175
- |---|---|---|
176
- | Flux Pro Erase | `fal-ai/flux-pro/v1/erase` | `tool-models/fal/flux-pro-erase` |
177
- | Bria Eraser | `fal-ai/bria/eraser` | `tool-models/fal/bria-eraser` |
178
- | Finegrain Eraser (Mask) | `fal-ai/finegrain-eraser/mask` | `tool-models/fal/finegrain-eraser` |
179
- | Qwen Image Edit 2509 多角度 | `fal-ai/qwen-image-edit-2509-lora-gallery/multiple-angles` | `@henjicc/ai-sdk/tool-models/fal/qwen-image-edit-2509-multiple-angles` |
180
- | 透视变换 | `fal-ai/image-apps-v2/perspective` | `@henjicc/ai-sdk/tool-models/fal/perspective-change` |
181
- | Image Apps v2 重打光 | `fal-ai/image-apps-v2/relighting` | `tool-models/fal/relighting` |
182
- | ControlLight | `fal-ai/control-light` | `tool-models/fal/control-light` |
183
- | Image Apps v2 扩图 | `fal-ai/image-apps-v2/outpaint` | `tool-models/fal/outpaint` |
184
- | Image Apps v2 商品摄影 | `fal-ai/image-apps-v2/product-photography` | `tool-models/fal/product-photography` |
185
- | Image Apps v2 照片修复 | `fal-ai/image-apps-v2/photo-restoration` | `tool-models/fal/photo-restoration` |
186
- | Pixelcut 背景移除 | `pixelcut/background-removal` | `tool-models/fal/pixelcut-background-removal` |
187
- | FLUX 2 多角度 | `fal-ai/flux-2-lora-gallery/multiple-angles` | `tool-models/fal/flux-2-multiple-angles` |
37
+ 共同验证边界:非流式单次 HTTP;空结果、缺失/重复/越界索引、非数值向量/分数不当作成功。向量维度必须一致;显式要求维度必须匹配。HTTP 错误、JSON 错误、供应商 error/code、取消和超时均失败;不自动重试付费 POST,不记录原文或凭据。未做真实付费验证。
188
38
 
189
- - 3 个消除工具聚合为 `@henjicc/ai-sdk/tool-packs/fal-image-edit-tools`,统一能力画像为 `operation=image-edit`、`feature=erase`。
190
- - 6 个重打光/修复/扩图/商品/抠图工具聚合为 `@henjicc/ai-sdk/tool-packs/fal-image-utility-tools`,再按单模型画像区分具体能力。
191
- - Qwen 2509、透视变换和 FLUX 2 聚合为 `@henjicc/ai-sdk/tool-packs/fal-multi-angle-tools`,与消除和实用工具 pack 分开分发。
39
+ 其他现有供应商也进行了检索核对:[APIMart 官方 AnythingLLM 指南](https://docs.apimart.ai/cn/integrations/chat/anythingllm) 提及 text-embedding-3-small/large,但本轮未取到独立端点字段表与价格,保留待核验;[魔搭 API 推理介绍](https://modelscope.cn/docs/model-service/API-Inference/intro) 本次无法读取正文,模型仓库存在权重不等于托管 API 已开放。Groq、DeepSeek、Fal、KIE、Grsai 的公开入口搜索未获得足以适配这两类能力的完整契约,不据此断言平台不支持。
192
40
 
193
- 三个 pack 都只是静态分发集合,执行继续走同一生成客户端和 Fal 公共协议。
194
-
195
- | 模型 | APIMart | KIE | Fal |
196
- |---|---|---|---|
197
- | Seedream 5.0 Pro | `seedream-5-0-pro` | `seedream/5-pro-text-to-image`、`5-pro-image-to-image`、`5-pro-layer-decomposition` | `bytedance/seedream/v5/pro/{text-to-image,edit}` |
198
- | Seedream 5.0 Lite | `seedream-5-0-lite` | `seedream/5-lite-text-to-image`、`5-lite-image-to-image` | `bytedance/seedream/v5/lite/{text-to-image,edit}` |
199
- | Qwen Image 3.0 | `qwen-image-3.0`、`qwen-image-3.0-pro` | `qwen3/{text-to-image,image-to-image}`、`qwen3-pro/text-to-image`、`qwen3/pro-image-to-image` | `alibaba/qwen-image-3/{text-to-image,edit}` |
200
- | Z-Image Turbo | `z-image-turbo` | `z-image` | `fal-ai/z-image/turbo`、`fal-ai/z-image/turbo/image-to-image`;魔搭 `Tongyi-MAI/Z-Image-Turbo` |
201
- | GPT-Image-2 | `gpt-image-2`(别名 `gpt-image-2-ext`)、`gpt-image-2-official` | `gpt-image-2-text-to-image`、`gpt-image-2-image-to-image` | `openai/gpt-image-2`、`openai/gpt-image-2/edit` |
202
- | IC-Light v2 | — | — | `fal-ai/iclight-v2` |
203
- | Topaz 图片放大 | — | — | `topaz/upscale/image/{precision,creative,generative}` |
204
- | Topaz 透明图放大 | — | — | `topaz/upscale/image/transparent` |
205
- | SeedVR2 图片放大 | — | — | `fal-ai/seedvr/upscale/image` |
206
- | Bria 创意放大 | — | — | `bria/upscale/creative` |
207
- | Ideogram 图片放大 | — | — | `fal-ai/ideogram/upscale` |
208
- | Nano Banana 2 | `gemini-3.1-flash-image-preview`(-official) | `nano-banana-2` | `fal-ai/nano-banana-2`、`fal-ai/nano-banana-2/edit` |
209
- | Nano Banana 2 Lite | `gemini-3.1-flash-lite-image`(-ext) | `nano-banana-2-lite` | ❌ 无(已探测确认) |
210
- | Nano Banana Pro | `gemini-3-pro-image-preview`(-official) | `nano-banana-pro` | `fal-ai/nano-banana-pro`、`fal-ai/nano-banana-pro/edit` |
211
- | Midjourney | `midjourney`(Imagine)、`midjourney-blend`、`midjourney-edit`、`midjourney-video`(新路由均自动注入平台模型 ID) | — | — |
212
- | Grok Imagine 2.0 | `grok-imagine-2.0-ext`、`grok-imagine-image-2.0` | `grok-imagine-image-2-0/{text-to-image,image-edit,segment-map,segment-edit}` | `xai/grok-imagine-image/v2.0/{text-to-image,edit}` |
213
- | Grok Imagine | | `grok-imagine/{text-to-image,image-to-image}` | |
214
- | Seedance 2.0 | `seedance-2.0` | `bytedance/seedance-2` | `bytedance/seedance-2.0/{text-to-video,image-to-video,reference-to-video}` |
215
- | Seedance 2.0 Fast | `seedance-2.0-fast` | `bytedance/seedance-2-fast` | `bytedance/seedance-2.0/fast/...` |
216
- | Seedance 2.0 Mini | `seedance-2.0-mini` | `bytedance/seedance-2-mini` | `bytedance/seedance-2.0/mini/...` |
217
- | Seedance 2.5 | `seedance-2.5` | `bytedance/seedance-2-5` | `bytedance/seedance-2.5/...` |
218
- | MiniMax H3 | `MiniMax-H3`、`MiniMax-H3-Context-IR`、`MiniMax-H3-Regeneration` | `minimax-h3/{text-to-video,image-to-video,reference-to-video}` | `minimax/h3/{text-to-video,image-to-video,reference-to-video}` |
219
- | MiniMax Hailuo 2.3 | — | `hailuo/2-3-image-to-video-{standard,pro}` | `fal-ai/minimax/hailuo-2.3/{standard,pro}/{text-to-video,image-to-video}`、`fal-ai/minimax/hailuo-2.3-fast/{standard,pro}/image-to-video` |
220
- | MiniMax Hailuo 02 | — | `hailuo/02-{text,image}-to-video-{standard,pro}` | `fal-ai/minimax/hailuo-02/{standard,pro}/{text-to-video,image-to-video}`、`fal-ai/minimax/hailuo-02-fast/image-to-video` |
221
- | Kling 3.0 | `kling-v3` | `kling-3.0/video` | `fal-ai/kling-video/v3/{pro,standard}/...`;派欧云 `/v3/async/kling-v3.0-{std,pro,4k}-{t2v,i2v}` |
222
- | Kling 3.0 Turbo | `kling-3.0-turbo` | `kling/v3-turbo-{text,image}-to-video` | `fal-ai/kling-video/v3/turbo/{pro,standard}/...` |
223
- | Kling 3.0 Omni | `kling-v3-omni` | `kling-3.0-omni/{text-to-video,image-to-video,reference-to-video,transformation}` | `fal-ai/kling-video/o3/{pro,standard}/...` |
224
- | Gemini Omni Flash | `gemini-omni-flash-preview`、`Omni-Flash-Ext` | `gemini-omni-video` + `/api/v1/omni/audio/create` + `/api/v1/omni/character/create` | `google/gemini-omni-flash/{image-to-video,reference-to-video}` |
225
-
226
- Grsai 的平台模型 ID 未并入本表:它把「渠道」直接编码进 `model` 字段(如 `nano-banana-2-cl`、`gpt-image-2-vip`),渠道数量比 APIMart/KIE/Fal 常见的 2 档更多(GPT-Image-2 2 个、Nano Banana 2 主模型 4 个、Nano Banana 2 Lite 1 个、Nano Banana Pro 5 个),完整枚举见对应模型文档:[GPT-Image-2_Grsai.md](GPT-Image-2/GPT-Image-2_Grsai.md)、[Nano-Banana-2_Grsai.md](Nano-Banana-2/Nano-Banana-2_Grsai.md)、[Nano-Banana-2-Lite_Grsai.md](Nano-Banana-2-Lite/Nano-Banana-2-Lite_Grsai.md)、[Nano-Banana-Pro_Grsai.md](Nano-Banana-Pro/Nano-Banana-Pro_Grsai.md)。
227
-
228
- ## 五、供应商通用协议速查
229
-
230
- | 供应商 | Base URL | 鉴权 | 提交 | 查询 | 结果字段 |
231
- |---|---|---|---|---|---|
232
- | **火山引擎** | 方舟 `https://ark.cn-beijing.volces.com`;语音 `https://openspeech.bytedance.com` | 方舟 Bearer;语音 `X-Api-Key` + Resource/Request ID | 图片同步;SeedASR 实时 WSS;文件 `POST /api/v3/auc/bigmodel/submit` | 文件 `POST /api/v3/auc/bigmodel/query` | 图片 `data[]`;ASR `result.text/utterances[]` |
233
- | **百炼** | `https://{WorkspaceId}.<region>.maas.aliyuncs.com` | `Authorization: Bearer sk-xxxx` | 生成/短 ASR/Qwen-MT 为同步 HTTP;实时 ASR 为 `/api-ws/v1/inference` 或 `/api-ws/v1/realtime` WSS;文件 ASR 为 `POST /api/v1/services/audio/asr/transcription` + `X-DashScope-Async: enable` | 异步 `GET /api/v1/tasks/{task_id}` | 按能力分为图像 `content[].image`、转写事件/文本、`transcription_url`、翻译 `choices[].message.content` |
234
- | **Groq** | `https://api.groq.com/openai/v1` | `Authorization: Bearer <GROQ_API_KEY>` | `POST /chat/completions`(同步/SSE);`POST /audio/transcriptions`;`GET /models` | — | Chat choices/delta;ASR `text/language/duration/segments/words` |
235
- | **硅基流动** | `https://api.siliconflow.cn/v1` | `Authorization: Bearer <API Key>` | `POST /audio/transcriptions` multipart | — | `text`;响应头 `x-siliconcloud-trace-id` |
236
- | **智谱** | 国内 `https://open.bigmodel.cn/api/paas/v4`;国际 `https://api.z.ai/api/paas/v4` | `Authorization: Bearer <API Key>`;两区凭据分离 | `POST /chat/completions`(同步/SSE) | — | `choices[].message.content/reasoning_content/tool_calls`;流式 `choices[].delta` |
237
- | **APIMart** | `https://api.apimart.ai`;大陆备用线路见[基础文档](供应商/APIMart.md) | `Authorization: Bearer <KEY>` | `POST /v1/images/generations`、`POST /v1/videos/generations` | `GET /v1/tasks/{task_id}` | `result.images[]` / `result.videos[]` |
238
- | **KIE** | `https://api.kie.ai` | `Authorization: Bearer <KEY>` | `POST /api/v1/jobs/createTask` | `GET /api/v1/jobs/recordInfo?taskId=` | `JSON.parse(resultJson).resultUrls` |
239
- | **Fal** | `https://fal.run` / `https://queue.fal.run` | `Authorization: Key $FAL_KEY` | `POST https://queue.fal.run/<endpoint-id>` | `GET .../requests/{id}/status`、`GET .../requests/{id}` | `images[]` / `video` |
240
- | **派欧云** | `https://api.ppio.com/v3`(代码仍用旧域名 `api.ppinfra.com`,待核实) | `Authorization: Bearer <KEY>` | `POST /v3/async/<route>` | `GET /v3/async/task-result?task_id=` | `images[].image_url` / `videos[].video_url` / `audios[].audio_url`(TTL 3600s) |
241
- | **魔搭** | `https://api-inference.modelscope.cn` | `Authorization: Bearer <TOKEN>` | `POST /v1/images/generations` + 头 `X-ModelScope-Async-Mode: true`(代码写的是 `/api/v1/jobs/createTask`,待核实) | `GET /v1/tasks/{task_id}` + `X-ModelScope-Task-Type: image_generation` | `output_images[]` |
242
- | **Grsai** | `https://grsaiapi.com`(全球)/ `https://grsai.dakka.com.cn`(国内直连,已接入连通性探测+自动切换) | `Authorization: Bearer <API Key>` | `POST /v1/api/generate`(`replyType` 项目侧统一强制 `async`) | `GET /v1/api/result?id=` | `results[].url` |
243
-
244
- 任务状态枚举:
245
- - APIMart:`pending` / `processing` / `completed` / `failed` / `cancelled`(Midjourney 另有 MJ 风格状态 `NOT_START` / `SUBMITTED` / `IN_PROGRESS` / `MODAL` / `SUCCESS` / `FAILURE`)
246
- - KIE:`waiting` / `queuing` / `generating` / `success` / `fail`
247
- - Fal:`IN_QUEUE` / `IN_PROGRESS` / `COMPLETED`
248
- - 百炼异步:`PENDING` / `RUNNING` / `SUCCEEDED` / `FAILED` / `CANCELED` / `UNKNOWN`
249
- - 火山文件 ASR:响应头 `20000002` 排队 / `20000001` 处理中 / `20000000` 成功;`20000003` 静音终态;`45/55*` 失败
250
- - 派欧云:`TASK_STATUS_QUEUED` / `TASK_STATUS_PROCESSING` / `TASK_STATUS_SUCCEED` / `TASK_STATUS_FAILED`
251
- - 魔搭:`SUCCEED` / `FAILED`
252
- - Grsai:`running` / `violation`(违规,独立终态)/ `succeeded` / `failed`
253
-
254
- ## 六、常用外部入口
255
-
256
- | 用途 | 链接 | 是否需登录 |
257
- |---|---|---|
258
- | APIMart 文档总索引 | https://docs.apimart.ai/llms.txt | |
259
- | APIMart 定价中心 | https://apimart.ai/zh/pricing | |
260
- | KIE 文档总索引 | https://docs.kie.ai/llms.txt | |
261
- | KIE 定价页 | https://kie.ai/pricing | 否 |
262
- | Fal 平台总览 | https://fal.ai/llms.txt | |
263
- | Fal 单模型 schema | `https://fal.ai/models/<endpoint-id>/llms.txt` | 否 |
264
- | 火山方舟图片生成 API | https://www.volcengine.com/docs/82379/1541523 | 否 |
265
- | 火山方舟模型价格 | https://www.volcengine.com/docs/82379/1544106 | 否 |
266
- | 火山 SeedASR 实时 WebSocket | https://docs.volcengine.com/docs/6561/2630027?lang=zh | |
267
- | 火山 SeedASR 文件 submit/query | https://docs.volcengine.com/docs/6561/1354868?lang=zh | 否 |
268
- | 火山 SeedASR 价格/免费额度 | https://docs.volcengine.com/docs/6561/1359369?lang=zh | |
269
- | 百炼模型价格 | https://help.aliyun.com/zh/model-studio/model-pricing | |
270
- | 百炼 ASR 模型与音频规格 | https://help.aliyun.com/zh/model-studio/asr-model/ | |
271
- | 百炼 Qwen-MT API | https://help.aliyun.com/zh/model-studio/qwen-mt-api | |
272
- | Groq 模型/价格 | https://console.groq.com/docs/models | |
273
- | Groq API 参考 | https://console.groq.com/docs/api-reference | |
274
- | Groq Reasoning | https://console.groq.com/docs/reasoning | |
275
- | Groq Speech-to-Text | https://console.groq.com/docs/speech-to-text | |
276
- | 硅基流动音频转写 API | https://api-docs.siliconflow.cn/docs/api/audio-transcriptions-post | |
277
- | 硅基流动价格 | https://siliconflow.cn/pricing | 否 |
278
- | 可灵官方视频能力地图(仅参考) | https://www.klingai.com/document-api/guides/capability-map/video | 否 |
279
- | 派欧云文档索引 | https://ppio.com/docs/llms.txt | 否 |
280
- | 派欧云定价页 | https://ppio.com/pricing | |
281
- | 魔搭 API-Inference 文档 | https://www.modelscope.cn/docs/model-service/API-Inference/intro | 否 |
282
- | 魔搭魔粒说明 | https://www.modelscope.cn/docs/magicube/intro | 否 |
283
- | APIMart API Key | https://apimart.ai/keys | **是** |
284
- | KIE API Key | https://kie.ai/api-key | **是** |
285
- | Fal API Key | https://fal.ai/dashboard/keys | **是** |
286
- | 火山方舟 API Key | https://console.volcengine.com/ark/region:cn-beijing/apiKey | **是** |
287
- | 百炼 API Key | https://bailian.console.aliyun.com/?apiKey=1#/api-key | **是** |
288
- | Groq API Key | https://console.groq.com/keys | **是** |
289
- | 硅基流动 API Key | https://cloud.siliconflow.cn/account/ak | **是** |
290
- | 智谱 GLM-5.3-Flash 模型/API | https://docs.bigmodel.cn/cn/guide/models/vlm/glm-5.3-flash | 否 |
291
- | 智谱模型价格 | https://open.bigmodel.cn/pricing | 否 |
292
- | 国际 Z.AI GLM-5.3-Flash | https://docs.z.ai/guides/vlm/glm-5.3-flash | 否 |
293
- | 国际 Z.AI 模型价格 | https://docs.z.ai/guides/overview/pricing | 否 |
294
- | Grsai 文档总索引(llms.txt) | https://qmy27nhsd9.apifox.cn/llms.txt | |
295
- | Grsai dashboard 模型列表(权威价格) | https://grsai.com/zh/dashboard/models | |
296
- | Grsai dashboard 公告(渠道调价历史) | https://grsai.com/zh/dashboard/announcements | |
297
- | Grsai API Key | https://grsai.ai/zh/dashboard/api-keys | **是** |
298
-
299
- ## 七、维护方式
300
-
301
- - **动手前**:先读 [文档采集手册.md](文档采集手册.md)——官方来源、流式事件矩阵、fixture/parser 测试闭环和 SDK 首发顺序的详细门槛只维护在这里
302
- - **新增模型**:在本文件的清单表里加一行 → 建同名文件夹 → 为每个支持它的供应商写一份 `<模型名>_<供应商名>.md`
303
- - **更新已有模型**:直接改对应的供应商文件,并更新该文件头部的「最后更新」以及本文件的「最后更新」
304
- - **信息来源要求**:每条参数 / 价格都要能追溯到文件末尾「原始链接索引」中的某个链接;需要登录才能看到的页面必须标注
41
+ ```
42
+ docs/model-adaptation/
43
+ ├── README.md # 本文件:总清单 + 索引
44
+ ├── 文档采集手册.md # 官方调研、事件契约、fixture 与 SDK 首发验证的唯一详细规范
45
+ ├── 供应商/ # 先完成供应商公共协议,再做具体模型
46
+ │ ├── 快速适配供应商.md # 新供应商的统一核对与落地清单
47
+ │ ├── APIMart.md
48
+ │ ├── KIE.md
49
+ │ ├── Fal.md
50
+ │ ├── Groq.md
51
+ │ ├── 硅基流动.md
52
+ │ ├── 百炼.md
53
+ │ ├── 火山引擎.md
54
+ │ ├── 派欧云.md
55
+ │ └── 魔搭.md
56
+ ├── <模型名>/ # 一个模型一个文件夹
57
+ │ └── <模型名>_<供应商名>.md # 支持该模型的每个供应商一个文件
58
+ ```
59
+
60
+ - **供应商名**使用项目内统一的显示名:
61
+ 派欧云 / Fal / 魔搭 / KIE / APIMart / 百炼 / 火山引擎 / 硅基流动 / Grsai / Groq(Grsai/Groq 沿用官方品牌拼写)
62
+ - 每个供应商文件都是**自包含**的:接入协议、能力清单、请求参数、响应结构、价格、适配要点、原始链接索引,看完这一份就能完成该供应商的适配
63
+ - 每份文件都标注**信息来源链接**以及**该链接是否需要登录**
64
+
65
+ ### 供应商基础文档
66
+
67
+ 模型适配前先看 [快速适配供应商](供应商/快速适配供应商.md),确认供应商公共层已经覆盖端点、鉴权、上传、任务查询、结果解析和计价。现有十家基础文档:
68
+
69
+ - [APIMart](供应商/APIMart.md):含中国大陆备用线路与图片上传协议
70
+ - [KIE](供应商/KIE.md):生成任务、文件上传、实际扣费与余额查询
71
+ - [Fal](供应商/Fal.md):同步 / 队列协议、Fal CDN、官方计价 API
72
+ - [百炼](供应商/百炼.md):地域端点、临时 OSS 上传与同步 / 异步边界
73
+ - [火山引擎](供应商/火山引擎.md):方舟图片,以及 SeedASR 2.0 实时/文件两套独立协议与计费
74
+ - [硅基流动](供应商/硅基流动.md):multipart 文件转写、免费模型与在线文档/历史公告冲突
75
+ - [派欧云](供应商/派欧云.md):异步任务协议、**无官方上传(借用 KIE)**、实名认证前置、在售清单极小
76
+ - [魔搭](供应商/魔搭.md):API-Inference 统一接口、**魔粒积分计费**、仅图片、可用性判定方法
77
+ - [Grsai](供应商/Grsai.md):聚合中转,**同一模型拆多个「渠道」**(`-cl`/`-vip`/`-vt`/`-lite` 等后缀,价差可达 10 倍以上)、无独立上传接口(图片直接 base64/URL 塞进请求体)、双线路连通性探测
78
+ - [Groq](供应商/Groq.md):OpenAI 兼容 LLM、Whisper 文件转写、动态模型发现、价格/免费层与错误边界
79
+ - [智谱 GLM](../llm-adaptation/供应商/智谱GLM.md):国内 BigModel / 国际 Z.AI 两个 endpoint profile、OpenAI 兼容 Chat 与供应商级思考约定;模型专属契约仍以本目录模型文件为准
80
+
81
+ 尚未适配的图片工具供应商候选统一记录在[图片工具候选供应商调研](供应商/候选供应商调研-图片工具.md):当前优先级为 Segmind / WaveSpeedAI(P0)、PiAPI(P1)、Replicate(P2)、Novita AI(P3);中国内地访问友好性均需真实网络验证,不能仅凭中文站或支付方式下结论。
82
+
83
+ ## 二、通用适配规则
84
+
85
+ 1. **默认绝对不显示的参数**:随机种子 `seed`、负面提示词。
86
+ 即使供应商提供了这两个字段,也**不注册、不下发**。各文件的「适配要点」会逐一标出哪个供应商有这两个字段。
87
+ 2. **`output_format` 默认不显示且不请求**,保持供应商默认值。
88
+ 3. 价格以**供应商的定价页 / 模型 `llms.txt` / 任务返回的 `cost`** 为准;部分 API 文档正文中的价格已过期,出现不一致时文件内会同时列出两处并说明取舍。
89
+ 4. 各供应商同名参数的**默认值、取值集合、字段名经常不一致**(例如 `adaptive` vs `auto`、整数 `5` vs 字符串 `"5"`),必须按供应商分别声明 schema,不要共用常量。
90
+ 5. 价格依赖本地媒体像素、时长或文件指标时,由模型在 `pricing.mediaContext` 声明需求,宿主统一读取并注入 calculator;禁止在单个页面或画布节点复制媒体计价逻辑。指标未解析完成时不显示伪造的兜底总价。
91
+
92
+ ## 三、模型清单
93
+
94
+ ### 图片(11)
95
+
96
+ | 模型 | 火山引擎 | 百炼 | APIMart | KIE | Fal | 派欧云 | 魔搭 | Grsai |
97
+ |---|---|---|---|---|---|---|---|---|
98
+ | [Seedream 5.0 Pro](Seedream-5.0-Pro/) | [✅ 官方](Seedream-5.0-Pro/Seedream-5.0-Pro_火山引擎.md) | — | [✅](Seedream-5.0-Pro/Seedream-5.0-Pro_APIMart.md) | [✅](Seedream-5.0-Pro/Seedream-5.0-Pro_KIE.md) | [✅](Seedream-5.0-Pro/Seedream-5.0-Pro_Fal.md) | ❌ 无 | ❌ 无 | ❌ 无 |
99
+ | [Seedream 5.0 Lite](Seedream-5.0-Lite/) | [✅ 官方](Seedream-5.0-Lite/Seedream-5.0-Lite_火山引擎.md) | — | [✅](Seedream-5.0-Lite/Seedream-5.0-Lite_APIMart.md) | [✅](Seedream-5.0-Lite/Seedream-5.0-Lite_KIE.md) | [✅](Seedream-5.0-Lite/Seedream-5.0-Lite_Fal.md) | ❌ 无 | ❌ 无 | ❌ 无 |
100
+ | [Qwen Image 3.0](Qwen-Image-3.0/) | — | [✅ 官方](Qwen-Image-3.0/Qwen-Image-3.0_百炼.md) | [✅](Qwen-Image-3.0/Qwen-Image-3.0_APIMart.md) | [✅](Qwen-Image-3.0/Qwen-Image-3.0_KIE.md) | [✅](Qwen-Image-3.0/Qwen-Image-3.0_Fal.md) | ❌ 无 | ❌ 无 | ❌ 无 |
101
+ | [Z-Image Turbo](Z-Image-Turbo/) | — | [✅ 官方](Z-Image-Turbo/Z-Image-Turbo_百炼.md) | [✅](Z-Image-Turbo/Z-Image-Turbo_APIMart.md) | [✅](Z-Image-Turbo/Z-Image-Turbo_KIE.md) | [✅](Z-Image-Turbo/Z-Image-Turbo_Fal.md) | ❌ 无 | [✅](Z-Image-Turbo/Z-Image-Turbo_魔搭.md) | ❌ 无 |
102
+ | [GPT-Image-2](GPT-Image-2/) | — | — | [✅](GPT-Image-2/GPT-Image-2_APIMart.md) | [✅](GPT-Image-2/GPT-Image-2_KIE.md) | [✅](GPT-Image-2/GPT-Image-2_Fal.md) | ❌ 无 | ❌ 无 | [✅](GPT-Image-2/GPT-Image-2_Grsai.md) |
103
+ | [IC-Light v2](IC-Light-v2/) | — | — | ❌ 无 | ❌ 无 | [✅](IC-Light-v2/IC-Light-v2_Fal.md) | ❌ 无 | ❌ 无 | ❌ 无 |
104
+ | [Nano Banana 2](Nano-Banana-2/) | | | [✅](Nano-Banana-2/Nano-Banana-2_APIMart.md) | [✅](Nano-Banana-2/Nano-Banana-2_KIE.md) | [✅](Nano-Banana-2/Nano-Banana-2_Fal.md) | ❌ 无 | ❌ 无 | [✅](Nano-Banana-2/Nano-Banana-2_Grsai.md) |
105
+ | [Nano Banana 2 Lite](Nano-Banana-2-Lite/) | — | — | [✅](Nano-Banana-2-Lite/Nano-Banana-2-Lite_APIMart.md) | [✅](Nano-Banana-2-Lite/Nano-Banana-2-Lite_KIE.md) | ❌ 无(已探测确认,见 [Fal 文档说明](Nano-Banana-2/Nano-Banana-2_Fal.md)) | ❌ 无 | ❌ 无 | [✅](Nano-Banana-2-Lite/Nano-Banana-2-Lite_Grsai.md) |
106
+ | [Nano Banana Pro](Nano-Banana-Pro/) | — | — | [✅](Nano-Banana-Pro/Nano-Banana-Pro_APIMart.md) | [✅](Nano-Banana-Pro/Nano-Banana-Pro_KIE.md) | [✅](Nano-Banana-Pro/Nano-Banana-Pro_Fal.md) | | ❌ 无 | [✅](Nano-Banana-Pro/Nano-Banana-Pro_Grsai.md) |
107
+ | [Midjourney](Midjourney/) | | — | [✅](Midjourney/Midjourney_APIMart.md) | | ❌ 无 | ❌ 无 | ❌ 无 | ❌ 无 |
108
+ | [Grok Imagine 2.0](Grok-Imagine-2.0/) | — | — | [✅](Grok-Imagine-2.0/Grok-Imagine-2.0_APIMart.md) | [✅](Grok-Imagine-2.0/Grok-Imagine-2.0_KIE.md) | [✅](Grok-Imagine-2.0/Grok-Imagine-2.0_Fal.md) | ❌ 无 | ❌ 无 | ❌ 无 |
109
+
110
+ **Nano Banana 2 Lite 是独立模型,不是 Nano Banana 2 的分辨率/渠道选项**:代码里两者 `canonicalModelId` 不同(`nano-banana-2` vs `nano-banana-2-lite`),`generationModelDescriptions.ts` 也各自维护一条独立描述,APIMart/KIE 均已用独立 `.model.ts` 文件实现,因此本表把它列为单独一行,不再挂在 Nano Banana 2 文档里。
111
+
112
+ ### 视频(9)
113
+
114
+ | 模型 | APIMart | KIE | Fal | 派欧云 | 魔搭 |
115
+ |---|---|---|---|---|---|
116
+ | [Seedance 2.0](Seedance-2.0/) | [✅](Seedance-2.0/Seedance-2.0_APIMart.md) | [✅](Seedance-2.0/Seedance-2.0_KIE.md) | [✅](Seedance-2.0/Seedance-2.0_Fal.md) | 已下架 | ❌ 无 |
117
+ | [Seedance 2.0 Fast](Seedance-2.0-Fast/) | [✅](Seedance-2.0-Fast/Seedance-2.0-Fast_APIMart.md) | [✅](Seedance-2.0-Fast/Seedance-2.0-Fast_KIE.md) | [✅](Seedance-2.0-Fast/Seedance-2.0-Fast_Fal.md) | 已下架 | ❌ 无 |
118
+ | [Seedance 2.0 Mini](Seedance-2.0-Mini/) | [✅](Seedance-2.0-Mini/Seedance-2.0-Mini_APIMart.md) | [✅](Seedance-2.0-Mini/Seedance-2.0-Mini_KIE.md) | [✅](Seedance-2.0-Mini/Seedance-2.0-Mini_Fal.md) | 已下架 | ❌ 无 |
119
+ | [Seedance 2.5](Seedance-2.5/) | [✅](Seedance-2.5/Seedance-2.5_APIMart.md) | [✅](Seedance-2.5/Seedance-2.5_KIE.md) | [✅](Seedance-2.5/Seedance-2.5_Fal.md) | 已下架 | ❌ 无 |
120
+ | [MiniMax H3](MiniMax-H3/) | [✅](MiniMax-H3/MiniMax-H3_APIMart.md) | [✅](MiniMax-H3/MiniMax-H3_KIE.md) | [✅](MiniMax-H3/MiniMax-H3_Fal.md) | | ❌ 无 |
121
+ | [Kling 3.0](Kling-3.0/) | [✅](Kling-3.0/Kling-3.0_APIMart.md) | [✅](Kling-3.0/Kling-3.0_KIE.md) | [✅](Kling-3.0/Kling-3.0_Fal.md) | [✅](Kling-3.0/Kling-3.0_派欧云.md) | |
122
+ | [Kling 3.0 Turbo](Kling-3.0-Turbo/) | [✅](Kling-3.0-Turbo/Kling-3.0-Turbo_APIMart.md) | [✅](Kling-3.0-Turbo/Kling-3.0-Turbo_KIE.md) | [✅](Kling-3.0-Turbo/Kling-3.0-Turbo_Fal.md) | | ❌ 无 |
123
+ | [Kling 3.0 Omni](Kling-3.0-Omni/) | [✅](Kling-3.0-Omni/Kling-3.0-Omni_APIMart.md) | [✅](Kling-3.0-Omni/Kling-3.0-Omni_KIE.md) | [✅](Kling-3.0-Omni/Kling-3.0-Omni_Fal.md) | | ❌ 无 |
124
+ | [Gemini Omni Flash](Gemini-Omni-Flash/) | [✅](Gemini-Omni-Flash/Gemini-Omni-Flash_APIMart.md) | [✅](Gemini-Omni-Flash/Gemini-Omni-Flash_KIE.md) | [✅](Gemini-Omni-Flash/Gemini-Omni-Flash_Fal.md) | | ❌ 无 |
125
+
126
+ 视频类模型在**火山引擎与百炼上未纳入本清单**:Seedance 系列的官方接口不在本次适配范围,Kling 官方接口按用户要求**仅作参考、暂不适配**(能力地图:https://www.klingai.com/document-api/guides/capability-map/video )。
127
+
128
+ Grsai 暂未在此表出现:其站内「Veo API」旧版文档给出了 `veo3.1-fast` 视频模型的端点,但该模型**未出现在 dashboard「模型列表」的定价清单中**,可用性与计价未确认,详见 [Grsai 基础文档 §8](供应商/Grsai.md)。
129
+
130
+ ### 主清单外存量模型(12)
131
+
132
+ 以下模型**不在主清单里**,但已在代码中适配并对用户开放,文档用于维护现有实现。部分模型已有多个供应商入口,未来是否纳入主清单待定。
133
+
134
+ | 模型 | 模态 | 供应商 | 文档 |
135
+ |---|---|---|---|
136
+ | MiniMax Hailuo 2.3 | 视频 | 派欧云 / KIE / Fal | [派欧云](Hailuo-2.3/Hailuo-2.3_派欧云.md) / [KIE](Hailuo-2.3/Hailuo-2.3_KIE.md) / [Fal](Hailuo-2.3/Hailuo-2.3_Fal.md) |
137
+ | MiniMax Hailuo 02 | 视频 | KIE / Fal | [KIE](Hailuo-02/Hailuo-02_KIE.md) / [Fal](Hailuo-02/Hailuo-02_Fal.md) |
138
+ | Grok Imagine | 图片 | KIE | [KIE](Grok-Imagine/Grok-Imagine_KIE.md) |
139
+ | Wan 2.5 Preview | 视频 | 派欧云 | [Wan-2.5-Preview_派欧云.md](Wan-2.5-Preview/Wan-2.5-Preview_派欧云.md) |
140
+ | Wan 2.6 | 视频 | 派欧云 | [Wan-2.6_派欧云.md](Wan-2.6/Wan-2.6_派欧云.md) |
141
+ | Wan 2.7 | 视频 | 派欧云 | [Wan-2.7_派欧云.md](Wan-2.7/Wan-2.7_派欧云.md) |
142
+ | MiniMax Speech | 音频 | 派欧云 | [MiniMax-Speech_派欧云.md](MiniMax-Speech/MiniMax-Speech_派欧云.md) |
143
+ | Topaz 图片放大 | 图片 | Fal | [Topaz-图片放大_Fal.md](Topaz-图片放大/Topaz-图片放大_Fal.md) |
144
+ | Topaz 透明图放大 | 图片 | Fal | [Topaz-透明图放大_Fal.md](Topaz-透明图放大/Topaz-透明图放大_Fal.md) |
145
+ | SeedVR2 图片放大 | 图片 | Fal | [SeedVR2-图片放大_Fal.md](SeedVR2-图片放大/SeedVR2-图片放大_Fal.md) |
146
+ | Bria 创意放大 | 图片 | Fal | [Bria-创意放大_Fal.md](Bria-创意放大/Bria-创意放大_Fal.md) |
147
+ | Ideogram 图片放大 | 图片 | Fal | [Ideogram-图片放大_Fal.md](Ideogram-图片放大/Ideogram-图片放大_Fal.md) |
148
+
149
+ 魔搭侧另有 5 个已适配的开源图像模型(`Qwen/Qwen-Image`、`Qwen/Qwen-Image-Edit-2509`、`black-forest-labs/FLUX.1-Krea-dev`、`MusePublic/majicMIX_realistic`、`MusePublic/14_ckpt_SD_XL`)。它们共用魔搭 API-Inference 的同一套端点与参数,差异只有 model ID、尺寸上限和魔粒档位,因此不单独建文件,统一记在 [供应商/魔搭.md](供应商/魔搭.md) 7 节。
150
+
151
+ ### SDK / Say-It 语音识别(15)
152
+
153
+ 以下 15 个模型均有独立 SDK capability descriptor 与按需执行模块;消费应用仍需显式导入和注册,升级包版本不会自动增加界面入口。火山文件版只接受供应商可访问的公网 URL,Say-It 在没有受管上传/TOS 之前不得把它伪装成本地文件入口。
154
+
155
+ | 模型 | 模式 | 主协议 | 上限 | 文档 |
156
+ |---|---|---|---|---|
157
+ | `fun-asr-realtime-2026-02-28` | 实时 | WebSocket `run-task` | 时长无限制 | [百炼](Fun-ASR-Realtime-2026-02-28/Fun-ASR-Realtime-2026-02-28_百炼.md) |
158
+ | `fun-asr-realtime` | 实时稳定别名 | WebSocket `run-task` | 时长无限制 | [百炼](Fun-ASR-Realtime/Fun-ASR-Realtime_百炼.md) |
159
+ | `qwen3-asr-flash-realtime-2026-02-10` | 实时 | WebSocket session | 时长无限制 | [百炼](Qwen3-ASR-Flash-Realtime-2026-02-10/Qwen3-ASR-Flash-Realtime-2026-02-10_百炼.md) |
160
+ | `qwen3-asr-flash-realtime` | 实时稳定别名 | WebSocket session | 时长无限制 | [百炼](Qwen3-ASR-Flash-Realtime/Qwen3-ASR-Flash-Realtime_百炼.md) |
161
+ | `fun-asr-flash-2026-06-15` | 短音频 | DashScope HTTP/SSE 同步 | 5 分钟 / 2 GB | [百炼](Fun-ASR-Flash-2026-06-15/Fun-ASR-Flash-2026-06-15_百炼.md) |
162
+ | `qwen3-asr-flash` | 短音频稳定别名 | OpenAI 兼容 / DashScope 同步 | 5 分钟 / 10 MB | [百炼](Qwen3-ASR-Flash/Qwen3-ASR-Flash_百炼.md) |
163
+ | `qwen3-asr-flash-2026-02-10` | 短音频快照 | OpenAI 兼容 / DashScope 同步 | 5 分钟 / 10 MB | [百炼](Qwen3-ASR-Flash-2026-02-10/Qwen3-ASR-Flash-2026-02-10_百炼.md) |
164
+ | `fun-asr` | 文件异步 | 提交 + 轮询 + 结果 JSON | 12 小时 / 2 GB | [百炼](Fun-ASR/Fun-ASR_百炼.md) |
165
+ | `qwen3-asr-flash-filetrans` | 文件异步 | 提交 + 轮询 | 12 小时 / 2 GB | [百炼](Qwen3-ASR-Flash-Filetrans/Qwen3-ASR-Flash-Filetrans_百炼.md) |
166
+ | `seedasr-2.0-realtime` | 实时 | 火山二进制 WebSocket | PCM S16LE / 16kHz / mono | [火山引擎](SeedASR-2.0-Realtime/SeedASR-2.0-Realtime_火山引擎.md) |
167
+ | `seedasr-2.0-file` | 文件异步 | HTTP submit + query | 仅公网 URL;本地文件需外部 TOS/受管上传 | [火山引擎](SeedASR-2.0-File/SeedASR-2.0-File_火山引擎.md) |
168
+ | `FunAudioLLM/SenseVoiceSmall` | 文件同步 | multipart HTTP | 1 小时 / 50 MB | [硅基流动](SenseVoiceSmall/SenseVoiceSmall_硅基流动.md) |
169
+ | `TeleAI/TeleSpeechASR` | 文件同步 | multipart HTTP | 1 小时 / 50 MB | [硅基流动](TeleSpeechASR/TeleSpeechASR_硅基流动.md) |
170
+ | `whisper-large-v3-turbo` | 文件同步 | OpenAI-compatible multipart | 免费层 25 MB;Developer 100 MB | [Groq](Whisper-Large-v3-Turbo/Whisper-Large-v3-Turbo_Groq.md) |
171
+ | `whisper-large-v3` | 文件同步 | OpenAI-compatible multipart | 免费层 25 MB;Developer 100 MB | [Groq](Whisper-Large-v3/Whisper-Large-v3_Groq.md) |
172
+
173
+ ### Say-It 翻译(3)
174
+
175
+ | 模型 | 流式语义 | 北京原价(输入/输出,每百万 Token) | 文档 |
176
+ |---|---|---|---|
177
+ | `qwen-mt-flash` | 增量 | 0.7 / 1.95 元 | [百炼](Qwen-MT-Flash/Qwen-MT-Flash_百炼.md) |
178
+ | `qwen-mt-plus` | 累积序列 | 1.8 / 5.4 元 | [百炼](Qwen-MT-Plus/Qwen-MT-Plus_百炼.md) |
179
+ | `qwen-mt-lite` | 增量 | 0.6 / 1.6 元 | [百炼](Qwen-MT-Lite/Qwen-MT-Lite_百炼.md) |
180
+
181
+ ### SDK LLM(2)
182
+
183
+ | 模型 | 供应商 | 备注 | 文档 |
184
+ |---|---|---|---|
185
+ | `openai/gpt-oss-20b` | Groq | Say-It 默认;`reasoning_effort=low/medium/high`;不支持 `reasoning_format` | [Groq](GPT-OSS-20B/GPT-OSS-20B_Groq.md) |
186
+ | `glm-5.3-flash` | 智谱 | 原生多模态 Chat 模型;1M 上下文 / 128K 输出;完整能力仍有官方冲突项 | [智谱](GLM-5.3-Flash/GLM-5.3-Flash_智谱.md) |
187
+
188
+ ## 四、平台模型 ID 速查
189
+
190
+ ### 可选工具模型
191
+
192
+ | 工具 | Fal endpoint ID | SDK 入口 |
193
+ |---|---|---|
194
+ | Flux Pro Erase | `fal-ai/flux-pro/v1/erase` | `tool-models/fal/flux-pro-erase` |
195
+ | Bria Eraser | `fal-ai/bria/eraser` | `tool-models/fal/bria-eraser` |
196
+ | Finegrain Eraser (Mask) | `fal-ai/finegrain-eraser/mask` | `tool-models/fal/finegrain-eraser` |
197
+ | Qwen Image Edit 2511 多角度 | `fal-ai/qwen-image-edit-2511-multiple-angles` | `@henjicc/ai-sdk/tool-models/fal/qwen-image-edit-2511-multiple-angles` |
198
+ | 透视变换 | `fal-ai/image-apps-v2/perspective` | `@henjicc/ai-sdk/tool-models/fal/perspective-change` |
199
+ | Image Apps v2 重打光 | `fal-ai/image-apps-v2/relighting` | `tool-models/fal/relighting` |
200
+ | ControlLight | `fal-ai/control-light` | `tool-models/fal/control-light` |
201
+ | Image Apps v2 扩图 | `fal-ai/image-apps-v2/outpaint` | `tool-models/fal/outpaint` |
202
+ | Image Apps v2 商品摄影 | `fal-ai/image-apps-v2/product-photography` | `tool-models/fal/product-photography` |
203
+ | Image Apps v2 照片修复 | `fal-ai/image-apps-v2/photo-restoration` | `tool-models/fal/photo-restoration` |
204
+ | Pixelcut 背景移除 | `pixelcut/background-removal` | `tool-models/fal/pixelcut-background-removal` |
205
+ | FLUX 2 多角度 | `fal-ai/flux-2-lora-gallery/multiple-angles` | `tool-models/fal/flux-2-multiple-angles` |
206
+
207
+ - 3 个消除工具聚合为 `@henjicc/ai-sdk/tool-packs/fal-image-edit-tools`,统一能力画像为 `operation=image-edit`、`feature=erase`。
208
+ - 6 个重打光/修复/扩图/商品/抠图工具聚合为 `@henjicc/ai-sdk/tool-packs/fal-image-utility-tools`,再按单模型画像区分具体能力。
209
+ - Qwen 2509、透视变换和 FLUX 2 聚合为 `@henjicc/ai-sdk/tool-packs/fal-multi-angle-tools`,与消除和实用工具 pack 分开分发。
210
+
211
+ 三个 pack 都只是静态分发集合,执行继续走同一生成客户端和 Fal 公共协议。
212
+
213
+ | 模型 | APIMart | KIE | Fal |
214
+ |---|---|---|---|
215
+ | Seedream 5.0 Pro | `seedream-5-0-pro` | `seedream/5-pro-text-to-image`、`5-pro-image-to-image`、`5-pro-layer-decomposition` | `bytedance/seedream/v5/pro/{text-to-image,edit}` |
216
+ | Seedream 5.0 Lite | `seedream-5-0-lite` | `seedream/5-lite-text-to-image`、`5-lite-image-to-image` | `bytedance/seedream/v5/lite/{text-to-image,edit}` |
217
+ | Qwen Image 3.0 | `qwen-image-3.0`、`qwen-image-3.0-pro` | `qwen3/{text-to-image,image-to-image}`、`qwen3-pro/text-to-image`、`qwen3/pro-image-to-image` | `alibaba/qwen-image-3/{text-to-image,edit}` |
218
+ | Z-Image Turbo | `z-image-turbo` | `z-image` | `fal-ai/z-image/turbo`、`fal-ai/z-image/turbo/image-to-image`;魔搭 `Tongyi-MAI/Z-Image-Turbo` |
219
+ | GPT-Image-2 | `gpt-image-2`(别名 `gpt-image-2-ext`)、`gpt-image-2-official` | `gpt-image-2-text-to-image`、`gpt-image-2-image-to-image` | `openai/gpt-image-2`、`openai/gpt-image-2/edit` |
220
+ | IC-Light v2 | — | — | `fal-ai/iclight-v2` |
221
+ | Topaz 图片放大 | — | — | `topaz/upscale/image/{precision,creative,generative}` |
222
+ | Topaz 透明图放大 | — | — | `topaz/upscale/image/transparent` |
223
+ | SeedVR2 图片放大 | — | — | `fal-ai/seedvr/upscale/image` |
224
+ | Bria 创意放大 | — | — | `bria/upscale/creative` |
225
+ | Ideogram 图片放大 | — | — | `fal-ai/ideogram/upscale` |
226
+ | Nano Banana 2 | `gemini-3.1-flash-image-preview`(-official) | `nano-banana-2` | `fal-ai/nano-banana-2`、`fal-ai/nano-banana-2/edit` |
227
+ | Nano Banana 2 Lite | `gemini-3.1-flash-lite-image`(-ext) | `nano-banana-2-lite` | ❌ 无(已探测确认) |
228
+ | Nano Banana Pro | `gemini-3-pro-image-preview`(-official) | `nano-banana-pro` | `fal-ai/nano-banana-pro`、`fal-ai/nano-banana-pro/edit` |
229
+ | Midjourney | `midjourney`(Imagine)、`midjourney-blend`、`midjourney-edit`、`midjourney-video`(新路由均自动注入平台模型 ID) | — | — |
230
+ | Grok Imagine 2.0 | `grok-imagine-2.0-ext`、`grok-imagine-image-2.0` | `grok-imagine-image-2-0/{text-to-image,image-edit,segment-map,segment-edit}` | `xai/grok-imagine-image/v2.0/{text-to-image,edit}` |
231
+ | Grok Imagine | — | `grok-imagine/{text-to-image,image-to-image}` | — |
232
+ | Seedance 2.0 | `seedance-2.0` | `bytedance/seedance-2` | `bytedance/seedance-2.0/{text-to-video,image-to-video,reference-to-video}` |
233
+ | Seedance 2.0 Fast | `seedance-2.0-fast` | `bytedance/seedance-2-fast` | `bytedance/seedance-2.0/fast/...` |
234
+ | Seedance 2.0 Mini | `seedance-2.0-mini` | `bytedance/seedance-2-mini` | `bytedance/seedance-2.0/mini/...` |
235
+ | Seedance 2.5 | `seedance-2.5` | `bytedance/seedance-2-5` | `bytedance/seedance-2.5/...` |
236
+ | MiniMax H3 | `MiniMax-H3`、`MiniMax-H3-Context-IR`、`MiniMax-H3-Regeneration` | `minimax-h3/{text-to-video,image-to-video,reference-to-video}` | `minimax/h3/{text-to-video,image-to-video,reference-to-video}` |
237
+ | MiniMax Hailuo 2.3 | — | `hailuo/2-3-image-to-video-{standard,pro}` | `fal-ai/minimax/hailuo-2.3/{standard,pro}/{text-to-video,image-to-video}`、`fal-ai/minimax/hailuo-2.3-fast/{standard,pro}/image-to-video` |
238
+ | MiniMax Hailuo 02 | — | `hailuo/02-{text,image}-to-video-{standard,pro}` | `fal-ai/minimax/hailuo-02/{standard,pro}/{text-to-video,image-to-video}`、`fal-ai/minimax/hailuo-02-fast/image-to-video` |
239
+ | Kling 3.0 | `kling-v3` | `kling-3.0/video` | `fal-ai/kling-video/v3/{pro,standard}/...`;派欧云 `/v3/async/kling-v3.0-{std,pro,4k}-{t2v,i2v}` |
240
+ | Kling 3.0 Turbo | `kling-3.0-turbo` | `kling/v3-turbo-{text,image}-to-video` | `fal-ai/kling-video/v3/turbo/{pro,standard}/...` |
241
+ | Kling 3.0 Omni | `kling-v3-omni` | `kling-3.0-omni/{text-to-video,image-to-video,reference-to-video,transformation}` | `fal-ai/kling-video/o3/{pro,standard}/...` |
242
+ | Gemini Omni Flash | `gemini-omni-flash-preview`、`Omni-Flash-Ext` | `gemini-omni-video` + `/api/v1/omni/audio/create` + `/api/v1/omni/character/create` | `google/gemini-omni-flash/{image-to-video,reference-to-video}` |
243
+
244
+ Grsai 的平台模型 ID 未并入本表:它把「渠道」直接编码进 `model` 字段(如 `nano-banana-2-cl`、`gpt-image-2-vip`),渠道数量比 APIMart/KIE/Fal 常见的 2 档更多(GPT-Image-2 2 个、Nano Banana 2 主模型 4 个、Nano Banana 2 Lite 1 个、Nano Banana Pro 5 个),完整枚举见对应模型文档:[GPT-Image-2_Grsai.md](GPT-Image-2/GPT-Image-2_Grsai.md)、[Nano-Banana-2_Grsai.md](Nano-Banana-2/Nano-Banana-2_Grsai.md)、[Nano-Banana-2-Lite_Grsai.md](Nano-Banana-2-Lite/Nano-Banana-2-Lite_Grsai.md)、[Nano-Banana-Pro_Grsai.md](Nano-Banana-Pro/Nano-Banana-Pro_Grsai.md)。
245
+
246
+ ## 五、供应商通用协议速查
247
+
248
+ | 供应商 | Base URL | 鉴权 | 提交 | 查询 | 结果字段 |
249
+ |---|---|---|---|---|---|
250
+ | **火山引擎** | 方舟 `https://ark.cn-beijing.volces.com`;语音 `https://openspeech.bytedance.com` | 方舟 Bearer;语音 `X-Api-Key` + Resource/Request ID | 图片同步;SeedASR 实时 WSS;文件 `POST /api/v3/auc/bigmodel/submit` | 文件 `POST /api/v3/auc/bigmodel/query` | 图片 `data[]`;ASR `result.text/utterances[]` |
251
+ | **百炼** | `https://{WorkspaceId}.<region>.maas.aliyuncs.com` | `Authorization: Bearer sk-xxxx` | 生成/短 ASR/Qwen-MT 为同步 HTTP;实时 ASR 为 `/api-ws/v1/inference` 或 `/api-ws/v1/realtime` WSS;文件 ASR 为 `POST /api/v1/services/audio/asr/transcription` + `X-DashScope-Async: enable` | 异步 `GET /api/v1/tasks/{task_id}` | 按能力分为图像 `content[].image`、转写事件/文本、`transcription_url`、翻译 `choices[].message.content` |
252
+ | **Groq** | `https://api.groq.com/openai/v1` | `Authorization: Bearer <GROQ_API_KEY>` | `POST /chat/completions`(同步/SSE);`POST /audio/transcriptions`;`GET /models` | — | Chat choices/delta;ASR `text/language/duration/segments/words` |
253
+ | **硅基流动** | `https://api.siliconflow.cn/v1` | `Authorization: Bearer <API Key>` | `POST /audio/transcriptions` multipart | — | `text`;响应头 `x-siliconcloud-trace-id` |
254
+ | **智谱** | 国内 `https://open.bigmodel.cn/api/paas/v4`;国际 `https://api.z.ai/api/paas/v4` | `Authorization: Bearer <API Key>`;两区凭据分离 | `POST /chat/completions`(同步/SSE) | — | `choices[].message.content/reasoning_content/tool_calls`;流式 `choices[].delta` |
255
+ | **APIMart** | `https://api.apimart.ai`;大陆备用线路见[基础文档](供应商/APIMart.md) | `Authorization: Bearer <KEY>` | `POST /v1/images/generations`、`POST /v1/videos/generations` | `GET /v1/tasks/{task_id}` | `result.images[]` / `result.videos[]` |
256
+ | **KIE** | `https://api.kie.ai` | `Authorization: Bearer <KEY>` | `POST /api/v1/jobs/createTask` | `GET /api/v1/jobs/recordInfo?taskId=` | `JSON.parse(resultJson).resultUrls` |
257
+ | **Fal** | `https://fal.run` / `https://queue.fal.run` | `Authorization: Key $FAL_KEY` | `POST https://queue.fal.run/<endpoint-id>` | `GET .../requests/{id}/status`、`GET .../requests/{id}` | `images[]` / `video` |
258
+ | **派欧云** | `https://api.ppio.com/v3`(代码仍用旧域名 `api.ppinfra.com`,待核实) | `Authorization: Bearer <KEY>` | `POST /v3/async/<route>` | `GET /v3/async/task-result?task_id=` | `images[].image_url` / `videos[].video_url` / `audios[].audio_url`(TTL 3600s) |
259
+ | **魔搭** | `https://api-inference.modelscope.cn` | `Authorization: Bearer <TOKEN>` | `POST /v1/images/generations` + 头 `X-ModelScope-Async-Mode: true`(代码写的是 `/api/v1/jobs/createTask`,待核实) | `GET /v1/tasks/{task_id}` + 头 `X-ModelScope-Task-Type: image_generation` | `output_images[]` |
260
+ | **Grsai** | `https://grsaiapi.com`(全球)/ `https://grsai.dakka.com.cn`(国内直连,已接入连通性探测+自动切换) | `Authorization: Bearer <API Key>` | `POST /v1/api/generate`(`replyType` 项目侧统一强制 `async`) | `GET /v1/api/result?id=` | `results[].url` |
261
+
262
+ 任务状态枚举:
263
+ - APIMart:`pending` / `processing` / `completed` / `failed` / `cancelled`(Midjourney 另有 MJ 风格状态 `NOT_START` / `SUBMITTED` / `IN_PROGRESS` / `MODAL` / `SUCCESS` / `FAILURE`)
264
+ - KIE:`waiting` / `queuing` / `generating` / `success` / `fail`
265
+ - Fal:`IN_QUEUE` / `IN_PROGRESS` / `COMPLETED`
266
+ - 百炼异步:`PENDING` / `RUNNING` / `SUCCEEDED` / `FAILED` / `CANCELED` / `UNKNOWN`
267
+ - 火山文件 ASR:响应头 `20000002` 排队 / `20000001` 处理中 / `20000000` 成功;`20000003` 静音终态;`45/55*` 失败
268
+ - 派欧云:`TASK_STATUS_QUEUED` / `TASK_STATUS_PROCESSING` / `TASK_STATUS_SUCCEED` / `TASK_STATUS_FAILED`
269
+ - 魔搭:`SUCCEED` / `FAILED`
270
+ - Grsai:`running` / `violation`(违规,独立终态)/ `succeeded` / `failed`
271
+
272
+ ## 六、常用外部入口
273
+
274
+ | 用途 | 链接 | 是否需登录 |
275
+ |---|---|---|
276
+ | APIMart 文档总索引 | https://docs.apimart.ai/llms.txt | 否 |
277
+ | APIMart 定价中心 | https://apimart.ai/zh/pricing | 否 |
278
+ | KIE 文档总索引 | https://docs.kie.ai/llms.txt | 否 |
279
+ | KIE 定价页 | https://kie.ai/pricing | 否 |
280
+ | Fal 平台总览 | https://fal.ai/llms.txt | 否 |
281
+ | Fal 单模型 schema | `https://fal.ai/models/<endpoint-id>/llms.txt` | 否 |
282
+ | 火山方舟图片生成 API | https://www.volcengine.com/docs/82379/1541523 | 否 |
283
+ | 火山方舟模型价格 | https://www.volcengine.com/docs/82379/1544106 | 否 |
284
+ | 火山 SeedASR 实时 WebSocket | https://docs.volcengine.com/docs/6561/2630027?lang=zh | 否 |
285
+ | 火山 SeedASR 文件 submit/query | https://docs.volcengine.com/docs/6561/1354868?lang=zh | 否 |
286
+ | 火山 SeedASR 价格/免费额度 | https://docs.volcengine.com/docs/6561/1359369?lang=zh | 否 |
287
+ | 百炼模型价格 | https://help.aliyun.com/zh/model-studio/model-pricing | 否 |
288
+ | 百炼 ASR 模型与音频规格 | https://help.aliyun.com/zh/model-studio/asr-model/ | 否 |
289
+ | 百炼 Qwen-MT API | https://help.aliyun.com/zh/model-studio/qwen-mt-api | 否 |
290
+ | Groq 模型/价格 | https://console.groq.com/docs/models | 否 |
291
+ | Groq API 参考 | https://console.groq.com/docs/api-reference | 否 |
292
+ | Groq Reasoning | https://console.groq.com/docs/reasoning | 否 |
293
+ | Groq Speech-to-Text | https://console.groq.com/docs/speech-to-text | 否 |
294
+ | 硅基流动音频转写 API | https://api-docs.siliconflow.cn/docs/api/audio-transcriptions-post | 否 |
295
+ | 硅基流动价格 | https://siliconflow.cn/pricing | 否 |
296
+ | 可灵官方视频能力地图(仅参考) | https://www.klingai.com/document-api/guides/capability-map/video | 否 |
297
+ | 派欧云文档索引 | https://ppio.com/docs/llms.txt | 否 |
298
+ | 派欧云定价页 | https://ppio.com/pricing | 否 |
299
+ | 魔搭 API-Inference 文档 | https://www.modelscope.cn/docs/model-service/API-Inference/intro | 否 |
300
+ | 魔搭魔粒说明 | https://www.modelscope.cn/docs/magicube/intro | 否 |
301
+ | APIMart API Key | https://apimart.ai/keys | **是** |
302
+ | KIE API Key | https://kie.ai/api-key | **是** |
303
+ | Fal API Key | https://fal.ai/dashboard/keys | **是** |
304
+ | 火山方舟 API Key | https://console.volcengine.com/ark/region:cn-beijing/apiKey | **是** |
305
+ | 百炼 API Key | https://bailian.console.aliyun.com/?apiKey=1#/api-key | **是** |
306
+ | Groq API Key | https://console.groq.com/keys | **是** |
307
+ | 硅基流动 API Key | https://cloud.siliconflow.cn/account/ak | **是** |
308
+ | 智谱 GLM-5.3-Flash 模型/API | https://docs.bigmodel.cn/cn/guide/models/vlm/glm-5.3-flash | 否 |
309
+ | 智谱模型价格 | https://open.bigmodel.cn/pricing | 否 |
310
+ | 国际 Z.AI GLM-5.3-Flash | https://docs.z.ai/guides/vlm/glm-5.3-flash | 否 |
311
+ | 国际 Z.AI 模型价格 | https://docs.z.ai/guides/overview/pricing | 否 |
312
+ | Grsai 文档总索引(llms.txt) | https://qmy27nhsd9.apifox.cn/llms.txt | 否 |
313
+ | Grsai dashboard 模型列表(权威价格) | https://grsai.com/zh/dashboard/models | 否 |
314
+ | Grsai dashboard 公告(渠道调价历史) | https://grsai.com/zh/dashboard/announcements | 否 |
315
+ | Grsai API Key | https://grsai.ai/zh/dashboard/api-keys | **是** |
316
+
317
+ ## 七、维护方式
318
+
319
+ - **动手前**:先读 [文档采集手册.md](文档采集手册.md)——官方来源、流式事件矩阵、fixture/parser 测试闭环和 SDK 首发顺序的详细门槛只维护在这里
320
+ - **新增模型**:在本文件的清单表里加一行 → 建同名文件夹 → 为每个支持它的供应商写一份 `<模型名>_<供应商名>.md`
321
+ - **更新已有模型**:直接改对应的供应商文件,并更新该文件头部的「最后更新」以及本文件的「最后更新」
322
+ - **信息来源要求**:每条参数 / 价格都要能追溯到文件末尾「原始链接索引」中的某个链接;需要登录才能看到的页面必须标注