|
| 1 | +import type { AgentModelConfig } from "./types"; |
| 2 | + |
| 3 | +// 通过模型 ID 字符串检测是否支持视觉输入 |
| 4 | +export function supportsVisionByModelId(modelId: string): boolean { |
| 5 | + const m = modelId.toLowerCase(); |
| 6 | + |
| 7 | + // OpenAI 视觉模型 |
| 8 | + if (m.includes("gpt-4o") || m.includes("gpt-4-turbo") || m.includes("gpt-4-vision")) return true; |
| 9 | + if (m.startsWith("o1") || m.startsWith("o3") || m.startsWith("o4")) return true; |
| 10 | + |
| 11 | + // Anthropic Claude 3+ 全部支持视觉 |
| 12 | + if ( |
| 13 | + m.startsWith("claude-3") || |
| 14 | + m.startsWith("claude-sonnet") || |
| 15 | + m.startsWith("claude-opus") || |
| 16 | + m.startsWith("claude-haiku") |
| 17 | + ) |
| 18 | + return true; |
| 19 | + |
| 20 | + // Google Gemini 基本都支持视觉 |
| 21 | + if (m.startsWith("gemini")) return true; |
| 22 | + |
| 23 | + // Grok 视觉 |
| 24 | + if (m.includes("grok") && m.includes("vision")) return true; |
| 25 | + |
| 26 | + // Qwen-VL |
| 27 | + if (m.includes("qwen") && (m.includes("vl") || m.includes("vision"))) return true; |
| 28 | + |
| 29 | + // GLM-4V |
| 30 | + if (m.includes("glm") && m.includes("v")) return true; |
| 31 | + |
| 32 | + // Pixtral (Mistral 视觉模型) |
| 33 | + if (m.startsWith("pixtral")) return true; |
| 34 | + |
| 35 | + // DeepSeek-VL |
| 36 | + if (m.includes("deepseek") && m.includes("vl")) return true; |
| 37 | + |
| 38 | + // Llama 视觉 |
| 39 | + if (m.includes("llama") && (m.includes("vision") || m.includes("scout"))) return true; |
| 40 | + |
| 41 | + return false; |
| 42 | +} |
| 43 | + |
| 44 | +// 检测模型是否支持视觉输入(用户手动设置优先于自动检测) |
| 45 | +export function supportsVision(model: AgentModelConfig): boolean { |
| 46 | + if (model.supportsVision !== undefined) return model.supportsVision; |
| 47 | + return supportsVisionByModelId(model.model); |
| 48 | +} |
| 49 | + |
| 50 | +// 通过模型 ID 字符串检测是否支持图片输出 |
| 51 | +export function supportsImageOutputByModelId(modelId: string): boolean { |
| 52 | + const m = modelId.toLowerCase(); |
| 53 | + // GPT-4o 系列支持图片生成(不含 mini/audio) |
| 54 | + if (m.includes("gpt-4o") && !m.includes("mini") && !m.includes("audio")) return true; |
| 55 | + // Gemini 2.0 Flash 支持原生图片生成(不含 1.5 等旧版本) |
| 56 | + if (m.includes("gemini-2") && m.includes("flash") && !m.includes("lite")) return true; |
| 57 | + // Gemini 3+ 带 image 标识的模型支持图片生成 |
| 58 | + if (m.includes("gemini-") && m.includes("image")) return true; |
| 59 | + // DALL-E |
| 60 | + if (m.startsWith("dall-e")) return true; |
| 61 | + return false; |
| 62 | +} |
| 63 | + |
| 64 | +// 检测模型是否支持图片输出(用户手动设置优先于自动检测) |
| 65 | +export function supportsImageOutput(model: AgentModelConfig): boolean { |
| 66 | + if (model.supportsImageOutput !== undefined) return model.supportsImageOutput; |
| 67 | + return supportsImageOutputByModelId(model.model); |
| 68 | +} |
0 commit comments