支持在会话输入框直接进行文件/图片的拖拽和粘贴

src/lib/adapters/ai-elements-adapter.ts

@@ -43,6 +43,13 @@ export interface UserResourceDisplay {
   mime_type?: string | null
 }
 
+export interface UserImageDisplay {
+  name: string
+  data: string
+  mime_type: string
+  uri?: string | null
+}
+
 const BLOCKED_RESOURCE_MENTION_RE = /@([^\s@]+)\s*\[blocked[^\]]*\]/gi
 const MARKDOWN_LINK_RE = /\[([^\]]+)\]\(([^)]+)\)/g
 
@@ -54,6 +61,7 @@ export interface AdaptedMessage {
   role: MessageRole
   content: AdaptedContentPart[]
   userResources?: UserResourceDisplay[]
+  userImages?: UserImageDisplay[]
   timestamp: string
   usage?: TurnUsage | null
   duration_ms?: number | null
@@ -398,6 +406,20 @@ function addResource(
   resources.push(resource)
 }
 
+function addImage(images: UserImageDisplay[], image: UserImageDisplay) {
+  const key = `${image.mime_type}:${image.data.length}:${image.data.slice(0, 64)}`
+  if (
+    images.some(
+      (item) =>
+        `${item.mime_type}:${item.data.length}:${item.data.slice(0, 64)}` ===
+        key
+    )
+  ) {
+    return
+  }
+  images.push(image)
+}
+
 export function extractUserResourcesFromText(text: string): {
   text: string
   resources: UserResourceDisplay[]
@@ -480,6 +502,33 @@ function splitUserTextAndResources(
   return { parts: nextParts, resources }
 }
 
+function deriveImageNameFromBlock(
+  block: Extract<ContentBlock, { type: "image" }>
+): string {
+  if (block.uri && block.uri.trim().length > 0) {
+    return fileNameFromUri(block.uri)
+  }
+  const ext = block.mime_type.split("/")[1]?.split("+")[0] ?? "image"
+  return `image.${ext}`
+}
+
+function extractUserImagesFromBlocks(
+  blocks: ContentBlock[]
+): UserImageDisplay[] {
+  const images: UserImageDisplay[] = []
+  for (const block of blocks) {
+    if (block.type !== "image") continue
+    if (!block.data || !block.mime_type) continue
+    addImage(images, {
+      name: deriveImageNameFromBlock(block),
+      data: block.data,
+      mime_type: block.mime_type,
+      uri: block.uri ?? null,
+    })
+  }
+  return images
+}
+
 /**
  * Generate a stable tool call ID based on message ID and block index
  */
@@ -661,6 +710,8 @@ export function adaptMessageTurn(
     turn.role === "user"
       ? splitUserTextAndResources(adaptedContent, text.attachedResources)
       : { parts: adaptedContent, resources: [] as UserResourceDisplay[] }
+  const userImages =
+    turn.role === "user" ? extractUserImagesFromBlocks(turn.blocks) : []
 
   return {
     id: turn.id,
@@ -668,6 +719,7 @@ export function adaptMessageTurn(
     content: userSplit.parts,
     userResources:
       userSplit.resources.length > 0 ? userSplit.resources : undefined,
+    userImages: userImages.length > 0 ? userImages : undefined,
     timestamp: turn.timestamp,
     usage: turn.usage,
     duration_ms: turn.duration_ms,
@@ -695,6 +747,7 @@ export interface MessageGroup {
   role: "user" | "assistant" | "system"
   parts: AdaptedContentPart[]
   userResources?: UserResourceDisplay[]
+  userImages?: UserImageDisplay[]
   usage?: TurnUsage | null
   duration_ms?: number | null
   model?: string | null
@@ -738,6 +791,7 @@ export function groupAdaptedMessages(
         role: effectiveRole,
         parts: [...msg.content],
         userResources: msg.userResources,
+        userImages: msg.userImages,
       })
     } else {
       if (currentGroup && currentGroup.role === "assistant") {

src/lib/prompt-draft.ts

@@ -1,5 +1,6 @@
 import type {
   AdaptedContentPart,
+  UserImageDisplay,
   UserResourceDisplay,
 } from "@/lib/adapters/ai-elements-adapter"
 import type { PromptDraft, PromptInputBlock } from "@/lib/types"
@@ -10,6 +11,35 @@ function isResourceLinkBlock(
   return block.type === "resource_link"
 }
 
+function isEmbeddedResourceBlock(
+  block: PromptInputBlock
+): block is Extract<PromptInputBlock, { type: "resource" }> {
+  return block.type === "resource"
+}
+
+function isImageBlock(
+  block: PromptInputBlock
+): block is Extract<PromptInputBlock, { type: "image" }> {
+  return block.type === "image"
+}
+
+function deriveResourceNameFromUri(uri: string): string {
+  const fallback = "resource"
+  const normalized = uri.trim()
+  if (!normalized) return fallback
+  const withoutQuery = normalized.split(/[?#]/, 1)[0]
+  const candidate = withoutQuery.split(/[\\/]/).pop() ?? ""
+  let decoded = ""
+  if (candidate) {
+    try {
+      decoded = decodeURIComponent(candidate)
+    } catch {
+      decoded = candidate
+    }
+  }
+  return decoded || fallback
+}
+
 export function getPromptDraftDisplayText(
   draft: PromptDraft,
   attachedResourcesFallback: string
@@ -33,9 +63,40 @@ export function buildUserMessageTextPartsFromDraft(
 export function extractUserResourcesFromDraft(
   draft: PromptDraft
 ): UserResourceDisplay[] {
-  return draft.blocks.filter(isResourceLinkBlock).map((resource) => ({
+  const linked = draft.blocks.filter(isResourceLinkBlock).map((resource) => ({
     name: resource.name,
     uri: resource.uri,
     mime_type: resource.mime_type ?? null,
   }))
+  const embedded = draft.blocks
+    .filter(isEmbeddedResourceBlock)
+    .map((resource) => ({
+      name: deriveResourceNameFromUri(resource.uri),
+      uri: resource.uri,
+      mime_type: resource.mime_type ?? null,
+    }))
+  return [...linked, ...embedded]
+}
+
+function deriveImageName(
+  uri: string | null | undefined,
+  mimeType: string
+): string {
+  if (uri && uri.trim().length > 0) {
+    const name = deriveResourceNameFromUri(uri)
+    if (name !== "resource") return name
+  }
+  const ext = mimeType.split("/")[1]?.split("+")[0] ?? "image"
+  return `image.${ext}`
+}
+
+export function extractUserImagesFromDraft(
+  draft: PromptDraft
+): UserImageDisplay[] {
+  return draft.blocks.filter(isImageBlock).map((image) => ({
+    name: deriveImageName(image.uri, image.mime_type),
+    data: image.data,
+    mime_type: image.mime_type,
+    uri: image.uri ?? null,
+  }))
 }

src/lib/tauri.ts

@@ -743,6 +743,13 @@ export async function stopFileTreeWatch(rootPath: string): Promise<void> {
   return invoke("stop_file_tree_watch", { rootPath })
 }
 
+export async function readFileBase64(
+  path: string,
+  maxBytes?: number
+): Promise<string> {
+  return invoke("read_file_base64", { path, maxBytes: maxBytes ?? null })
+}
+
 export async function readFilePreview(
   rootPath: string,
   path: string,

src/lib/types.ts

@@ -60,6 +60,12 @@ export type MessageRole = "user" | "assistant" | "system" | "tool"
 
 export type ContentBlock =
   | { type: "text"; text: string }
+  | {
+      type: "image"
+      data: string
+      mime_type: string
+      uri?: string | null
+    }
   | {
       type: "tool_use"
       tool_use_id: string | null
@@ -304,8 +310,27 @@ export type ConnectionStatus =
   | "disconnected"
   | "error"
 
+export interface PromptCapabilitiesInfo {
+  image: boolean
+  audio: boolean
+  embedded_context: boolean
+}
+
 export type PromptInputBlock =
   | { type: "text"; text: string }
+  | {
+      type: "image"
+      data: string
+      mime_type: string
+      uri?: string | null
+    }
+  | {
+      type: "resource"
+      uri: string
+      mime_type?: string | null
+      text?: string | null
+      blob?: string | null
+    }
   | {
       type: "resource_link"
       uri: string
@@ -430,6 +455,11 @@ export type AcpEvent =
       type: "selectors_ready"
       connection_id: string
     }
+  | {
+      type: "prompt_capabilities"
+      connection_id: string
+      prompt_capabilities: PromptCapabilitiesInfo
+    }
   | {
       type: "mode_changed"
       connection_id: string