New pic #4858
New pic #4858
Conversation
|
|
|
|
|
Preview mcp_server Image: |
|
Preview sandbox Image: |
|
Preview mcp_server Image: |
|
Preview sandbox Image: |
|
Preview mcp_server Image: |
|
Preview sandbox Image: |
|
Preview mcp_server Image: |
|
Preview sandbox Image: |
|
Preview fastgpt Image: |
|
Preview mcp_server Image: |
|
Preview sandbox Image: |
|
Preview fastgpt Image: |
|
Preview mcp_server Image: |
|
Preview sandbox Image: |
|
Preview fastgpt Image: |
|
Preview mcp_server Image: |
|
Preview sandbox Image: |
|
Preview mcp_server Image: |
|
Preview sandbox Image: |
|
Preview fastgpt Image: |
|
Preview mcp_server Image: |
|
Preview sandbox Image: |
|
Deployment Status: ✅ Success |
…based on data source type, improving the import process for non-image datasets.
…ings, enhancing error messages, and streamlining the preview URL generation process.
|
Preview mcp_server Image: |
|
Deployment Status: ✅ Success |
|
Preview sandbox Image: |
|
Preview fastgpt Image: |
|
Preview sandbox Image: |
|
Deployment Status: ✅ Success |
|
Preview mcp_server Image: |
|
Preview fastgpt Image: |
|
Preview mcp_server Image: |
|
Preview sandbox Image: |
|
Deployment Status: ✅ Success |
|
Preview mcp_server Image: |
|
Preview sandbox Image: |
|
Deployment Status: ✅ Success |
|
Preview fastgpt Image: |
| @@ -11,7 +11,7 @@ export const fileImgs = [ | |||
| // { suffix: '.', src: '/imgs/files/file.svg' } | |||
| ]; | |||
|
|
|||
| export function getFileIcon(name = '', defaultImg = 'file/fill/file') { | |||
| @@ -81,6 +81,7 @@ export type ApiDatasetCreateDatasetCollectionParams = ApiCreateDatasetCollection | |||
| }; | |||
| export type FileIdCreateDatasetCollectionParams = ApiCreateDatasetCollectionParams & { | |||
| fileId: string; | |||
| collectionId?: string; | |||
| @@ -129,6 +130,7 @@ export type PushDatasetDataChunkProps = { | |||
| a?: string; // bonus content | |||
| chunkIndex?: number; | |||
| indexes?: Omit<DatasetDataIndexItemType, 'dataId'>[]; | |||
| imageFileId?: string; //file id preview | |||
There was a problem hiding this comment.
检查其他地方的也是，不要把 q,a,imageId 分开写
| @@ -0,0 +1,19 @@ | |||
| export interface DatasetCollectionImageSchema { | |||
There was a problem hiding this comment.
系统都是用 type，非必要不用 interface
There was a problem hiding this comment.
dataset/image/type.d.ts
这个表是图库，不是图片知识库。
| @@ -0,0 +1,19 @@ | |||
| export interface DatasetCollectionImageSchema { | |||
| @@ -56,6 +59,7 @@ export const createCollectionAndInsertData = async ({ | |||
|
|
|||
| billId?: string; | |||
| session?: ClientSession; | |||
| parentCollectionId?: string; | |||
| @@ -156,7 +170,26 @@ export const createCollectionAndInsertData = async ({ | |||
| return newBillId; | |||
| })(); | |||
|
|
|||
| // 5. insert to training queue | |||
| // 5. Update the collectionId field in the image record (for image collections) | |||
| if (createCollectionParams.metadata?.isImageCollection && createCollectionParams.fileId) { | |||
There was a problem hiding this comment.
都说了，不能主动指定啥 isImageCollection
There was a problem hiding this comment.
createImageCollection 是一个单独的 api，这个函数不应该有任何改动。
| trainingType: DatasetCollectionSchemaType['trainingType']; | ||
| autoIndexes?: DatasetCollectionSchemaType['autoIndexes']; | ||
| imageIndex?: DatasetCollectionSchemaType['imageIndex']; | ||
| export const getTrainingModeByCollection = ({ |
There was a problem hiding this comment.
不用改，trainingType 已经有图片处理了
| @@ -372,3 +411,38 @@ export async function delCollection({ | |||
| ); | |||
| }); | |||
| } | |||
|
|
|||
| export async function pushImageFileToTrainingQueue({ | |||
There was a problem hiding this comment.
重写这个文件。
- 在 rawText 改成可选。 rawText 统计增加一个 imageIdList 的字段，表示插入的图片 ID。
- 修复第一步：const chunks = 有 rawText 的话走分块。没有的话，把图片 ID 作为 chunks，chunks 结构改成: q,a,imageId
- 最后批量删除过期索引
|
|
||
| /* ============= dataset images ========== */ | ||
|
|
||
| export async function createDatasetImage({ |
There was a problem hiding this comment.
放到 dataset/image/controller.ts 里
| return String(image._id); | ||
| } | ||
|
|
||
| export async function getDatasetImage(imageId: string): Promise<any> { |
|
|
||
| export async function getDatasetImage(imageId: string): Promise<any> { | ||
| try { | ||
| if (!imageId || imageId.length !== 24) { |
There was a problem hiding this comment.
Types.ObjectId(id).isValid() 可以校验
| @@ -145,3 +147,69 @@ try { | |||
| } | |||
|
|
|||
| export const MongoDataset = getMongoModel<DatasetSchemaType>(DatasetCollectionName, DatasetSchema); | |||
|
|
|||
| export const DatasetCollectionImageCollectionName = 'dataset_collection_images'; | |||
There was a problem hiding this comment.
写到 dataset/image/schema.ts 里
| @@ -184,6 +184,8 @@ | |||
| "comfirm_leave_page": "确认离开该页面？", | |||
| "comfirn_create": "确认创建", | |||
| "commercial_function_tip": "请升级商业版后使用该功能：https://doc.fastgpt.cn/docs/commercial/intro/", | |||
| "common.Create Failed": "创建集合失败", | |||
| @@ -1,4 +1,5 @@ | |||
| { | |||
| " error?.message || t('core.dataset.collection.Create Failed'),\n status.core.dataset.collection.Create Failed": "", | |||
| @@ -54,11 +54,14 @@ const QuoteList = React.memo(function QuoteList({ | |||
| const processedData = rawSearch.map((item) => { | |||
| if (chatItemDataId && quoteList) { | |||
| const currentFilterItem = quoteList.find((res) => res._id === item.id); | |||
| return { | |||
| const result = { | |||
| // indexes: DatasetDataSchemaType['indexes']; | ||
| imageFileId?: string; | ||
| imageSize?: number; | ||
| previewUrl?: string; |
| {t('dataset:retain_collection')} | ||
| </Button> | ||
| )} | ||
| {datasetDetail.type !== 'websiteDataset' && !!collection?.chunkSize && ( |
| @@ -283,11 +327,79 @@ const DataCard = () => { | |||
|
|
|||
| {/* Data content */} | |||
| <Box wordBreak={'break-all'} fontSize={'sm'}> | |||
| <Markdown source={item.q} isDisabled /> | |||
| {!!item.a && ( | |||
| {isImageCollection ? ( | |||
There was a problem hiding this comment.
不用管是不是 imageCollection，直接根据 data 里有没有 imagePreviewUrl 决定。有的话就渲染图片模式
| alignSelf="stretch" | ||
| borderRadius="md" | ||
| overflow="hidden" | ||
| bg="var(--Gray-Modern-100, #F4F4F7)" |
| padding="8px 8px 10px 8px" | ||
| justifyContent="center" | ||
| alignItems="center" | ||
| alignSelf="stretch" |
| <Box | ||
| flex="1 0 0" | ||
| color="var(--Gray-Modern-800, #1D2532)" | ||
| fontFamily="PingFang SC" |
There was a problem hiding this comment.
不用指定字体。
fontSize={"sm"}
lineheight，letterSpacing，maxHeight 不用指定
There was a problem hiding this comment.
大部分情况下，只需要指定 padding,margin fontSize,fontweight,color 即可。其他样式都是全局的了。
| @@ -230,6 +233,15 @@ export async function insertData2Dataset({ | |||
| { session, ordered: true } | |||
| ); | |||
|
|
|||
| // 4. Remove TTL from image if imageFileId exists (prevent image expiration during training) | |||
There was a problem hiding this comment.
insertData 不需要删 ttl，插入集合时候已经删过了。
需要再 insertData接口里进行 TTL 删除
| await deleteDatasetDataVector({ | ||
| teamId: data.teamId, | ||
| idList: data.indexes.map((item) => item.dataId) | ||
| }); | ||
|
|
||
| // 3. If there are any image files, delete the image records. | ||
| if (data.imageFileId) { |
There was a problem hiding this comment.
不要 catch，就是需要抛错，事务是保障全部成功，如果失败就要回退。
|
|
||
| // Handle imageFileId update | ||
| if (imageFileId !== undefined) { | ||
| await MongoDatasetData.findByIdAndUpdate(dataId, { |
知识库导入图片