feat: 完成数据处理接口与前端接入

This commit is contained in:
caoxiaozhu
2026-07-23 15:10:13 +08:00
parent f453234057
commit f04dc479bb
29 changed files with 7126 additions and 1144 deletions

View File

@@ -10,7 +10,7 @@ import SourceUploadStep from './create/SourceUploadStep.vue'
import PreviewCompareStep from './create/PreviewCompareStep.vue'
import GenerationStep from './create/GenerationStep.vue'
import ResultEditorStep from './create/ResultEditorStep.vue'
import { buildPreviewItems, DEFAULT_SOURCE_TEXT } from './create/previewModel'
import { DEFAULT_SOURCE_TEXT, estimateTokenCount } from './create/previewModel'
import {
createDefaultStructuredOptions,
createDefaultUnstructuredOptions,
@@ -21,6 +21,25 @@ import {
} from './create/useDataProcessDraft'
import { useDataProcessGeneration } from './create/useDataProcessGeneration'
import { useModelsStore } from '@/stores/models'
import {
buildDataProcessPreview,
createDataProcessPreview,
createDataProcessTask,
deleteDataProcessPreview,
deleteDataProcessSourceFile,
getDataProcessPreview,
getDataProcessSourceContent,
getDataProcessTask,
pullDataProcessExternalSource,
testDataProcessExternalSource,
updateDataProcessPreview,
updateDataProcessTask,
uploadDataProcessSourceFiles,
type DataProcessExternalSourcePayload,
type DataProcessPreviewItem,
type DataProcessSourceFile,
} from '@/api/modules/dataProcess'
import type { DataProcessConfig } from '@/types/dataProcess'
import type {
ExternalDataSource,
GenerationControlOptions,
@@ -35,11 +54,14 @@ import type {
const router = useRouter()
const modelsStore = useModelsStore()
const { list: modelList } = storeToRefs(modelsStore)
const generationModels = computed(() => modelList.value.filter((model) => model.type === 'LLM'))
const generationModels = computed(() => modelList.value.filter((model) => (
model.type === 'LLM'
&& (model.model_source === 'api' || model.model_source === 'online' || Boolean(model.api_url))
)))
const taskSetupRef = ref<InstanceType<typeof TaskSetupStep>>()
const modelSelectionRef = ref<InstanceType<typeof ModelSelectionStep>>()
const confirmDialogRef = ref<InstanceType<typeof AppConfirmDialog>>()
const PREVIEW_MODEL_VERSION = 'document-chunk-v2'
const PREVIEW_MODEL_VERSION = 'backend-pipeline-v1'
const WIZARD_STEPS = [
{ id: 'create', title: '创建任务', desc: '填写任务信息与处理配置' },
@@ -52,6 +74,7 @@ const WIZARD_STEPS = [
const currentStep = ref(0)
const currentStepId = computed<StepId>(() => WIZARD_STEPS[currentStep.value]?.id ?? 'create')
const task = reactive({ name: '', description: '' })
const taskId = ref<string | null>(null)
const processType = ref<ProcessType>('structured')
const structuredOptions = ref<StructuredProcessOptions>(createDefaultStructuredOptions())
const unstructuredOptions = ref<UnstructuredProcessOptions>(createDefaultUnstructuredOptions())
@@ -61,18 +84,17 @@ const modelSelectionOptions = computed<GenerationControlOptions>(() => (
const uploadedFiles = ref<UploadedDataFile[]>([])
const externalSource = reactive<ExternalDataSource>({
type: 'mysql',
type: 'postgresql',
url: '',
authMode: 'none',
username: '',
password: '',
token: '',
limit: 1000,
query: '',
fileName: 'external-data.jsonl',
})
const externalPulling = ref(false)
const externalConnected = ref(false)
let connectionTimer: ReturnType<typeof setTimeout> | null = null
let pullTimer: ReturnType<typeof setTimeout> | null = null
const fileName = computed(() => uploadedFiles.value.map(f => f.name).join(', '))
const previewSignature = ref('')
@@ -88,6 +110,7 @@ const {
generation,
results,
selectedResultId,
persistResultChanges,
resetDownstream,
restoreResult,
startGeneration,
@@ -96,11 +119,9 @@ const {
updateResultField,
validateResults,
} = useDataProcessGeneration({
previewItems,
processType,
structuredOptions,
unstructuredOptions,
taskId,
dirty,
beforeGenerate: syncPreviewChanges,
})
const modifiedPreviewCount = computed(() => previewItems.value.filter((item) => item.status !== 'original').length)
@@ -160,7 +181,101 @@ function updateModelSelectionOptions(value: GenerationControlOptions) {
structuredOptions.value = { ...structuredOptions.value, ...value }
}
function toBackendConfig(): DataProcessConfig {
const options = processType.value === 'unstructured'
? unstructuredOptions.value
: structuredOptions.value
const common = {
preprocess_options: [...options.preprocessOptions],
semantic_enrichment: options.semanticEnrichment,
dataset_split: { ...options.datasetSplit },
generation_model_id: options.generationModelId,
generation_prompt: options.generationPrompt,
temperature: options.temperature,
max_tokens: options.maxTokens,
json_mode: options.jsonMode,
quality_filter_enabled: options.qualityFilterEnabled,
filter_low_quality: options.filterLowQuality,
filter_short_content: options.filterShortContent,
min_output_length: options.minOutputLength,
}
if (processType.value === 'unstructured') {
return {
...common,
chunk_method: unstructuredOptions.value.chunkMethod,
chunk_size: unstructuredOptions.value.chunkSize,
chunk_overlap: unstructuredOptions.value.chunkOverlap,
min_chunk_size: unstructuredOptions.value.minChunkSize,
custom_delimiter: unstructuredOptions.value.customDelimiter,
preserve_tables: unstructuredOptions.value.preserveTables,
preserve_code_blocks: unstructuredOptions.value.preserveCodeBlocks,
preserve_lists: unstructuredOptions.value.preserveLists,
qa_pairs_per_chunk: unstructuredOptions.value.qaPairsPerChunk,
}
}
return {
...common,
qa_pairs_per_row: structuredOptions.value.qaPairsPerRow,
}
}
function taskPayload() {
return {
name: task.name.trim(),
description: task.description.trim(),
process_type: processType.value,
config: toBackendConfig(),
}
}
function externalPayload(): DataProcessExternalSourcePayload {
return {
type: externalSource.type,
url: externalSource.url.trim(),
auth_mode: externalSource.authMode,
username: externalSource.username || undefined,
password: externalSource.password || undefined,
limit: externalSource.limit,
query: externalSource.query?.trim() || undefined,
file_name: externalSource.fileName || 'external-data.jsonl',
}
}
function mapPreviewItem(item: DataProcessPreviewItem): PreviewItem {
return {
id: String(item.id),
sourceFileId: String(item.source_file_id),
originalContent: item.original_content,
editedContent: item.edited_content,
sourceStart: item.source_start,
sourceEnd: item.source_end,
sourceStartLine: item.source_start_line,
sourceEndLine: item.source_end_line,
tokenCount: item.token_count,
status: item.status,
updatedAt: item.updated_at,
}
}
function mapSourceFile(file: DataProcessSourceFile, content = ''): UploadedDataFile {
return {
uid: String(file.id),
sourceFileId: String(file.id),
name: file.name,
size: file.size_bytes,
count: file.record_count,
content,
fileFormat: file.file_format,
checksumSha256: file.checksum_sha256,
status: 'ready',
}
}
const { persistDraft, restoreDraft } = useDataProcessDraft({
taskId,
currentStepId,
task,
processType,
@@ -255,7 +370,7 @@ const generationOptionsSignature = computed(() => JSON.stringify(generationAffec
function buildPreviewSignature() {
const filesSignature = uploadedFiles.value
.map((file) => `${file.uid}:${file.name}:${file.size}:${file.count}`)
.map((file) => `${file.uid}:${file.name}:${file.size}:${file.checksumSha256 || file.count}`)
.join('|')
return `${PREVIEW_MODEL_VERSION}:${processType.value}:${JSON.stringify(previewAffectingOptions())}:${filesSignature}`
}
@@ -280,7 +395,7 @@ watch(generationOptionsSignature, (currentSignature, previousSignature) => {
})
watch(
[task, processType, structuredOptions, unstructuredOptions, externalSource],
[taskId, task, processType, structuredOptions, unstructuredOptions, externalSource],
persistDraft,
{ deep: true },
)
@@ -294,48 +409,85 @@ watch(currentStep, () => nextTick(scrollToStepTop))
async function handleFileChange(uploadFile: UploadFile) {
const raw = uploadFile.raw
if (!raw) return
if (!taskId.value) {
ElMessage.error('任务尚未创建,请返回模型选择步骤后重试')
return
}
if (raw.size > 200 * 1024 * 1024) {
ElMessage.warning('单文件不能超过 200MB')
return
}
const extension = raw.name.split('.').pop()?.toLowerCase() ?? ''
const textExtensions = ['txt', 'md', 'json', 'jsonl', 'csv']
const textExtensions = new Set(['txt', 'md', 'json', 'jsonl', 'csv'])
if (!textExtensions.has(extension)) {
ElMessage.error('当前仅支持 TXT、Markdown、JSON、JSONL 和 CSV不会用示例内容替代无法解析的文件')
return
}
if (uploadedFiles.value.some((file) => file.name === raw.name && file.size === raw.size)) {
ElMessage.warning('同名且同大小的文件已经上传')
return
}
let content = ''
if (textExtensions.includes(extension)) {
try {
content = await raw.text()
} catch {
content = ''
}
try {
content = new TextDecoder('utf-8', { fatal: true }).decode(await raw.arrayBuffer())
} catch {
ElMessage.error('文件不是有效的 UTF-8 文本,请转换编码后重试')
return
}
if (!content.trim()) {
ElMessage.warning('不能上传空文件')
return
}
const fileContent = content.trim() ? content : DEFAULT_SOURCE_TEXT
const linesCount = fileContent.split('\n').filter((line) => line.trim()).length
// Prevent duplicate upload of the same file
if (!uploadedFiles.value.some(f => f.name === raw.name && f.size === raw.size)) {
uploadedFiles.value.push({
uid: uploadFile.uid || Date.now() + Math.random(),
name: raw.name,
size: raw.size,
count: linesCount,
content: fileContent
})
try {
const uploaded = await uploadDataProcessSourceFiles(taskId.value, [raw])
const source = uploaded.files[0]
if (!source) throw new Error('后端未返回源文件记录')
uploadedFiles.value.push(mapSourceFile(source, content))
previewSignature.value = ''
resetDownstream()
dirty.value = true
ElMessage.success(`文件 ${source.name} 上传成功`)
} catch {
// 请求层已展示后端的解析或格式错误。
}
dirty.value = true
}
function useSampleFile() {
uploadedFiles.value = [{
uid: 'sample-1',
name: 'finance_qa.jsonl',
size: 128 * 1024 * 1024,
count: DEFAULT_SOURCE_TEXT.split('\n').filter((line) => line.trim()).length,
content: DEFAULT_SOURCE_TEXT
}]
dirty.value = true
async function useSampleFile() {
if (!taskId.value) {
ElMessage.error('任务尚未创建,请返回模型选择步骤后重试')
return
}
const sample = new File([DEFAULT_SOURCE_TEXT], 'finance_qa.jsonl', { type: 'application/x-ndjson' })
await handleFileChange({ raw: sample, uid: Date.now(), name: sample.name } as UploadFile)
}
async function restoreRegisteredSources() {
if (!taskId.value) return
try {
const savedTask = await getDataProcessTask(taskId.value)
const sources = savedTask.source_files || []
const restoredFiles = await Promise.all(sources.map(async (file) => {
try {
const source = await getDataProcessSourceContent(taskId.value!, file.id, {
start_line: 1,
line_count: 5000,
})
return mapSourceFile(file, source.content)
} catch {
return mapSourceFile(file)
}
}))
uploadedFiles.value = restoredFiles
if (restoredFiles.length) {
ElMessage.success(`已同步 ${restoredFiles.length} 个已登记源文件`)
}
} catch {
ElMessage.warning('草稿任务暂时无法从后端同步,请检查服务后重试')
}
}
function updateExternalSource(value: ExternalDataSource) {
@@ -343,58 +495,81 @@ function updateExternalSource(value: ExternalDataSource) {
externalConnected.value = false
}
function handleTestConnection() {
async function handleTestConnection() {
if (!taskId.value) {
ElMessage.error('任务尚未创建,请返回模型选择步骤后重试')
return
}
if (!externalSource.url.trim()) {
ElMessage.warning('请先填写数据源地址')
return
}
if (connectionTimer) clearTimeout(connectionTimer)
externalPulling.value = true
connectionTimer = setTimeout(() => {
connectionTimer = null
try {
const result = await testDataProcessExternalSource(taskId.value, externalPayload())
externalConnected.value = result.connected
if (result.connected) ElMessage.success(result.message || '数据源连接测试成功')
else ElMessage.warning(result.message || '数据源连接失败')
} catch {
externalConnected.value = false
} finally {
externalPulling.value = false
externalConnected.value = true
ElMessage.success('数据源连接测试成功')
}, 1500)
}
}
function handlePullData() {
async function handlePullData() {
if (!taskId.value) {
ElMessage.error('任务尚未创建,请返回模型选择步骤后重试')
return
}
if (!externalSource.url.trim()) {
ElMessage.warning('请先填写数据源地址')
return
}
if (pullTimer) clearTimeout(pullTimer)
if (!externalSource.query?.trim()) {
ElMessage.warning('请先填写只读 SELECT 查询语句')
return
}
externalPulling.value = true
pullTimer = setTimeout(() => {
pullTimer = null
externalPulling.value = false
try {
const response = await pullDataProcessExternalSource(taskId.value, externalPayload())
const newFiles: UploadedDataFile[] = []
for (const file of response.files) {
const source = await getDataProcessSourceContent(taskId.value, file.id, {
start_line: 1,
line_count: 5000,
})
newFiles.push(mapSourceFile(file, source.content))
}
uploadedFiles.value.push(...newFiles)
externalConnected.value = true
const typeName = externalSource.type.toUpperCase()
const id = `external-${Date.now()}`
uploadedFiles.value.push({
uid: id,
name: `${typeName} 拉取数据 ${new Date().toLocaleString('zh-CN')}`,
size: Math.min(externalSource.limit, 5000) * 64,
count: Math.min(externalSource.limit, DEFAULT_SOURCE_TEXT.split('\n').filter((line) => line.trim()).length),
content: DEFAULT_SOURCE_TEXT,
})
dirty.value = true
ElMessage.success(`已成功拉取 ${uploadedFiles.value[uploadedFiles.value.length - 1].count.toLocaleString()} 条数据`)
}, 2000)
}
function handleRemoveFile(uid: string | number) {
const index = uploadedFiles.value.findIndex(f => f.uid === uid)
if (index > -1) {
uploadedFiles.value.splice(index, 1)
previewSignature.value = ''
previewItems.value = []
selectedPreviewId.value = null
resetDownstream()
dirty.value = true
ElMessage.success(`已成功登记 ${newFiles.length} 个外部源文件`)
} catch {
externalConnected.value = false
} finally {
externalPulling.value = false
}
}
async function handleRemoveFile(uid: string | number) {
const index = uploadedFiles.value.findIndex(f => f.uid === uid)
if (index < 0 || !taskId.value) return
try {
await deleteDataProcessSourceFile(taskId.value, uid)
} catch {
return
}
uploadedFiles.value.splice(index, 1)
previewSignature.value = ''
previewItems.value = []
selectedPreviewId.value = null
resetDownstream()
dirty.value = true
}
function resetSourceDataForProcessTypeChange() {
uploadedFiles.value = []
previewSignature.value = ''
@@ -415,10 +590,24 @@ async function nextFromCreate() {
async function nextFromModel() {
const valid = await modelSelectionRef.value?.validate()
if (!valid) return
goToStep('upload')
try {
const saved = taskId.value
? await updateDataProcessTask(taskId.value, taskPayload())
: await createDataProcessTask(taskPayload())
taskId.value = String(saved.id)
dirty.value = true
persistDraft()
goToStep('upload')
} catch {
// 请求层已展示名称冲突或配置非法等具体原因。
}
}
function nextFromUpload() {
async function nextFromUpload() {
if (!taskId.value) {
ElMessage.error('任务尚未创建,请返回模型选择步骤后重试')
return
}
if (uploadedFiles.value.length === 0) {
ElMessage.warning(processType.value === 'external' ? '请先拉取至少一个数据源' : '请上传至少一个源数据文件')
return
@@ -426,14 +615,25 @@ function nextFromUpload() {
const signature = buildPreviewSignature()
if (signature !== previewSignature.value) {
previewItems.value = uploadedFiles.value.flatMap((file) =>
buildPreviewItems(
file.content,
processType.value,
String(file.uid),
processType.value === 'unstructured' ? unstructuredOptions.value : undefined,
),
)
try {
await buildDataProcessPreview(taskId.value, {
source_file_ids: uploadedFiles.value.map((file) => file.sourceFileId || file.uid),
})
const first = await getDataProcessPreview(taskId.value, { page: 1, page_size: 500 })
const items = [...first.items]
const pages = Math.ceil(first.total / first.page_size)
for (let page = 2; page <= pages; page += 1) {
const next = await getDataProcessPreview(taskId.value, { page, page_size: 500 })
items.push(...next.items)
}
previewItems.value = items.map(mapPreviewItem)
} catch {
return
}
if (!previewItems.value.length) {
ElMessage.warning('源文件没有生成可用的预览条目,请检查文件内容和预处理配置')
return
}
selectedPreviewFileId.value = String(uploadedFiles.value[0]?.uid ?? '') || null
selectedPreviewId.value = activePreviewItems.value[0]?.id ?? null
selectedPreviewIdsByFile.value = selectedPreviewId.value && selectedPreviewFileId.value
@@ -464,40 +664,50 @@ function updatePreviewContent(id: string, value: string) {
const item = previewItems.value.find((entry) => entry.id === id)
if (!item) return
item.editedContent = value
item.tokenCount = Math.max(1, Math.ceil(value.length / 2))
item.tokenCount = estimateTokenCount(value)
item.status = value === item.originalContent ? 'original' : item.sourceStart == null ? 'manual' : 'modified'
resetDownstream()
dirty.value = true
}
async function syncPreviewChanges() {
if (!taskId.value) throw new Error('任务尚未创建')
const changedItems = previewItems.value.filter((item) => item.status === 'modified' || item.status === 'manual')
for (const item of changedItems) {
const saved = await updateDataProcessPreview(taskId.value, item.id, {
edited_content: item.editedContent,
expected_updated_at: item.updatedAt,
})
const index = previewItems.value.findIndex((entry) => entry.id === item.id)
if (index >= 0) previewItems.value[index] = mapPreviewItem(saved)
}
}
function restorePreviewItem(id: string) {
const item = previewItems.value.find((entry) => entry.id === id)
if (!item || item.sourceStart == null) return
item.editedContent = item.originalContent
item.tokenCount = Math.max(1, Math.ceil(item.originalContent.length / 2))
item.tokenCount = estimateTokenCount(item.originalContent)
item.status = 'original'
resetDownstream()
dirty.value = true
}
function addPreviewItem() {
if (!selectedPreviewFileId.value) return
const id = `manual-${Date.now()}`
previewItems.value.push({
id,
sourceFileId: selectedPreviewFileId.value,
originalContent: '',
editedContent: '',
sourceStart: null,
sourceEnd: null,
sourceStartLine: null,
sourceEndLine: null,
tokenCount: 1,
status: 'manual',
})
selectPreviewItem(id)
resetDownstream()
dirty.value = true
async function addPreviewItem() {
if (!selectedPreviewFileId.value || !taskId.value) return
try {
const created = await createDataProcessPreview(taskId.value, {
source_file_id: selectedPreviewFileId.value,
edited_content: '',
})
const item = mapPreviewItem(created)
previewItems.value.push(item)
selectPreviewItem(item.id)
resetDownstream()
dirty.value = true
} catch {
// 请求层已展示错误。
}
}
async function removePreviewItem(id: string) {
@@ -512,6 +722,12 @@ async function removePreviewItem(id: string) {
const index = previewItems.value.findIndex((item) => item.id === id)
if (index < 0) return
if (!taskId.value) return
try {
await deleteDataProcessPreview(taskId.value, id)
} catch {
return
}
previewItems.value.splice(index, 1)
selectedPreviewId.value = activePreviewItems.value[Math.min(index, activePreviewItems.value.length - 1)]?.id ?? null
if (selectedPreviewFileId.value && selectedPreviewId.value) {
@@ -531,7 +747,7 @@ async function handlePrimaryAction() {
return
}
if (currentStepId.value === 'upload') {
nextFromUpload()
await nextFromUpload()
return
}
if (currentStepId.value === 'preview') {
@@ -546,7 +762,7 @@ async function handlePrimaryAction() {
if (generation.status === 'success') {
goToStep('results')
} else if (generation.status !== 'running') {
startGeneration()
await startGeneration()
}
return
}
@@ -567,6 +783,15 @@ async function saveTask() {
ElMessage.warning('请先修正校验失败的结果')
return
}
try {
await persistResultChanges()
} catch {
return
}
if (!validateResults()) {
ElMessage.warning('仍有结果未通过后端质量校验,请继续修正')
return
}
dirty.value = false
localStorage.removeItem(DATA_PROCESS_DRAFT_STORAGE_KEY)
allowLeave = true
@@ -607,12 +832,10 @@ onBeforeRouteLeave(async () => {
onBeforeUnmount(() => {
stopGenerationTimer()
if (connectionTimer) clearTimeout(connectionTimer)
if (pullTimer) clearTimeout(pullTimer)
})
onMounted(() => {
onMounted(async () => {
restoreDraft()
modelsStore.load()
await Promise.all([modelsStore.load(), restoreRegisteredSources()])
})
</script>