feat: export reproducible retrieval history snapshots

frontend/package.json

@@ -6,7 +6,8 @@
   "scripts": {
     "dev": "vite",
     "build": "vue-tsc -b && vite build",
-    "preview": "vite preview"
+    "preview": "vite preview",
+    "test:session": "node --test tests/session.test.mjs"
   },
   "dependencies": {
     "lucide-vue-next": "^0.468.0",

frontend/src/services/exports.ts

@@ -0,0 +1,19 @@
+import { apiDownload } from './session'
+import type { RetrievalRunResult } from '../api'
+import type { RagContext } from './rag'
+
+export type ExportSelection = { run_ids: string[]; include_contexts?: boolean }
+export type ExportManifest = {
+  record_type: 'manifest'; schema_version: 'nalati-retrieval-export-v1'; export_id: string; exported_at: string;
+  selection: Required<ExportSelection>; counts: { retrieval_runs: number; rag_contexts: number }; records_sha256: string;
+  snapshot_semantics: 'stored_history_not_current_corpus';
+  generation_measurements: { status: 'not_called'; usage: null; cost: null; currency: null }
+}
+export type ExportRecord = ExportManifest
+  | { record_type: 'retrieval_run'; id: string; ranking_unit: 'knowledge_item' | 'chunk'; payload: RetrievalRunResult }
+  | { record_type: 'rag_context'; id: string; payload: RagContext }
+
+// Returns bytes/metadata; callers explicitly choose whether and where to save the file.
+export const exportRetrievalHistory = (payload: ExportSelection) => apiDownload(
+  '/api/retrieval/exports', { method: 'POST', body: JSON.stringify(payload) }, true,
+)

frontend/src/services/session.ts

@@ -6,10 +6,10 @@ export class ApiError extends Error {
 
 export function clearSessionToken() { csrfToken = null }
 
-export async function apiRequest<T>(path: string, init: RequestInit = {}, protect = false): Promise<T> {
+async function request(path: string, init: RequestInit, protect: boolean, accept: string): Promise<Response> {
   if (protect && !csrfToken) await apiRequest('/api/auth/me')
   const headers = new Headers(init.headers)
-  headers.set('Accept', 'application/json')
+  headers.set('Accept', accept)
   if (init.body) headers.set('Content-Type', 'application/json')
   if (protect && csrfToken) headers.set('X-CSRF-Token', csrfToken)
   const response = await fetch(path, { ...init, credentials: 'include', headers })
@@ -19,6 +19,11 @@ export async function apiRequest<T>(path: string, init: RequestInit = {}, protec
     const detail = body && typeof body === 'object' && 'detail' in body ? body.detail : null
     throw new ApiError(response.status, typeof detail === 'string' ? detail : `请求失败（${response.status}），请检查登录状态或填写内容。`)
   }
+  return response
+}
+
+export async function apiRequest<T>(path: string, init: RequestInit = {}, protect = false): Promise<T> {
+  const response = await request(path, init, protect, 'application/json')
   if (response.status === 204) return undefined as T
   const data = await response.json()
   if (path === '/api/auth/login' || path === '/api/auth/me') {
@@ -26,3 +31,12 @@ export async function apiRequest<T>(path: string, init: RequestInit = {}, protec
   }
   return data as T
 }
+
+export type ApiDownload = { blob: Blob; filename: string; recordsSha256: string | null }
+
+export async function apiDownload(path: string, init: RequestInit = {}, protect = false): Promise<ApiDownload> {
+  const response = await request(path, init, protect, 'application/x-ndjson')
+  const disposition = response.headers.get('Content-Disposition') ?? ''
+  const filename = /^attachment; filename="([A-Za-z0-9._-]+)"$/.exec(disposition)?.[1] ?? 'download.jsonl'
+  return { blob: await response.blob(), filename, recordsSha256: response.headers.get('X-Records-SHA256') }
+}

frontend/tests/session.test.mjs

@@ -0,0 +1,70 @@
+import assert from 'node:assert/strict'
+import { readFile } from 'node:fs/promises'
+import test from 'node:test'
+import ts from 'typescript'
+
+async function session(t, fetch_) {
+  const previous = globalThis.fetch
+  globalThis.fetch = fetch_
+  t.after(() => { globalThis.fetch = previous })
+  const source = await readFile(new URL('../src/services/session.ts', import.meta.url), 'utf8')
+  const { outputText } = ts.transpileModule(source, { compilerOptions: { target: ts.ScriptTarget.ES2022, module: ts.ModuleKind.ESNext } })
+  return import(`data:text/javascript;base64,${Buffer.from(outputText).toString('base64')}#${t.name}`)
+}
+
+test('JSON login and binary export share Cookie/CSRF transport without parsing JSONL as JSON', async t => {
+  const content = '{"record_type":"manifest"}\n{"query":"合成测试"}\n'
+  const calls = []
+  const api = await session(t, async (path, init) => {
+    calls.push([path, init])
+    if (path === '/api/auth/login') return Response.json({ csrf_token: 'fixture-csrf' })
+    return new Response(content, { headers: { 'Content-Disposition': 'attachment; filename="fixture.jsonl"', 'X-Records-SHA256': 'a'.repeat(64) } })
+  })
+  await api.apiRequest('/api/auth/login', { method: 'POST', body: '{}' })
+  const file = await api.apiDownload('/api/retrieval/exports', { method: 'POST', body: '{"run_ids":[]}' }, true)
+  assert.equal(await file.blob.text(), content)
+  assert.equal(file.filename, 'fixture.jsonl')
+  assert.equal(file.recordsSha256, 'a'.repeat(64))
+  assert.equal(calls[1][1].credentials, 'include')
+  assert.equal(calls[1][1].headers.get('X-CSRF-Token'), 'fixture-csrf')
+  assert.equal(calls[1][1].headers.get('Content-Type'), 'application/json')
+  assert.equal(calls[1][1].headers.get('Accept'), 'application/x-ndjson')
+  assert.equal(calls[0][1].headers.get('Accept'), 'application/json')
+})
+
+test('download refreshes missing session token and preserves server authorization errors', async t => {
+  const calls = []
+  const api = await session(t, async (path, init) => {
+    calls.push([path, init])
+    if (path === '/api/auth/me') return Response.json({ csrf_token: 'refreshed-csrf' })
+    return Response.json({ detail: '需要管理员权限' }, { status: 403 })
+  })
+  await assert.rejects(api.apiDownload('/api/retrieval/exports', { method: 'POST', body: '{}' }, true), error => {
+    assert.ok(error instanceof api.ApiError)
+    assert.equal(error.status, 403)
+    assert.equal(error.message, '需要管理员权限')
+    return true
+  })
+  assert.equal(calls.length, 2)
+  assert.equal(calls[0][0], '/api/auth/me')
+  assert.equal(calls[1][1].headers.get('X-CSRF-Token'), 'refreshed-csrf')
+})
+
+test('401 download clears stale token and unsafe filename is not propagated', async t => {
+  let current = 'old-csrf'
+  let failed = false
+  const api = await session(t, async (path, init) => {
+    if (path === '/api/auth/login' || path === '/api/auth/me') return Response.json({ csrf_token: current })
+    assert.equal(init.headers.get('X-CSRF-Token'), current)
+    if (!failed) {
+      failed = true; current = 'new-csrf'
+      return Response.json({ detail: '请先登录' }, { status: 401 })
+    }
+    return new Response('fixture', { headers: { 'Content-Disposition': 'attachment; filename="../../fixture.jsonl"' } })
+  })
+  await api.apiRequest('/api/auth/login', { method: 'POST', body: '{}' })
+  await assert.rejects(api.apiDownload('/api/retrieval/exports', { method: 'POST' }, true), error => error.status === 401)
+  const file = await api.apiDownload('/api/retrieval/exports', { method: 'POST' }, true)
+  assert.equal(file.filename, 'download.jsonl')
+  assert.equal(file.recordsSha256, null)
+})