Files
arcrun-collector/direct_extract_test.go
T
Leo 1738acaca7 fix(t108 🔴🔴): 逐帳號同步遺失萃取設定→原文直送——三層修+契約保險絲
事故(leo 機 07-28 17:3x 總管現場抓到):t104 config 重寫吃掉機器層 extractor/key、
per-account DirectConfig 不繼承 ⇒ extractor 空 ⇒ 6 筆走 rag_ingest_direct 原文出機。
修:①config 讀改寫全程保留既有欄位(帶測試)②帳號層無值垂直繼承機器層
extractor/claude_bin/gemini_api_key/llm_model/CardIngestWF/RemovedWF(帶 fake server 測試:
必須收到 rag_ingest_card 非 direct)③契約保險絲:extractor 空時非 .md/.txt 禁直送、
標 failed「萃取器未設定,已跳過(不直送原文)」——同類 bug 永不再成外洩。
三模組 go test 全綠(總管親跑)。(實作=子 CC;驗證+commit=總管)
2026-07-28 17:59:42 +08:00

317 lines
12 KiB
Go
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
// direct_extract_test.go — task 6extractor 模式端到端(本地萃卡→POST rag_ingest_card)。
package main
import (
"encoding/json"
"io"
"net/http"
"net/http/httptest"
"os"
"path/filepath"
"strings"
"testing"
)
// 完整鏈(claude stub 版):丟原稿 → 萃卡落地本地 → 只有「卡片」被 POST 到 rag_ingest_card
// → 原文從未離開本機 → manifest 標 ingested(下一輪不重送)。
func TestDirectExtractorModeE2E(t *testing.T) {
root := t.TempDir()
if err := os.WriteFile(filepath.Join(root, "報銷規則.md"), []byte("# 原稿機密內容 XYZZY"), 0o644); err != nil {
t.Fatal(err)
}
// 假 cypher:收 rag_ingest_card、驗 payload、記帳
var posted []map[string]any
srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if !strings.HasSuffix(r.URL.Path, "/webhooks/named/demo/rag_ingest_card/trigger") {
t.Errorf("打錯端點:%s", r.URL.Path)
}
body, _ := io.ReadAll(r.Body)
var m map[string]any
_ = json.Unmarshal(body, &m)
posted = append(posted, m)
_ = json.NewEncoder(w).Encode(map[string]any{"success": true})
}))
defer srv.Close()
// stub claude:把原稿萃成卡(模擬 /wiki-capture 行為)
stubDir := t.TempDir()
stub := filepath.Join(stubDir, "claude")
script := "#!/bin/sh\nmkdir -p system-dev/wiki/cards\nprintf '# 報銷規則\\n## 一句話定義\\n測試卡\\n## 關聯\\n- 報銷規則 >> 屬於 >> 財務\\n' > 'system-dev/wiki/cards/報銷規則.md'\n"
if err := os.WriteFile(stub, []byte(script), 0o755); err != nil {
t.Fatal(err)
}
cfg := &DirectConfig{
WatchFolders: []string{root},
Manifest: filepath.Join(t.TempDir(), "m.json"),
CypherURL: srv.URL, Namespace: "demo", APIKey: "demo",
Library: "kb", Extractor: "claude", ClaudeBin: stub,
CardIngestWF: "rag_ingest_card", MaxRemoved: DefaultMaxRemovedRatio,
}
results, exit, _ := RunDirectOnce(cfg, false)
if exit != 0 {
t.Fatalf("exit=%d results=%+v", exit, results)
}
if len(results) != 1 || results[0].Status != "ingested" {
t.Fatalf("results=%+v", results)
}
// 卡片落地本地(用戶看得到自己的 wiki)
if _, err := os.Stat(filepath.Join(root, "system-dev", "wiki", "cards", "報銷規則.md")); err != nil {
t.Fatalf("卡片未落地:%v", err)
}
// 上雲的是卡片、不是原文
if len(posted) != 1 {
t.Fatalf("應恰好 POST 一張卡,got %d", len(posted))
}
cc, _ := posted[0]["card_content"].(string)
if !strings.Contains(cc, "## 一句話定義") {
t.Fatalf("card_content 不是卡片:%.80s", cc)
}
if strings.Contains(cc, "XYZZY") {
t.Fatal("原文內容洩上雲=違反四步定稿邊界")
}
// path 必須是「原檔路徑」(takedown 比對鍵+B4 溯源)——不是卡片路徑(07-24 第五枚坑)
if p, _ := posted[0]["path"].(string); p != "報銷規則.md" {
t.Fatalf("path=%q(應為原檔路徑)", p)
}
// 第二輪:原稿沒變 → 不重萃不重送
results2, exit2, _ := RunDirectOnce(cfg, false)
if exit2 != 0 || len(results2) != 0 || len(posted) != 1 {
t.Fatalf("第二輪應零事件:results=%+v posted=%d", results2, len(posted))
}
}
// t15extractor 模式刪原檔 → 雲端 takedown 成功後,本地萃出的卡也要被清掉。
func TestDirectExtractorRemovedClearsLocalCard(t *testing.T) {
root := t.TempDir()
if err := os.WriteFile(filepath.Join(root, "報銷規則.md"), []byte("# 原稿"), 0o644); err != nil {
t.Fatal(err)
}
// 假 cypher:收 rag_ingest_card 與 rag_takedown_direct
var takedowns []map[string]any
srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if strings.HasSuffix(r.URL.Path, "/webhooks/named/demo/rag_takedown_direct/trigger") {
body, _ := io.ReadAll(r.Body)
var m map[string]any
_ = json.Unmarshal(body, &m)
takedowns = append(takedowns, m)
}
_ = json.NewEncoder(w).Encode(map[string]any{"success": true})
}))
defer srv.Close()
// stub claude:萃卡落地
stubDir := t.TempDir()
stub := filepath.Join(stubDir, "claude")
script := "#!/bin/sh\nmkdir -p system-dev/wiki/cards\nprintf '# 報銷規則\\n## 一句話定義\\n測試卡\\n' > 'system-dev/wiki/cards/報銷規則.md'\n"
if err := os.WriteFile(stub, []byte(script), 0o755); err != nil {
t.Fatal(err)
}
cfg := &DirectConfig{
WatchFolders: []string{root},
Manifest: filepath.Join(t.TempDir(), "m.json"),
CypherURL: srv.URL, Namespace: "demo", APIKey: "demo",
Library: "kb", Extractor: "claude", ClaudeBin: stub,
CardIngestWF: "rag_ingest_card", RemovedWF: "rag_takedown_direct",
// 單檔刪除=removed ratio 100%,預設 0.4 防呆會壓下事件;本測試聚焦下架路,放寬到 1.0
//1 > 1.0×1 為 false → 事件放行)。
MaxRemoved: 1.0,
}
// 第一輪:萃卡+上雲,本地卡存在
if _, exit, _ := RunDirectOnce(cfg, false); exit != 0 {
t.Fatalf("第一輪 ingest 失敗 exit=%d", exit)
}
cardPath := filepath.Join(root, "system-dev", "wiki", "cards", "報銷規則.md")
if _, err := os.Stat(cardPath); err != nil {
t.Fatalf("前置失敗:卡片未落地 %v", err)
}
// 刪原檔 → 第二輪:takedown 打出去、本地卡也被清
if err := os.Remove(filepath.Join(root, "報銷規則.md")); err != nil {
t.Fatal(err)
}
results, exit, _ := RunDirectOnce(cfg, false)
if exit != 0 {
t.Fatalf("第二輪 exit=%d results=%+v", exit, results)
}
if len(results) != 1 || results[0].Status != "removed" {
t.Fatalf("results=%+v", results)
}
if len(takedowns) != 1 {
t.Fatalf("應恰好一次 takedowngot %d", len(takedowns))
}
if pn, _ := takedowns[0]["page_name"].(string); pn != "報銷規則" {
t.Fatalf("takedown page_name=%q", pn)
}
if _, err := os.Stat(cardPath); !os.IsNotExist(err) {
t.Fatalf("本地卡應已被清(err=%v", err)
}
}
// t15:本地卡不存在時(存在才刪)下架照常成功,不多出 warning。
func TestDirectExtractorRemovedNoLocalCardOK(t *testing.T) {
root := t.TempDir()
if err := os.WriteFile(filepath.Join(root, "a.md"), []byte("x"), 0o644); err != nil {
t.Fatal(err)
}
srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
_ = json.NewEncoder(w).Encode(map[string]any{"success": true})
}))
defer srv.Close()
stubDir := t.TempDir()
stub := filepath.Join(stubDir, "claude")
script := "#!/bin/sh\nmkdir -p system-dev/wiki/cards\nprintf '# a\\n## 一句話定義\\n卡\\n' > system-dev/wiki/cards/a.md\n"
if err := os.WriteFile(stub, []byte(script), 0o755); err != nil {
t.Fatal(err)
}
cfg := &DirectConfig{
WatchFolders: []string{root},
Manifest: filepath.Join(t.TempDir(), "m.json"),
CypherURL: srv.URL, Namespace: "demo", APIKey: "demo",
Extractor: "claude", ClaudeBin: stub,
CardIngestWF: "rag_ingest_card", RemovedWF: "rag_takedown_direct",
MaxRemoved: 1.0,
}
if _, exit, _ := RunDirectOnce(cfg, false); exit != 0 {
t.Fatal("第一輪失敗")
}
// 模擬用戶已手動清走本地卡 → removed 分支「存在才刪」不應報錯或多出 warning
if err := os.Remove(filepath.Join(root, "system-dev", "wiki", "cards", "a.md")); err != nil {
t.Fatal(err)
}
if err := os.Remove(filepath.Join(root, "a.md")); err != nil {
t.Fatal(err)
}
results, exit, _ := RunDirectOnce(cfg, false)
if exit != 0 || len(results) != 1 || results[0].Status != "removed" {
t.Fatalf("exit=%d results=%+v", exit, results)
}
}
// 萃取失敗=該檔標 failed、exit=1、manifest 不標(下輪重試),其他檔不受影響。
func TestDirectExtractorFailKeepsRetry(t *testing.T) {
root := t.TempDir()
if err := os.WriteFile(filepath.Join(root, "a.md"), []byte("x"), 0o644); err != nil {
t.Fatal(err)
}
stubDir := t.TempDir()
stub := filepath.Join(stubDir, "claude")
if err := os.WriteFile(stub, []byte("#!/bin/sh\nexit 3\n"), 0o755); err != nil {
t.Fatal(err)
}
cfg := &DirectConfig{
WatchFolders: []string{root},
Manifest: filepath.Join(t.TempDir(), "m.json"),
CypherURL: "https://x.example", Namespace: "demo", APIKey: "demo",
Extractor: "claude", ClaudeBin: stub, MaxRemoved: DefaultMaxRemovedRatio,
}
results, exit, _ := RunDirectOnce(cfg, false)
if exit != 1 || len(results) != 1 || results[0].Status != "failed" {
t.Fatalf("exit=%d results=%+v", exit, results)
}
// 再跑一輪:仍是同一個事件(manifest 沒標 ingested=會重試)
results2, _, _ := RunDirectOnce(cfg, false)
if len(results2) != 1 {
t.Fatalf("失敗檔應重試:%+v", results2)
}
}
// t108 Test BmakeAccountSubConfig 必須繼承機器層 Extractor/GeminiAPIKey/CardIngestWF 等,
// 帳號層(AccountConfig)無這些欄位時一律繼承機器層——驗收到 rag_ingest_card 而非 rag_ingest_direct。
func TestMultiAccountInheritsExtractor(t *testing.T) {
root := t.TempDir()
if err := os.WriteFile(filepath.Join(root, "doc.md"), []byte("# 知識"), 0o644); err != nil {
t.Fatal(err)
}
// stub claude:輸出一張最簡卡片
stubDir := t.TempDir()
stub := filepath.Join(stubDir, "claude")
script := "#!/bin/sh\nmkdir -p system-dev/wiki/cards\nprintf '# doc\\n## 定義\\n測試\\n## 關聯\\n- doc >> 屬於 >> kb\\n' > 'system-dev/wiki/cards/doc.md'\n"
if err := os.WriteFile(stub, []byte(script), 0o755); err != nil {
t.Fatal(err)
}
var hitCard, hitDirect bool
srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if strings.Contains(r.URL.Path, "rag_ingest_card") {
hitCard = true
} else {
hitDirect = true
}
_ = json.NewEncoder(w).Encode(map[string]any{"success": true})
}))
defer srv.Close()
// 機器層有 Extractor+ClaudeBin;帳號層 AccountConfig 無這些欄位(正是 t108 場景)
cfg := &DirectConfig{
Manifest: filepath.Join(t.TempDir(), "m.json"),
Library: "kb",
Extractor: "claude",
ClaudeBin: stub,
CardIngestWF: "rag_ingest_card",
IngestWF: "rag_ingest_direct",
RemovedWF: "rag_takedown_direct",
MaxRemoved: DefaultMaxRemovedRatio,
Accounts: []AccountConfig{{
CypherURL: srv.URL,
Namespace: "demo",
APIKey: "demo",
WatchFolders: []string{root},
}},
}
results, exit, _ := RunDirectOnce(cfg, false)
if exit != 0 {
t.Fatalf("exit=%d results=%+v", exit, results)
}
if hitDirect {
t.Error("不應打 rag_ingest_direct(原文不出機,違反四步定稿)")
}
if !hitCard {
t.Error("應打 rag_ingest_card(機器層 extractor=claude 應被帳號繼承)")
}
}
// t108 Test Cextractor 空時,非 .md/.txt 檔禁止直送——標 failed 且絕不打任何 ingest 端點。
func TestExtractorEmptyBlocksNonTextDirect(t *testing.T) {
root := t.TempDir()
if err := os.WriteFile(filepath.Join(root, "report.pdf"), []byte("%PDF-1.4 機密原文"), 0o644); err != nil {
t.Fatal(err)
}
var serverCalled bool
srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
serverCalled = true
_ = json.NewEncoder(w).Encode(map[string]any{"success": true})
}))
defer srv.Close()
cfg := &DirectConfig{
WatchFolders: []string{root},
Manifest: filepath.Join(t.TempDir(), "m.json"),
CypherURL: srv.URL, Namespace: "demo", APIKey: "demo",
Library: "kb",
Extractor: "", // 舊制直送模式
IngestWF: "rag_ingest_direct",
RemovedWF: "rag_takedown_direct",
MaxRemoved: DefaultMaxRemovedRatio,
}
results, exit, _ := RunDirectOnce(cfg, false)
if exit != 1 {
t.Fatalf("exit=%d,應是 1(非文字檔無萃取器=失敗)", exit)
}
if serverCalled {
t.Error("防禦閘失效:PDF 被直送上雲(契約破壞)")
}
if len(results) != 1 || results[0].Status != "failed" {
t.Fatalf("results=%+v", results)
}
if !strings.Contains(results[0].Error, "萃取器未設定") {
t.Errorf("錯誤訊息不符:%q", results[0].Error)
}
}