@chainlesschain/personal-data-hub 0.2.1 → 0.2.3

package/lib/adapters/wechat/frida-agent/wechat-key-hook.js

@@ -32,7 +32,10 @@
 "use strict";
 
 (function () {
-  var TARGET_MODULE = "libwcdb.so";
+  // sjqz-verified module name is `libWCDB.so` (uppercase); some WeChat
+  // builds ship lowercase. Try both — first match wins, no extra cost
+  // because Process.findModuleByName is a cheap lookup.
+  var TARGET_MODULES = ["libWCDB.so", "libwcdb.so"];
   // Primary symbol per §18.3. Add fallbacks below — version drift will
   // shift the export name; host treats first hit as authoritative.
   var SYMBOLS = [
@@ -60,63 +63,182 @@
   // the host detaches quickly (anti-detection §18.6 #4).
   var fired = false;
 
+  // Sig-aware arg index map. The host treats the first 'key' event as
+  // authoritative, so picking the wrong index for v2 = host gets the
+  // database NAME pointer (e.g. "main") and DB opens fail silently.
+  //   sqlite3_key(sqlite3 *db, const void *pKey, int nKey)
+  //     args[0]=db,            args[1]=key, args[2]=len
+  //   sqlite3_key_v2(sqlite3 *db, const char *zDbName, const void *pKey, int nKey)
+  //     args[0]=db, args[1]=name, args[2]=key, args[3]=len
+  //   wcdb_setkey / WCDBKeyDerive: unknown sig — assume sqlite3_key shape
+  //   Mangled C++: WCDB::Database::setCipherKey(*this, const std::string&)
+  //     args[0]=this, args[1]=&string (length needs .size()) — not handled
+  //     here; emit error so the host falls back to MD5 path.
+  function argIndicesFor(symbolName) {
+    if (symbolName === "sqlite3_key_v2") {
+      return { key: 2, len: 3, sig: "v2" };
+    }
+    if (symbolName.indexOf("_ZN4WCDB") === 0) {
+      return { key: -1, len: -1, sig: "mangled-cpp" };
+    }
+    return { key: 1, len: 2, sig: "v1" };
+  }
+
+  // sjqz extract_wechat_key.py uses Memory.readCString(args[1]) for the
+  // key — meaning some WeChat builds pass the key as a NUL-terminated
+  // 64-char ASCII hex string. Other builds (and the original SQLCipher
+  // contract) pass 32 raw bytes. We can disambiguate by `len`:
+  //   - len === 32 → raw 32-byte key → readByteArray + bytesToHex
+  //   - len === 64 → ASCII hex string → readCString
+  //   - anything else → emit error, host falls back to MD5 path
   function makeHook(symbolName) {
+    var idx = argIndicesFor(symbolName);
     return {
       onEnter: function (args) {
         if (fired) return;
+        if (idx.key < 0) {
+          send({
+            kind: "error",
+            message:
+              "unsupported symbol signature: " +
+              symbolName +
+              " — host should fall back to MD5(IMEI+UIN) key path",
+          });
+          return;
+        }
         try {
-          // sqlite3_key signature: int sqlite3_key(sqlite3 *db, const void *pKey, int nKey)
-          // args[1] = key bytes, args[2] = key length
-          var len = args[2].toInt32();
+          var len = args[idx.len].toInt32();
           if (len <= 0 || len > 256) {
-            send({ kind: "error", message: "implausible key length " + len + " at " + symbolName });
+            send({
+              kind: "error",
+              message:
+                "implausible key length " + len + " at " + symbolName,
+            });
+            return;
+          }
+          var hex;
+          var format;
+          if (len === 64) {
+            // ASCII hex string (sjqz-verified path on WeChat 7.x/8.0 libWCDB)
+            var s = Memory.readCString(args[idx.key], len);
+            if (!s || s.length === 0) {
+              send({
+                kind: "error",
+                message: "readCString returned empty at " + symbolName,
+              });
+              return;
+            }
+            hex = s.toLowerCase();
+            format = "ascii-hex";
+          } else if (len === 32) {
+            // Raw 32-byte key — convert to 64-char hex
+            var buf = args[idx.key].readByteArray(len);
+            hex = bytesToHex(buf);
+            format = "raw-bytes";
+          } else {
+            // Ambiguous length — could be either. Emit both interpretations
+            // and let the host try each against the DB until one succeeds.
+            var bufAmb = args[idx.key].readByteArray(len);
+            var hexFromBytes = bytesToHex(bufAmb);
+            var hexFromString = null;
+            try {
+              var sAmb = Memory.readCString(args[idx.key], len);
+              if (sAmb) hexFromString = sAmb.toLowerCase();
+            } catch (_e) {
+              // readCString may fault on non-NUL-terminated bytes; ignore.
+            }
+            fired = true;
+            send({
+              kind: "key",
+              hex: hexFromBytes,
+              alt: hexFromString,
+              source: symbolName,
+              sig: idx.sig,
+              format: "ambiguous",
+              length: len,
+            });
             return;
           }
-          var buf = args[1].readByteArray(len);
-          var hex = bytesToHex(buf);
           if (!hex) {
-            send({ kind: "error", message: "empty key buffer at " + symbolName });
+            send({
+              kind: "error",
+              message: "empty key buffer at " + symbolName,
+            });
             return;
           }
           fired = true;
-          send({ kind: "key", hex: hex, source: symbolName });
+          send({
+            kind: "key",
+            hex: hex,
+            source: symbolName,
+            sig: idx.sig,
+            format: format,
+            length: len,
+          });
         } catch (e) {
-          send({ kind: "error", message: "hook exception at " + symbolName + ": " + (e && e.message ? e.message : String(e)) });
+          send({
+            kind: "error",
+            message:
+              "hook exception at " +
+              symbolName +
+              ": " +
+              (e && e.message ? e.message : String(e)),
+          });
         }
       },
     };
   }
 
-  function tryAttach() {
-    var mod = Process.findModuleByName(TARGET_MODULE);
+  function tryAttachOnModule(moduleName) {
+    var mod = Process.findModuleByName(moduleName);
     if (!mod) return false;
     var attached = 0;
     for (var i = 0; i < SYMBOLS.length; i++) {
-      var addr = Module.findExportByName(TARGET_MODULE, SYMBOLS[i]);
+      var addr = Module.findExportByName(moduleName, SYMBOLS[i]);
       if (!addr) continue;
       try {
         Interceptor.attach(addr, makeHook(SYMBOLS[i]));
-        send({ kind: "hooked", symbol: SYMBOLS[i], module: TARGET_MODULE });
+        send({ kind: "hooked", symbol: SYMBOLS[i], module: moduleName });
         attached++;
       } catch (e) {
-        send({ kind: "error", message: "Interceptor.attach failed for " + SYMBOLS[i] + ": " + (e && e.message ? e.message : String(e)) });
+        send({
+          kind: "error",
+          message:
+            "Interceptor.attach failed for " +
+            SYMBOLS[i] +
+            ": " +
+            (e && e.message ? e.message : String(e)),
+        });
       }
     }
     return attached > 0;
   }
 
+  function tryAttach() {
+    for (var i = 0; i < TARGET_MODULES.length; i++) {
+      if (tryAttachOnModule(TARGET_MODULES[i])) {
+        return true;
+      }
+    }
+    return false;
+  }
+
   // Module-load polling — §18.6 #1 "hook at module-load time before
-  // anti-detection thread runs". WeChat lazy-loads libwcdb when the
+  // anti-detection thread runs". WeChat lazy-loads libWCDB when the
   // first DB opens, so we can't always find it at script start.
   if (!tryAttach()) {
-    send({ kind: "module-waiting", module: TARGET_MODULE });
+    send({ kind: "module-waiting", module: TARGET_MODULES.join("|") });
     var attempts = 0;
     var poll = function () {
       attempts++;
       if (tryAttach()) return;
       if (attempts >= 60) {
         // 60 attempts × 500ms = 30s ceiling, matches host timeoutMs
-        send({ kind: "error", message: TARGET_MODULE + " did not load within 30s" });
+        send({
+          kind: "error",
+          message:
+            TARGET_MODULES.join("|") + " did not load within 30s",
+        });
         return;
       }
       setTimeout(poll, 500);

package/lib/adapters/wechat/key-providers/frida-key-provider.js

@@ -191,6 +191,14 @@ class FridaKeyProvider extends KeyProvider {
         if (evt.kind === "key") {
           settled = true;
           telemetry.keySource = evt.source;
+          // Phase 12.6 (post-sjqz audit) — capture sig/format/length so a
+          // failed DB open can be diagnosed: ascii-hex vs raw-bytes
+          // determines whether sqlite3_key got the expected key bytes,
+          // and sig=v1/v2 confirms args index resolution.
+          telemetry.keyFormat = evt.format || null;
+          telemetry.keySig = evt.sig || null;
+          telemetry.keyLength = evt.length || null;
+          telemetry.keyAlt = evt.alt || null;
           telemetry.durationMs = Date.now() - telemetry.startedAt;
           cleanup().then(() => resolve(String(evt.hex || "").toLowerCase()));
           return;

package/lib/adapters/wechat/normalize.js

@@ -203,11 +203,20 @@ function contactDisplayName(byUsername, wxid) {
 function guessContactSubtype(row) {
   // rcontact.type bits: official accounts / group / regular contact /
   // black list. Detailed mapping in WeChat reverse-eng community —
-  // for v0.5 we keep it simple: anything that's not the user's self is
-  // "contact". Phase 12.6 will refine with full bit mapping.
-  if (typeof row.username === "string" && row.username.endsWith("@chatroom")) {
+  // for v0.5 we keep it simple: chatroom → unknown (not a Person),
+  // `gh_*` username → merchant (公众号 / Official Account — brand /
+  // business pushing content; closest enum match), rest → contact.
+  // Phase 12.6 will refine with full bit mapping + rcontact.type bits.
+  // (sjqz parity wechat.py:282 — get_friends() excludes gh_* from
+  // friends view but keeps them in contacts; we keep as Person with
+  // distinct subtype so Ask flow / EntityResolver can filter cleanly.)
+  if (typeof row.username !== "string") return "contact";
+  if (row.username.endsWith("@chatroom")) {
     return "unknown"; // chat group, not a Person
   }
+  if (row.username.startsWith("gh_")) {
+    return "merchant"; // 公众号 / Official Account
+  }
   return "contact";
 }
 

package/lib/index.js

@@ -39,10 +39,12 @@ const { Train12306Adapter } = require("./adapters/travel-12306");
 const { CtripAdapter } = require("./adapters/travel-ctrip");
 const { AmapAdapter } = require("./adapters/travel-amap");
 const { BaiduMapAdapter } = require("./adapters/travel-baidu-map");
+const { TencentMapAdapter } = require("./adapters/travel-tencent-map");
 const shoppingBase = require("./adapters/shopping-base");
 const { TaobaoAdapter } = require("./adapters/shopping-taobao");
 const { JdAdapter } = require("./adapters/shopping-jd");
 const { MeituanAdapter } = require("./adapters/shopping-meituan");
+const { PinduoduoAdapter } = require("./adapters/shopping-pinduoduo");
 const { BilibiliAdapter } = require("./adapters/social-bilibili");
 const { WeiboAdapter } = require("./adapters/social-weibo");
 const { DouyinAdapter } = require("./adapters/social-douyin");
@@ -221,13 +223,14 @@ module.exports = {
   wxidToWeChatPersonId: wechatAdapter.wxidToWeChatPersonId,
   WECHAT_PRAGMA_PROFILES: wechatAdapter.WECHAT_PRAGMA_PROFILES,
 
-  // Phase 9 — Travel four-pack
+  // Phase 9 + §2.5b 地图三联 — Travel five-pack
   normalizeTravelRecord: travelBase.normalizeTravelRecord,
   parseChineseDateTime: travelBase.parseChineseDateTime,
   Train12306Adapter,
   CtripAdapter,
   AmapAdapter,
   BaiduMapAdapter,
+  TencentMapAdapter,
 
   // Phase 7 — Shopping three-pack
   normalizeOrderRecord: shoppingBase.normalizeOrderRecord,
@@ -235,6 +238,7 @@ module.exports = {
   TaobaoAdapter,
   JdAdapter,
   MeituanAdapter,
+  PinduoduoAdapter,
 
   // Phase 13+ — long-tail social + messaging (借 sjqz parsers)
   BilibiliAdapter,

package/lib/vault.js

@@ -44,19 +44,71 @@ function newGroupId() {
   return `mg-${r()}${r()}-${Date.now().toString(36)}`;
 }
 
+/**
+ * Translate a bs3mc load-failure error into an actionable, user-readable
+ * message. Detects NODE_MODULE_VERSION mismatch (the single most common
+ * failure: Node 23/24/25 has no prebuild — bs3mc upstream only ships for
+ * Node LTS ABIs 108/115/127). See memory `node_23_native_dep_trap.md`.
+ *
+ * Pure function so it can be unit-tested without poisoning require cache.
+ *
+ * @param {Error|unknown} err Original throw from `require("better-sqlite3-multiple-ciphers")`.
+ * @param {string} [nodeVer] process.versions.node (override for tests).
+ * @returns {Error} Wrapped Error with `cause` and (when ABI-related) `code: "BS3MC_ABI_MISMATCH"`.
+ */
+function formatDriverLoadError(err, nodeVer) {
+  const originalMsg = err && err.message ? err.message : String(err);
+  const runtimeNodeVer = nodeVer || process.versions.node;
+
+  const abiMatch = originalMsg.match(
+    /NODE_MODULE_VERSION\s+(\d+)[\s\S]+?requires\s+NODE_MODULE_VERSION\s+(\d+)/,
+  );
+  if (abiMatch) {
+    const compiledAbi = abiMatch[1];
+    const runtimeAbi = abiMatch[2];
+    const lines = [
+      "better-sqlite3-multiple-ciphers ABI mismatch — Node " +
+        runtimeNodeVer +
+        " has ABI " +
+        runtimeAbi +
+        " but bs3mc prebuild is ABI " +
+        compiledAbi +
+        ".",
+      "",
+      "修法（任选其一）：",
+      "  1. 切 Node 22 LTS (推荐) — nvm-windows: `nvm install 22.12.0 && nvm use 22.12.0`",
+      "  2. 源码重编 — `npm rebuild better-sqlite3-multiple-ciphers --build-from-source`",
+      "     （需要本机有 Visual Studio Build Tools / node-gyp toolchain，慢且不推荐）",
+      "",
+      "为什么 bs3mc 没 ABI " + runtimeAbi + " prebuild：",
+      "  bs3mc 上游只 ship 主流 Node LTS 的 prebuild (ABI 108/115/127)。",
+      "  Node 23/24/25 是 Current 系列，上游不给 prebuild。",
+      "",
+      "项目 engines.node 允许 >=22.12 是为了兼容未来 LTS，但实际推荐 22.x。",
+    ];
+    const wrapped = new Error(lines.join("\n"));
+    wrapped.cause = err;
+    wrapped.code = "BS3MC_ABI_MISMATCH";
+    return wrapped;
+  }
+
+  const wrapped = new Error(
+    "Failed to load better-sqlite3-multiple-ciphers. " +
+      "Install it as a workspace dep or pin the version in your package. " +
+      "Original error: " +
+      originalMsg,
+  );
+  wrapped.cause = err;
+  return wrapped;
+}
+
 function loadDriver() {
   // Lazy require so consumers that only need schemas don't pay for the
   // native binding load. Errors surface here with a precise message.
   try {
     return require("better-sqlite3-multiple-ciphers");
   } catch (err) {
-    const msg =
-      "Failed to load better-sqlite3-multiple-ciphers. " +
-      "Install it as a workspace dep or pin the version in your package. " +
-      "Original error: " + (err && err.message ? err.message : String(err));
-    const wrapped = new Error(msg);
-    wrapped.cause = err;
-    throw wrapped;
+    throw formatDriverLoadError(err);
   }
 }
 
@@ -1223,4 +1275,4 @@ class LocalVault {
   }
 }
 
-module.exports = { LocalVault };
+module.exports = { LocalVault, _internal: { loadDriver, formatDriverLoadError } };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@chainlesschain/personal-data-hub",
-  "version": "0.2.1",
+  "version": "0.2.3",
   "description": "Personal Data Hub — UnifiedSchema + validators + KG ingest helpers for the data-back-to-the-individual middleware",
   "type": "commonjs",
   "main": "lib/index.js",
@@ -46,6 +46,7 @@
     "./adapters/travel-ctrip": "./lib/adapters/travel-ctrip/index.js",
     "./adapters/travel-amap": "./lib/adapters/travel-amap/index.js",
     "./adapters/travel-baidu-map": "./lib/adapters/travel-baidu-map/index.js",
+    "./adapters/travel-tencent-map": "./lib/adapters/travel-tencent-map/index.js",
     "./adapters/shopping-base": "./lib/adapters/shopping-base/index.js",
     "./adapters/shopping-taobao": "./lib/adapters/shopping-taobao/index.js",
     "./adapters/shopping-jd": "./lib/adapters/shopping-jd/index.js",