fix(dataflows): zero-pad HK ticker codes to Yahoo's 4-digit form (#1342)

- normalize_symbol maps 09992.HK to 9992.HK and 700.HK to 0700.HK (#957)

tests/test_symbol_utils.py

@@ -52,6 +52,23 @@ class TestNormalizeSymbol(unittest.TestCase):
     def test_empty_input_passthrough(self):
         self.assertEqual(normalize_symbol(""), "")
 
+    def test_hk_five_digit_code_repadded_to_four(self):
+        # HKEX lists up to 5-digit codes; Yahoo only accepts 4 (#957).
+        self.assertEqual(normalize_symbol("09992.HK"), "9992.HK")
+        self.assertEqual(normalize_symbol("00700.HK"), "0700.HK")
+        self.assertEqual(normalize_symbol("00001.HK"), "0001.HK")
+
+    def test_hk_four_digit_code_unchanged(self):
+        self.assertEqual(normalize_symbol("0700.HK"), "0700.HK")
+        self.assertEqual(normalize_symbol("9992.HK"), "9992.HK")
+        self.assertEqual(normalize_symbol("80737.HK"), "80737.HK")
+
+    def test_hk_short_code_padded_to_four(self):
+        self.assertEqual(normalize_symbol("700.HK"), "0700.HK")
+
+    def test_hk_code_case_insensitive_suffix(self):
+        self.assertEqual(normalize_symbol("09992.hk"), "9992.HK")
+
 
 @pytest.mark.unit
 class TestNoMarketDataError(unittest.TestCase):

tradingagents/dataflows/symbol_utils.py

@@ -10,6 +10,7 @@ differ from the broker / TradingView / MT5 style symbols users often type:
     EURUSD            EURUSD=X          spot forex pairs take a ``=X`` suffix
     BTCUSD            BTC-USD           crypto pairs use a ``-`` separator
     SPX500, US500     ^GSPC             index CFDs map to Yahoo index symbols
+    09992.HK, 700.HK  9992.HK, 0700.HK  HK codes are zero-padded to 4 digits
 
 Passing the raw broker symbol to Yahoo returns an empty result, which the
 agents previously received as free text and could hallucinate a price
@@ -72,6 +73,9 @@ _ALIASES = {
 # Yahoo symbols may contain letters, digits, and these structural characters.
 _YAHOO_SAFE = re.compile(r"^[A-Za-z0-9._\-\^=]+$")
 
+# HKEX codes as Yahoo spells them: the number zero-padded to 4 digits (#957).
+_HK_CODE = re.compile(r"^(\d{1,5})\.HK$")
+
 
 # Crypto quote currencies that all map to Yahoo's USD pair. Yahoo lists only
 # ``<BASE>-USD`` (not the USDT/USDC stablecoin pairs), so a broker symbol quoted
@@ -109,7 +113,9 @@ def normalize_symbol(raw: str) -> str:
       2. Crypto rule: a known crypto base quoted in USD/USDT/USDC (dashed or
          not) -> ``BASE-USD``.
       3. Forex rule: six letters that are two ISO currency codes -> ``PAIR=X``.
-      4. Otherwise the upper-cased symbol is returned unchanged (plain
+      4. HK rule: a numeric ``.HK`` code -> Yahoo's 4-digit padding
+         (``09992.HK`` -> ``9992.HK``, ``700.HK`` -> ``0700.HK``).
+      5. Otherwise the upper-cased symbol is returned unchanged (plain
          equities, ETFs, Yahoo-native symbols like ``GC=F`` or ``^GSPC``).
 
     A trailing ``+`` (broker CFD marker, e.g. ``XAUUSD+``) is stripped before
@@ -130,6 +136,8 @@ def normalize_symbol(raw: str) -> str:
         canonical = crypto
     elif len(s) == 6 and s[:3] in _FOREX_CURRENCIES and s[3:] in _FOREX_CURRENCIES:
         canonical = f"{s}=X"
+    elif hk := _HK_CODE.match(s):
+        canonical = f"{int(hk.group(1)):04d}.HK"
     else:
         canonical = s