diff --git a/sample_annotator/clients/gold_client.py b/sample_annotator/clients/gold_client.py
index 75da84b..cfd0201 100644
--- a/sample_annotator/clients/gold_client.py
+++ b/sample_annotator/clients/gold_client.py
@@ -13,6 +13,11 @@
 
 from requests.auth import HTTPBasicAuth
 
+import hashlib
+import pickle
+
+cache = None
+
 
 USERPASS = Tuple[str, str]
 URL = str
@@ -33,10 +38,32 @@
 # but leaving this in as a stub in case this happens again
 EXCLUSION_LIST = []
 
+def build_cache_key(endpoint_url: str, params: dict, user: str, passwd: str) -> str:
+    normalized_params = tuple(sorted(params.items()))
+    key_data = (endpoint_url, normalized_params, user, passwd)
+    return key_data
+
+def ensure_cache_initialized():
+    global cache
+    if cache is None:
+        path = os.environ.get("GOLD_CACHE_DIR", "cachedir")
+        logging.warning(f"[cache] cache was None. Falling back to: {path}")
+        cache = Cache(path)
+
+def set_cache_directory(path: str):
+    """
+    Override the default cache directory globally.
+    Must be called before any memoized function is used.
+    """
+    global cache
+    # Clear any previous cache instance
+    if cache is not None:
+        cache.close()
+    cache = Cache(path)
+    logging.info(f"Cache directory set to: {path}")
 
-@cache.memoize()
-@cache.memoize()
 def _fetch_url(endpoint_url, params, user, passwd) -> JSON:
+    logging.warning(f"[cache] API call made: {endpoint_url} {params}")
     attempt = 0
     while attempt < 4:
         results = requests.get(
@@ -53,6 +80,22 @@ def _fetch_url(endpoint_url, params, user, passwd) -> JSON:
             attempt += 1
     raise Exception(f"API call to {endpoint_url} failed after {attempt} attempts")
 
+def get_fetch_url():
+    ensure_cache_initialized()
+
+    @cache.memoize()
+    def _memoized_fetch_url(endpoint_url, params, user, passwd):
+        key_data = build_cache_key(endpoint_url, params, user, passwd)
+        key_bytes = pickle.dumps(key_data)
+        hashed_key = hashlib.sha256(key_bytes).hexdigest()
+
+        study_id = params.get("studyGoldId", "<none>")
+        logging.warning(f"[cache] memoize key: {hashed_key} for studyGoldId={study_id}")
+
+        return _fetch_url(endpoint_url, params, user, passwd)
+
+    return _memoized_fetch_url
+
 
 class GoldClient:
     """
@@ -91,7 +134,23 @@ def _normalize_id(self, id: str) -> str:
     def _call(self, endpoint: str, params: Dict = {}) -> JSON:
         (user, passwd) = self.gold_key
         endpoint_url = f"{self.url}/{endpoint}"
-        obj = _fetch_url(endpoint_url, params, user, passwd)
+
+        # Try direct cache lookup first
+        key_data = build_cache_key(endpoint_url, params, user, passwd)
+        if cache and key_data in cache:
+            logging.info(f"[cache] Direct cache hit for {endpoint} {params}")
+            return cache.get(key_data)
+
+        # Also try with simpler key format that might be in cache
+        study_id = params.get("studyGoldId", "")
+        if study_id:
+            simple_key = f"{endpoint_url}-{study_id}-{user}-{passwd}"
+            if cache and simple_key in cache:
+                logging.info(f"[cache] Simple key cache hit for {endpoint} {params}")
+                return cache.get(simple_key)
+
+        # Fall back to memoized function if not found directly
+        obj = get_fetch_url()(endpoint_url, params, user, passwd)
         self.num_calls += 1
         return obj
 
@@ -109,7 +168,7 @@ def fetch_projects_by_study(self, id: str) -> List[SampleDict]:
         return results
 
     def fetch_biosamples_by_study(
-        self, id: str, include_project=True
+            self, id: str, include_project=True
     ) -> List[SampleDict]:
         """
         Fetches all samples for a study
@@ -124,6 +183,8 @@ def fetch_biosamples_by_study(
         else:
             biosamples = self._call("biosamples", {"studyGoldId": id})
             if include_project:
+                logging.info(f"Cache contains {len(cache) if cache else 0} records")
+                logging.info(f"Cache directory is at {cache.directory if cache else 'None'}")
                 projects = self.fetch_projects_by_study(id)
                 # weave projects in samples
                 samples_by_id = {