diff --git a/plugins/python_providers/resources/amda.py b/plugins/python_providers/resources/amda.py
index 353176c..9d350b9 100644
--- a/plugins/python_providers/resources/amda.py
+++ b/plugins/python_providers/resources/amda.py
@@ -27,8 +27,8 @@ def get_sample(metadata,start,stop):
                 elif value == 'multicomponent':
                     ts_type = pysciqlopcore.MultiComponentTimeSerie
                     default_ctor_args = (0,2)
-        tstart=datetime.datetime.fromtimestamp(start, tz=timezone.utc)
-        tend=datetime.datetime.fromtimestamp(stop, tz=timezone.utc)
+        tstart=datetime.fromtimestamp(start, tz=timezone.utc)
+        tend=datetime.fromtimestamp(stop, tz=timezone.utc)
         df = amda.get_parameter(start_time=tstart, stop_time=tend, parameter_id=param_id, method="REST")
         t = np.array([d.timestamp() for d in df.index])
         values = df.values
@@ -50,7 +50,7 @@ for name,component in amda.component.items():
 
 products = []
 for key,parameter in parameters.items():
-    path = f"/AMDA/{parameter['mission']}/{parameter['instrument']}/{parameter['dataset']}/{parameter['name']}" 
+    path = f"/AMDA/{parameter['mission']}/{parameter.get('observatory','')}/{parameter['instrument']}/{parameter['dataset']}/{parameter['name']}"
     components = [component['name'] for component in parameter.get('components',[])]
     metadata = [ (key,item) for key,item in parameter.items() if key is not 'components' ]
     n_components = parameter.get('size',0)
diff --git a/plugins/python_providers/resources/test.py b/plugins/python_providers/resources/test.py
index d236313..cf5a6dc 100644
--- a/plugins/python_providers/resources/test.py
+++ b/plugins/python_providers/resources/test.py
@@ -4,42 +4,84 @@ import PythonProviders
 import pysciqlopcore
 import numpy as np
 import math
+from spwc.cache import _cache
+from spwc.common.datetime_range import DateTimeRange
+from functools import partial
+from datetime import datetime, timedelta, timezone
 
 someglobal = 1
 
 def make_scalar(x):
     y = np.cos(x/10.)
-    return pysciqlopcore.ScalarTimeSerie(x,y)
+    return pds.DataFrame(index=[datetime.fromtimestamp(t, tz=timezone.utc) for t in x], data=y)
 
 def make_vector(x):
     v=np.ones((len(x),3))
     for i in range(3):
-        v.transpose()[:][i] = np.cos(x/10. + float(i))
-    return pysciqlopcore.VectorTimeSerie(x,v)
+        v.transpose()[:][i] = np.cos(x/10. + float(i)) + (100. * np.cos(x/10000. + float(i)))
+    return pds.DataFrame(index=[datetime.fromtimestamp(t, tz=timezone.utc) for t in x], data=v)
 
 
 def make_multicomponent(x):
     v=np.ones((len(x),4))
     for i in range(4):
         v.transpose()[:][i] = float(i+1) * np.cos(x/10. + float(i))
-    return pysciqlopcore.MultiComponentTimeSerie(x,v)
+    return pds.DataFrame(index=[datetime.fromtimestamp(t, tz=timezone.utc) for t in x], data=v)
 
 
-def get_data(metadata,start,stop):
+def _get_data(p_type, start, stop):
+    if type(start) is datetime:
+        start = start.timestamp()
+        stop = stop.timestamp()
     x = np.arange(math.ceil(start), math.floor(stop))
-    for key,value in metadata:
-        if key == 'xml:id':
-            param_id = value
-        elif key == 'type':
-            if value == 'vector':
-                return make_vector(x)
-            elif value == 'multicomponent':
-                return make_multicomponent(x)
-    return make_scalar(x)
-
-
+    if p_type == 'scalar':
+        return make_scalar(x)
+    if p_type == 'vector':
+        return make_vector(x)
+    if p_type == 'multicomponent':
+        return make_multicomponent(x)
+    return None
 
+def get_data(metadata,start,stop):
+    ts_type = pysciqlopcore.ScalarTimeSerie
+    default_ctor_args = 1
+    use_cache = False
+    p_type = 'scalar'
+    try:
+        for key,value in metadata:
+            if key == 'type':
+                p_type = value
+                if value == 'vector':
+                    ts_type = pysciqlopcore.VectorTimeSerie
+                elif value == 'multicomponent':
+                    ts_type = pysciqlopcore.MultiComponentTimeSerie
+                    default_ctor_args = (0,2)
+            if key == 'cache' and value == 'true':
+                use_cache = True
+        if use_cache:
+            cache_product = f"tests/{p_type}"
+            df = _cache.get_data(cache_product, DateTimeRange(datetime.fromtimestamp(start, tz=timezone.utc), datetime.fromtimestamp(stop, tz=timezone.utc)),
+                                     partial(_get_data, p_type),
+                                     fragment_hours=24)
+        else:
+            print("No Cache")
+            df = _get_data(p_type, start, stop)
+        t = np.array([d.timestamp() for d in df.index])
+        values = df.values
+        return ts_type(t,values)
+    except Exception as e:
+        print(traceback.format_exc())
+        print("Error in test.py ",str(e))
+        return ts_type(default_ctor_args)
 
-PythonProviders.register_product([("/tests/scalar",[],[("type","scalar")]), ("/tests/vector",[],[("type","vector")]), ("/tests/multicomponent",[],[("type","multicomponent"),('size','4')])],get_data)
+products = [
+    ("/tests/without_cache/scalar",[],[("type","scalar")]),
+    ("/tests/without_cache/vector",[],[("type","vector")]),
+    ("/tests/without_cache/multicomponent",[],[("type","multicomponent"),('size','4')]),
+    ("/tests/with_cache/scalar",[],[("type","scalar"), ("cache","true")]),
+    ("/tests/with_cache/vector",[],[("type","vector"), ("cache","true")]),
+    ("/tests/with_cache/multicomponent",[],[("type","multicomponent"),('size','4'), ("cache","true")])
+    ]
 
 
+PythonProviders.register_product(products ,get_data)