feat: add support for parameters / end_user_metadata

examples/vertex_ai_conversation/evaluation_tool__autoeval__colab.ipynb

@@ -184,19 +184,21 @@
         "- `query` _(the input question)_\n",
         "- `expected_answer` _(the ideal or ground truth answer)_\n",
         "- `expected_uri` _(the webpage url or more generally the uri that contains the answer to `query`)_.\n",
+        "- `user_metadata` _(optional user metadata passed to datastore route with user query. Can be one of [str, dict])_\n",
+        "- `parameters` _(optional session parameters to include with the user query. Can be one of [str, dict])_\n",
         "\n",
         "In addition to the required columns the RougeL metric can also use the following optional column:\n",
         "\n",
         "- `golden_snippet` _(the extracted document snippet or segment that contains the `expected_answer`)_\n",
         "\n",
         "An example for the queryset can be seen in this table:\n",
         "\n",
-        "| conversation_id | turn_index | query | expected_answer | expected_uri |\n",
-        "| --- | --- | --- | --- | --- |\n",
-        "| 0 | 1 | What is the capital of France? | Capital of France is Paris. | exampleurl.com/france |\n",
-        "| 0 | 2 | How many people live there? | 2.1 million people live in Paris. | exampleurl.com/paris |\n",
-        "| 1 | 1 | What is the color of the sky? | It is blue. | exampleurl.com/common |\n",
-        "| 2 | 1 | How many legs does an octopus have? | It has 8 limbs. | exampleurl.com/octopus |\n",
+        "| conversation_id | turn_index | query | expected_answer | expected_uri | user_metadata | parameters\n",
+        "| --- | --- | --- | --- | --- | --- | --- |\n",
+        "| 0 | 1 | What is the capital of France? | Capital of France is Paris. | exampleurl.com/france | None | None |\n",
+        "| 0 | 2 | How many people live there? | 2.1 million people live in Paris. | exampleurl.com/paris | {\"some_end_user_key\": \"some_value\"} | {\"param1\": 1, \"param2\": \"some_string\"} |\n",
+        "| 1 | 1 | What is the color of the sky? | It is blue. | exampleurl.com/common | None | None |\n",
+        "| 2 | 1 | How many legs does an octopus have? | It has 8 limbs. | exampleurl.com/octopus | None | None |\n",
         "\n",
         "---\n",
         "\n",
@@ -224,14 +226,14 @@
       "source": [
         "# @markdown `run this cell to load data manually`\n",
         "INPUT_SCHEMA_REQUIRED_COLUMNS = [\n",
-        "    \"conversation_id\", \"turn_index\", \"query\", \"expected_answer\", \"expected_uri\", \"user_metadata\"\n",
+        "    \"conversation_id\", \"turn_index\", \"query\", \"expected_answer\", \"expected_uri\", \"user_metadata\", \"parameters\"\n",
         "]\n",
         "\n",
         "sample_df = pd.DataFrame(columns=INPUT_SCHEMA_REQUIRED_COLUMNS)\n",
         "\n",
-        "sample_df.loc[0] = [\"0\", 1 ,\"Who are you?\", \"I am an assistant\", \"www.google.com\", None]\n",
-        "sample_df.loc[1] = [\"1\", 1 ,\"Which is the cheapest plan?\", \"Basic plan\", \"www.google.com\", None]\n",
-        "sample_df.loc[2] = [\"1\", 2, \"How much?\", \"The Basic plan costs 20$/month\", \"www.google.com\", None]\n",
+        "sample_df.loc[0] = [\"0\", 1 ,\"Who are you?\", \"I am an assistant\", \"www.google.com\", None, None]\n",
+        "sample_df.loc[1] = [\"1\", 1 ,\"Who is the CEO?\", \"The CEO is Matt Reinjes.\", \"www.yeti.com\", {\"some_end_user_key\": \"some_value\"}, {\"param1\": 1, \"param2\": \"some_string\"}]\n",
+        "sample_df.loc[2] = [\"1\", 2, \"How much?\", \"The Basic plan costs 20$/month\", \"www.google.com\", None, None]\n",
         "queryset = sample_df"
       ]
     },

src/dfcx_scrapi/tools/datastore_scraper.py

@@ -36,6 +36,7 @@
     "expected_answer",
     "expected_uri",
     "user_metadata",
+    "parameters"
 ]
 
 def load_spreadsheet(
@@ -191,23 +192,33 @@ def scrape_detect_intent(
         query: str,
         session_id: Union[str, None] = None,
         user_metadata: Union[str, None] = None,
+        parameters: Union[str, None] = None
     ) -> AgentResponse:
         if session_id is None:
             session_id = self.sessions.build_session_id(self.agent_id)
 
         if user_metadata:
             try:
-                user_metadata = json.loads(user_metadata)
+                if isinstance(user_metadata, str):
+                    user_metadata = json.loads(user_metadata)
             except ValueError as err:
                 raise UserWarning("Invalid user metadata") from err
 
+        if parameters:
+            try:
+                if isinstance(parameters, str):
+                    parameters = json.loads(parameters)
+            except ValueError as err:
+                raise UserWarning("Invalid parameters") from err
+
         response = self.sessions.detect_intent(
             agent_id=self.agent_id,
             session_id=session_id,
             text=query,
             language_code=self.language_code,
             end_user_metadata=user_metadata,
             populate_data_store_connection_signals=True,
+            parameters=parameters
         )
 
         ar = AgentResponse()
@@ -224,7 +235,10 @@ def run(
 
         def scrape(row):
             result = self.scrape_detect_intent(
-                row["query"], row["session_id"], row["user_metadata"]
+                query=row["query"],
+                session_id=row["session_id"],
+                user_metadata=row["user_metadata"],
+                parameters=row["parameters"]
             )
             progress_bar.update()