Add Parameter for detect intent

examples/vertex_ai_conversation/evaluation_tool__autoeval__colab.ipynb

@@ -272,6 +272,7 @@
         "\n",
         "GLOBAL_SCOPE = [\"https://spreadsheets.google.com/feeds\"]\n",
         "\n",
+        "PARAMETERS = \"parameters\"\n",
         "CONVERSATION_ID = \"conversation_id\"\n",
         "TURN_INDEX = \"turn_index\"\n",
         "QUERY = \"query\"\n",
@@ -285,7 +286,7 @@
         "AGENT_URI = \"projects/{project_id}/locations/{location}/agents/{agent_id}\"\n",
         "\n",
         "INPUT_SCHEMA_REQUIRED_COLUMNS = [\n",
-        "    CONVERSATION_ID, TURN_INDEX, QUERY, REFERENCE, EXPECTED_URI, USER_METADATA\n",
+        "    PARAMETERS,CONVERSATION_ID, TURN_INDEX, QUERY, REFERENCE, EXPECTED_URI, USER_METADATA\n",
         "]\n",
         "\n",
         "_EXECUTION_SEQUENCE_KEY = \"DataStore Execution Sequence\"\n",
@@ -783,6 +784,7 @@
         "    query_param_mapping = {}\n",
         "\n",
         "    if parameters:\n",
+        "      parameters = json.loads(parameters)\n",
         "      query_param_mapping[\"parameters\"] = parameters\n",
         "\n",
         "    if end_user_metadata:\n",
@@ -807,6 +809,7 @@
         "      self,\n",
         "      query: str,\n",
         "      session_id: str | None = None,\n",
+        "      parameters: str | None = None,\n",
         "      user_metadata: str | None = None,\n",
         "      ) -> VertexConversationResponse:\n",
         "    if session_id is None:\n",
@@ -825,6 +828,7 @@
         "        language_code=self.language_code,\n",
         "        end_user_metadata=user_metadata,\n",
         "        populate_data_store_connection_signals=True,\n",
+        "        parameters=parameters,\n",
         "    )\n",
         "    return VertexConversationResponse.from_query_result(response._pb)\n",
         "\n",
@@ -2251,6 +2255,7 @@
       "outputs": [],
       "source": [
         "# test agent on a single query\n",
+        "parameter_str = '{\"channel\" : \"chat\"}'\n",
         "response = scraper.scrape_detect_intent(query=\"who is the ceo?\")\n",
         "print(json.dumps(dataclasses.asdict(response), indent=4))"
       ]
@@ -2264,24 +2269,26 @@
         "## Data Loading\n",
         "\n",
         "The queryset must be in a tabular format that has to contain the following columns:\n",
+        "- `parameters` _(parameter which needs to be passed as part of detect intent for agent to route to correct flow/page)_\n",
         "- `conversation_id` _(unique identifier of a conversation, which must be the same for each row that are part of the same conversation)_\n",
         "- `turn_index` _(index of the query - expected answer pair within a conversation)_\n",
         "- `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` _(Dict of CX Session endUserMetadata to set in the conversation.)_.\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",
+        "|  parameters | conversation_id | turn_index | query | expected_answer | expected_uri | user_metadata |\n",
+        "| --- | --- | --- | --- | --- | --- | --- |\n",
+        "| {\"channel\" : \"chat\"} | 0 | 1 | What is the capital of France? | Capital of France is Paris. | exampleurl.com/france | None |\n",
+        "| {\"channel\" : \"chat\"} | 0 | 2 | How many people live there? | 2.1 million people live in Paris. | exampleurl.com/paris | None |\n",
+        "| {\"channel\" : \"chat\"} | 1 | 1 | What is the color of the sky? | It is blue. | exampleurl.com/common | None |\n",
+        "| {\"channel\" : \"chat\"} | 2 | 1 | How many legs does an octopus have? | It has 8 limbs. | exampleurl.com/octopus | None |\n",
         "\n",
         "---\n",
         "\n",
@@ -2311,9 +2318,9 @@
         "\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$/month\", \"www.google.com\", None]\n",
+        "sample_df.loc[0] = ['{\"channel\" : \"chat\"}', \"0\", 1 ,\"Who are you?\", \"I am an assistant\", \"www.google.com\", None]\n",
+        "sample_df.loc[1] = ['{\"channel\" : \"chat\"}', \"1\", 1 ,\"Which is the cheapest plan?\", \"Basic plan\", \"www.google.com\", None]\n",
+        "sample_df.loc[2] = ['{\"channel\" : \"chat\"}', \"1\", 2, \"How much?\", \"The Basic plan costs 20$/month\", \"www.google.com\", None]\n",
         "queryset = sample_df"
       ]
     },