feat: use websocket for ai assistant (pinterest#1311)

* feat: use websocket for ai assistant

* fix node test

* comments

* remove memory and add keep button

* fix linter

querybook/server/const/ai_assistant.py

@@ -0,0 +1,14 @@
+from enum import Enum
+
+
+# KEEP IT CONSISTENT AS webapp/const/aiAssistant.ts
+class AICommandType(Enum):
+    SQL_FIX = "SQL_FIX"
+    SQL_TITLE = "SQL_TITLE"
+    TEXT_TO_SQL = "TEXT_TO_SQL"
+    RESET_MEMORY = "RESET_MEMORY"
+
+
+AI_ASSISTANT_NAMESPACE = "/ai_assistant"
+AI_ASSISTANT_REQUEST_EVENT = "ai_assistant_request"
+AI_ASSISTANT_RESPONSE_EVENT = "ai_assistant_response"

querybook/server/datasources/__init__.py

@@ -16,7 +16,6 @@
 from . import event_log
 from . import data_element
 from . import comment
-from . import ai_assistant
 
 # Keep this at the end of imports to make sure the plugin APIs override the default ones
 try:
@@ -43,5 +42,4 @@
 event_log
 data_element
 comment
-ai_assistant
 api_plugin

querybook/server/datasources_socketio/__init__.py

@@ -1,7 +1,9 @@
 from . import query_execution
 from . import datadoc
 from . import connect
+from . import ai_assistant
 
 connect
 query_execution
 datadoc
+ai_assistant

querybook/server/datasources_socketio/ai_assistant.py

@@ -0,0 +1,13 @@
+from const.ai_assistant import (
+    AI_ASSISTANT_NAMESPACE,
+    AI_ASSISTANT_REQUEST_EVENT,
+)
+
+from .helper import register_socket
+
+
+@register_socket(AI_ASSISTANT_REQUEST_EVENT, namespace=AI_ASSISTANT_NAMESPACE)
+def ai_assistant_request(command_type: str, payload={}):
+    from lib.ai_assistant import ai_assistant
+
+    ai_assistant.handle_ai_command(command_type, payload)

querybook/server/datasources_socketio/connect.py

@@ -1,6 +1,7 @@
 from flask_login import current_user
 from flask_socketio import ConnectionRefusedError
 
+from const.ai_assistant import AI_ASSISTANT_NAMESPACE
 from const.data_doc import DATA_DOC_NAMESPACE
 from const.query_execution import QUERY_EXECUTION_NAMESPACE
 
@@ -20,3 +21,8 @@ def connect_query_execution(auth):
 @register_socket("connect", namespace=DATA_DOC_NAMESPACE)
 def connect_datadoc(auth):
     on_connect()
+
+
+@register_socket("connect", namespace=AI_ASSISTANT_NAMESPACE)
+def connect_ai_assistant(auth):
+    on_connect()

querybook/server/lib/ai_assistant/__init__.py

@@ -1,11 +1,14 @@
 from env import QuerybookSettings
 
+from .all_ai_assistants import get_ai_assistant_class
+
 
 if QuerybookSettings.AI_ASSISTANT_PROVIDER:
-    from .ai_assistant import AIAssistant
+    ai_assistant = get_ai_assistant_class(QuerybookSettings.AI_ASSISTANT_PROVIDER)
+    ai_assistant.set_config(QuerybookSettings.AI_ASSISTANT_CONFIG)
 
-    ai_assistant = AIAssistant(
-        QuerybookSettings.AI_ASSISTANT_PROVIDER, QuerybookSettings.AI_ASSISTANT_CONFIG
-    )
 else:
     ai_assistant = None
+
+
+__all__ = ["ai_assistant"]

querybook/server/lib/ai_assistant/assistants/openai_assistant.py

@@ -3,11 +3,6 @@
 
 from langchain.chat_models import ChatOpenAI
 from langchain.callbacks.manager import CallbackManager
-from langchain.prompts.chat import (
-    ChatPromptTemplate,
-    SystemMessage,
-    HumanMessagePromptTemplate,
-)
 import openai
 
 
@@ -29,174 +24,9 @@ def _get_error_msg(self, error) -> str:
 
         return super()._get_error_msg(error)
 
-    @property
-    def title_generation_prompt_template(self) -> ChatPromptTemplate:
-        system_message_prompt = SystemMessage(
-            content="You are a helpful assistant that can summerize SQL queries."
-        )
-        human_template = (
-            "Generate a brief 10-word-maximum title for the SQL query below. "
-            "===Query\n"
-            "{query}\n\n"
-            "===Response Guidelines\n"
-            "1. Only respond with the title without any explanation\n"
-            "2. Dont use double quotes to enclose the title\n"
-            "3. Dont add a final period to the title\n\n"
-            "===Example response\n"
-            "This is a title\n"
-        )
-        human_message_prompt = HumanMessagePromptTemplate.from_template(human_template)
-        return ChatPromptTemplate.from_messages(
-            [system_message_prompt, human_message_prompt]
-        )
-
-    @property
-    def query_auto_fix_prompt_template(self) -> ChatPromptTemplate:
-        system_message_prompt = SystemMessage(
-            content=(
-                "You are a SQL expert that can help fix SQL query errors.\n\n"
-                "Please follow the format below for your response:\n"
-                "<@key-1@>\n"
-                "value-1\n\n"
-                "<@key-2@>\n"
-                "value-2\n\n"
-            )
-        )
-        human_template = (
-            "Please help fix the query below based on the given error message and table schemas. \n\n"
-            "===SQL dialect\n"
-            "{dialect}\n\n"
-            "===Query\n"
-            "{query}\n\n"
-            "===Error\n"
-            "{error}\n\n"
-            "===Table Schemas\n"
-            "{table_schemas}\n\n"
-            "===Response Format\n"
-            "<@key-1@>\n"
-            "value-1\n\n"
-            "<@key-2@>\n"
-            "value-2\n\n"
-            "===Example response:\n"
-            "<@explanation@>\n"
-            "This is an explanation about the error\n\n"
-            "<@fix_suggestion@>\n"
-            "This is a recommended fix for the error\n\n"
-            "<@fixed_query@>\n"
-            "The fixed SQL query\n\n"
-            "===Response Guidelines\n"
-            "1. For the <@fixed_query@> section, it can only be a valid SQL query without any explanation.\n"
-            "2. If there is insufficient context to address the query error, you may leave the fixed_query section blank and provide a general suggestion instead.\n"
-            "3. Maintain the original query format and case in the fixed_query section, including comments, except when correcting the erroneous part.\n"
-        )
-        human_message_prompt = HumanMessagePromptTemplate.from_template(human_template)
-        return ChatPromptTemplate.from_messages(
-            [system_message_prompt, human_message_prompt]
-        )
-
-    @property
-    def generate_sql_query_prompt_template(self) -> ChatPromptTemplate:
-        system_message_prompt = SystemMessage(
-            content=(
-                "You are a SQL expert that can help generating SQL query.\n\n"
-                "Please follow the key/value pair format below for your response:\n"
-                "<@key-1@>\n"
-                "value-1\n\n"
-                "<@key-2@>\n"
-                "value-2\n\n"
-            )
-        )
-        human_template = (
-            "Please help to generate a new SQL query or modify the original query to answer the following question. Your response should ONLY be based on the given context.\n\n"
-            "===SQL Dialect\n"
-            "{dialect}\n\n"
-            "===Tables\n"
-            "{table_schemas}\n\n"
-            "===Original Query\n"
-            "{original_query}\n\n"
-            "===Question\n"
-            "{question}\n\n"
-            "===Response Format\n"
-            "<@key-1@>\n"
-            "value-1\n\n"
-            "<@key-2@>\n"
-            "value-2\n\n"
-            "===Example Response:\n"
-            "Example 1: Sufficient Context\n"
-            "<@query@>\n"
-            "A generated SQL query based on the provided context with the asked question at the beginning is provided here.\n\n"
-            "Example 2: Insufficient Context\n"
-            "<@explanation@>\n"
-            "An explanation of the missing context is provided here.\n\n"
-            "===Response Guidelines\n"
-            "1. If the provided context is sufficient, please respond only with a valid SQL query without any explanations in the <@query@> section. The query should start with a comment containing the question being asked.\n"
-            "2. If the provided context is insufficient, please explain what information is missing.\n"
-            "3. If the original query is provided, please modify the original query to answer the question. The original query may start with a comment containing a previously asked question. If you find such a comment, please use both the original question and the new question to generate the new query.\n"
-            "4. The <@key_name@> in the response can only be <@explanation@> or <@query@>.\n\n"
-        )
-        human_message_prompt = HumanMessagePromptTemplate.from_template(human_template)
-        return ChatPromptTemplate.from_messages(
-            [system_message_prompt, human_message_prompt]
-        )
-
-    def _generate_title_from_query(
-        self, query, stream=True, callback_handler=None, user_id=None
-    ):
-        """Generate title from SQL query using OpenAI's chat model."""
-        messages = self.title_generation_prompt_template.format_prompt(
-            query=query
-        ).to_messages()
-        chat = ChatOpenAI(
-            **self._config,
-            streaming=stream,
-            callback_manager=CallbackManager([callback_handler]),
-        )
-        ai_message = chat(messages)
-        return ai_message.content
-
-    def _query_auto_fix(
-        self,
-        language,
-        query,
-        error,
-        table_schemas,
-        stream,
-        callback_handler,
-        user_id=None,
-    ):
-        """Query auto fix using OpenAI's chat model."""
-        messages = self.query_auto_fix_prompt_template.format_prompt(
-            dialect=language, query=query, error=error, table_schemas=table_schemas
-        ).to_messages()
-        chat = ChatOpenAI(
-            **self._config,
-            streaming=stream,
-            callback_manager=CallbackManager([callback_handler]),
-        )
-        ai_message = chat(messages)
-        return ai_message.content
-
-    def _generate_sql_query(
-        self,
-        language: str,
-        table_schemas: str,
-        question: str,
-        original_query: str,
-        stream,
-        callback_handler,
-        user_id=None,
-    ):
-        """Generate SQL query using OpenAI's chat model."""
-        messages = self.generate_sql_query_prompt_template.format_prompt(
-            dialect=language,
-            question=question,
-            table_schemas=table_schemas,
-            original_query=original_query,
-        ).to_messages()
-        chat = ChatOpenAI(
+    def _get_llm(self, callback_handler):
+        return ChatOpenAI(
             **self._config,
-            streaming=stream,
+            streaming=True,
             callback_manager=CallbackManager([callback_handler]),
         )
-        ai_message = chat(messages)
-        return ai_message.content