Merge branch 'sirocco-ventures:main' into main

app/plugins/bigquery/__init__.py

@@ -25,7 +25,7 @@
     service_account_json=ConnectionArgument(
         type= 7,
         generic_name= 'Service account JSON',
-        description= 'Service account fetails',
+        description= 'Service account details',
         order= 2,
         required = True,
         value = None,

app/plugins/loader.py

@@ -1,4 +1,5 @@
 from app.plugins.postgresql.handler import Postresql
+from app.plugins.mysql.handler import Mysql
 from app.plugins.bigquery.handler import Bigquery
 from app.plugins.airtable.handler import Airtable
 from app.plugins.website.handler import Website
@@ -12,6 +13,7 @@ def __init__(self, configs):
     def load_ds(self):
         db_classes = {
             "postgres": Postresql,
+            "mysql": Mysql,
             "bigquery": Bigquery,
             "airtable": Airtable,
             "website": Website,

app/plugins/mysql/__init__.py

@@ -0,0 +1,186 @@
+from app.models.prompt import Prompt
+from collections import OrderedDict
+from app.models.request import ConnectionArgument
+
+# Plugin Metadata
+__version__ = '1.0.0'
+__plugin_name__ = 'mysql'
+__display_name__ = "MySQL DB"
+__description__ = 'MySQL integration for handling MySQL database operations.'
+__icon__ = '/assets/plugins/logos/mysql.svg'
+__category__ = 2
+
+
+# Connection arguments
+__connection_args__ = OrderedDict(
+    db_name= ConnectionArgument(
+        type = 1,
+        generic_name= 'MySQL Database name',
+        description = 'Database name',
+        order= 5,
+        required = True,
+        value = None,
+        slug = "db_name"
+    ),
+    db_user=ConnectionArgument(
+        type= 1,
+        generic_name= 'MySQL User name',
+        description= 'Database username',
+        order= 2,
+        required = True,
+        value = None,
+        slug = "db_user"
+    ),
+    db_password=ConnectionArgument(
+        type= 2,
+        generic_name= 'MySQL Password',
+        description= 'Database password',
+        order= 3,
+        required = True,
+        value = None,
+        slug = "db_password"
+    ),
+    db_host=ConnectionArgument(
+        type= 1,
+        generic_name= 'MySQL Database host',
+        description= 'Database hostname',
+        order= 1,
+        required = True,
+        value = None,
+        slug = "db_host"
+    ),
+    db_port=ConnectionArgument(
+        type= 3,
+        generic_name= 'MySQL Database port',
+        description= 'Database port',
+        order = 4,
+        required = True,
+        value = None,
+        slug = "db_port"
+    ),
+)
+
+# Prompt
+__prompt__ = Prompt(**{
+        "base_prompt": "{system_prompt}{user_prompt}",
+        "system_prompt": {
+            "template": """
+            You are an Mysql expert. Your job is to answer questions about a Mysql database using only the provided schema details and rules.
+
+            go through the schema details given below
+            -- start db schema section--
+            {schema}
+            -- end db schema section--
+
+            A brief description about the schema is given below
+
+            -- start db context section--
+            {context}
+            -- end db context section--
+
+            Sample sql queries with their questions are given below
+
+            -- start query samples section--
+            $suggestions
+            -- end query samples section--
+
+            Adhere to the given rules without failure
+
+            -- start rules section --
+            - Use Table Aliases always to prevent ambiguity . For example, `SELECT table1.col1, table2.col1 FROM table1 JOIN table2 ON table1.id = table2.id`.
+            - use LIKE operator with LOWER function for string comparison or equality
+            - Always use alias/table name for fields in WHERE condition
+            - Do not use non existing tables or fields
+            - id columns are mandatory for all operations
+            - Do not use JSON_BUILD_OBJECT operation
+            - Do not use unwanted joins
+            - Do not return incomplete queries
+            - Adher to sysql query syntax
+            -- end rules section --
+            """
+        },
+        "user_prompt":{
+            "template": """
+            Follow these steps to generate query to solve the question `$question`
+
+            1. Deliberately go through schema, context, rules deliberately
+            2. Understand the question and check whether it's doable with the given context
+            3. Do only the task asked, Don't hallucinate and overdo the task
+            4. Strictly return all the fields in the schema during listing operations
+            5. Strictly return at least 1 text fields and an id field during aggregation/group by operations
+            6. Generate a query to solve the problem using the schema, context, and strictly follow the rules
+            7. output in the given json format, extra explanation is strictly prohibited
+
+            {
+                "explanation": "Explain how you finalized the sql query using the schemas and rules provided",
+                "query" : "mysql query",
+                "operation_kind" : "aggregation|list",
+                "schema": "used schema details separated by comma",
+                "confidence" : "confidence in 100",
+                "visualisation": {
+                    "type": "chart type (bar chart, line chart, pie chart) or 'table' for tabular format; 'none' if operation_kind is 'list'",
+                    "x-axis": ["fields that can be used as x axis"],
+                    "y-axis": ["fields that can be used as y axis"],
+                    "title": "layout title name"
+                },
+                "general_message": "a general message describing the answers like 'here is your list of incidents' or 'look what i found'",
+                "main_entity" : "main entity  for the query",
+                "next_questions" : [Produce 3 related questions(maximum 8 words) aligned with the current question, db context and which can be answered with only two table . While creating questions strictly prohibit questions which tells to specify for a specific item]
+
+            }
+            """
+        },
+        "regeneration_prompt": {
+            "template": """
+            You were trying to answer the following user question by writing SQL query to answer the question given in `[question][/question]`
+            [question]
+            $question
+            [/question]
+
+            You generated this query given in `[query][/query]`
+            [query]
+            {query_generated}
+            [/query]
+
+            But upon execution you encountered some error , error traceback is given in [query_error][/query_error]
+            [query_error]
+            {exception_log}
+            [/query_error]
+
+            Follow these steps to generate the query
+
+            1. Deliberately go through schema, context, rules deliberately
+            2. Understand the question and check whether it's doable with the given context
+            3. Use survey answers if available and include it in query for filtering values
+            4. Do only the task asked, Don't hallucinate and overdo the task
+            5. Strictly return all the fields in the schema during listing operations
+            6. Strictly return at least 1 text fields and an id field during aggregation/group by operations
+            7. Generate a query to solve the problem using the schema, context and the rules and based on the previous query try to rectify the query error
+            8. output in the given json format, extra explanation is strictly prohibited
+
+            {
+                "explanation": "Explain how you finalized the sql query using the schemas and rules provided",
+                "query" : "mysql query",
+                "operation_kind" : "aggregation|list",
+                "visualisation": {
+                    "type": "chart type (bar chart, line chart, pie chart) or 'table' for tabular format; 'none' if operation_kind is 'list'",
+                    "value_field": "fields in which values are stored",
+                    "x-axis": "field that can be used as x axis",
+                    "y-axis": "field that can be used as y axis",
+                    "title": "layout title name"
+                },
+                "schema": "used schema details separated by comma",
+                "confidence" : "confidence in 100",
+                "general_message": "a general message describing the answers like 'here is your list of incidents' or 'look what i found'",
+                "main_entity" : "main entity  for the query",
+                "next_questions" : [Produce 3 related questions(maximum 8 words) aligned with the current question, db context and which can be answered with only two table . While creating questions strictly prohibit questions which tells to specify for a specific item]
+            }
+            """
+        }
+    })
+
+
+
+__all__ = [
+    __version__, __plugin_name__, __display_name__ , __description__, __icon__, __category__, __prompt__
+]

app/plugins/mysql/formatter.py

@@ -0,0 +1,100 @@
+from typing import Any
+from loguru import logger
+
+
+class Formatter:
+    def format(self, data: Any,input) -> (dict):
+        """
+        Main entry point for formatting the data based on the input parameters.
+        Handles different formatting strategies based on operation kind.
+
+        :param data: The data to format.
+        :param input_params: Dictionary containing operation and formatting details.
+        :return: A dictionary containing the formatted response.
+        """
+        response = {}
+        self.main_entity = input.get("main_entity")
+        self.kind = input.get("operation_kind", "").lower()
+        self.general_message = input.get("general_message")
+        self.empty_message = input.get("empty_message")
+
+        logger.info("Formatting output using inference for mysql")
+
+        if self.kind  == "list":
+            response = self.basic_formatter(data, input)
+        elif self.kind == "aggregation":
+            response = self.aggregation_formatter(data, input)
+        else:
+            response["data"] = data
+            response["kind"] = "list"
+
+
+        response.update({
+            "main_entity": self.main_entity,
+            "main_format": self.kind,
+            "role": "assistant",
+            "content": self.general_message,
+            "empty_message": self.empty_message,
+        })
+
+        return response
+
+    def basic_formatter(self, data: Any, input:Any) -> dict :
+        """
+        Formats data as a list, handling cases for none, single, and multiple entries.
+
+        :param data: The data to format.
+        :return: A dictionary containing the formatted list response.
+        """
+        logger.info("Formatting data as a list")
+
+        if data is None:
+            response = {"data": [], "kind": "none"}
+        elif len(data) == 1:
+            response = {"data": data, "kind": "single"}
+        else:
+            response = {"data": data, "kind": "list"}
+
+        return response
+
+
+    def aggregation_formatter(self, data:Any, input:Any) -> dict :
+        """
+        Formats data for aggregation visualisation, supporting table and chart formats.
+
+        :param data: The data to format.
+        :param visualisation: Dictionary containing visualisation details (e.g., x-axis, y-axis, chart type).
+        :return: A dictionary containing the formatted aggregation response.
+        """
+
+        logger.info("Formatting data as aggregation")
+
+        visualisation = input.get("visualisation", {})
+        response = {}
+
+
+        if data is None or len(data) == 0:
+            response = {"data": [], "kind": "none"}
+        elif len(data) == 1:
+            response = {"data": data, "kind": "table"}
+        else:
+            value_fields = visualisation.get("y-axis", [])
+            key_fields = visualisation.get("x-axis", [])
+            title = visualisation.get("title", "")
+
+            visualisaton_kind = visualisation["type"].replace(" ", "_") if visualisation["type"] is not None else "table"
+
+            if visualisaton_kind.lower() in ["bar_chart", "line_chart", "pie_chart"] and len(value_fields) > 0 and len(key_fields) > 0:
+                response["kind"] = visualisaton_kind
+                response["data"] = data
+                response["x"] = key_fields
+                response["y"] = value_fields
+                response["title"] = title
+            else:
+                response = {"kind": "table", "data": data}
+
+        return response
+
+
+
+