MindsDB 25.2.4.0__py3-none-any.whl → 25.3.1.0__py3-none-any.whl

mindsdb/integrations/handlers/youtube_handler/youtube_handler.py

@@ -10,9 +10,7 @@ from mindsdb.integrations.libs.response import (
 from mindsdb.utilities import log
 from mindsdb_sql_parser import parse_sql
 
-from collections import OrderedDict
 from mindsdb.utilities.config import Config
-from mindsdb.integrations.libs.const import HANDLER_CONNECTION_ARG_TYPE as ARG_TYPE
 
 from googleapiclient.discovery import build
 
@@ -38,12 +36,10 @@ class YoutubeHandler(APIHandler):
             name of a handler instance
         """
         super().__init__(name)
-
-        connection_data = kwargs.get("connection_data", {})
+        self.connection_data = kwargs.get("connection_data", {})
+        self.kwargs = kwargs
 
         self.parser = parse_sql
-        self.connection_data = connection_data
-        self.kwargs = kwargs
         self.connection = None
         self.is_connected = False
 
@@ -104,7 +100,6 @@ class YoutubeHandler(APIHandler):
             Status confirmation
         """
         response = StatusResponse(False)
-        need_to_close = self.is_connected is False
 
         try:
             self.connect()
@@ -131,34 +126,3 @@ class YoutubeHandler(APIHandler):
         """
         ast = parse_sql(query)
         return self.query(ast)
-
-
-connection_args = OrderedDict(
-    youtube_access_token={
-        "type": ARG_TYPE.STR,
-        "description": "API Key",
-        "label": "API Key",
-    },
-    credentials_url={
-        'type': ARG_TYPE.STR,
-        'description': 'URL to OAuth2 Credentials',
-        'label': 'URL to OAuth2 Credentials',
-    },
-    credentials_file={
-        'type': ARG_TYPE.STR,
-        'description': 'Location of OAuth2 Credentials',
-        'label': 'Location of OAuth2 Credentials',
-    },
-    credentials={
-        'type': ARG_TYPE.PATH,
-        'description': 'OAuth2 Credentials',
-        'label': 'Upload OAuth2 Credentials',
-    },
-    code={
-        'type': ARG_TYPE.STR,
-        'description': 'Authentication Code',
-        'label': 'Authentication Code',
-    }
-)
-
-connection_args_example = OrderedDict(youtube_api_token="<your-youtube-api-token>")

mindsdb/integrations/libs/llm/utils.py

@@ -16,6 +16,7 @@ from mindsdb.integrations.libs.llm.config import (
     NvidiaNIMConfig,
     MindsdbConfig,
 )
+from mindsdb.utilities.config import config
 from langchain_text_splitters import Language, RecursiveCharacterTextSplitter
 
 
@@ -211,7 +212,7 @@ def get_llm_config(provider: str, args: Dict) -> BaseLLMConfig:
     if provider == "mindsdb":
         return MindsdbConfig(
             model_name=args["model_name"],
-            project_name=args.get("project_name", "mindsdb"),
+            project_name=args.get("project_name", config.get("default_project")),
         )
     if provider == "vllm":
         return OpenAIConfig(

mindsdb/integrations/utilities/handlers/auth_utilities/google/google_user_oauth_utilities.py

@@ -1,9 +1,8 @@
-import os
 import json
+from pathlib import Path
 import requests
 import datetime as dt
 from flask import request
-from shutil import copyfile
 
 from mindsdb.utilities import log
 
@@ -29,73 +28,65 @@ class GoogleUserOAuth2Manager:
         creds = None
 
         if self.credentials_file or self.credentials_url:
-            # get the current directory and checks tokens & creds
-            curr_dir = self.handler_storage.folder_get('config')
+            oauth_user_info = self.handler_storage.encrypted_json_get('oauth_user_info')
 
-            creds_file = os.path.join(curr_dir, 'creds.json')
-            secret_file = os.path.join(curr_dir, 'secret.json')
-
-            if os.path.isfile(creds_file):
-                creds = Credentials.from_authorized_user_file(creds_file, self.scopes)
+            if oauth_user_info:
+                creds = Credentials.from_authorized_user_info(oauth_user_info, self.scopes)
 
             if not creds or not creds.valid:
                 logger.debug("Credentials do not exist or are invalid, attempting to authorize again")
 
-                if self._download_secret_file(secret_file):
-                    # save to storage
-                    self.handler_storage.folder_sync('config')
-                else:
-                    raise ValueError('No valid Gmail Credentials filepath or S3 url found.')
+                oauth_user_info = self._download_oauth_user_info()
 
                 if creds and creds.expired and creds.refresh_token:
                     creds.refresh(Request())
                     logger.debug("Credentials refreshed successfully")
                 else:
-                    creds = self._execute_google_auth_flow(secret_file, self.scopes, self.code)
+                    creds = self._execute_google_auth_flow(oauth_user_info)
                     logger.debug("New credentials obtained")
 
-                self._save_credentials_to_file(creds, creds_file)
-                logger.debug(f"saved session credentials to {creds_file}")
-                self.handler_storage.folder_sync('config')
+                self.handler_storage.encrypted_json_set('oauth_user_info', self._convert_credentials_to_dict(creds))
+                logger.debug("Saving credentials to storage")
 
         return creds
 
-    def _download_secret_file(self, secret_file):
-        # if credentials_url is set, attempt to download the file
+    def _download_oauth_user_info(self):
+        # if credentials_url is set, attempt to download the contents of the files
         # this will be given preference over credentials_file
         if self.credentials_url:
             response = requests.get(self.credentials_url)
             if response.status_code == 200:
-                with open(secret_file, 'w') as creds:
-                    creds.write(response.text)
-                return True
+                return response.json()
             else:
-                logger.error("Failed to get credentials from S3", response.status_code)
+                logger.error("Failed to get credentials from URL", response.status_code)
+
+        # if credentials_file is set, attempt to read the contents of the file
+        if self.credentials_file:
+            path = Path(self.credentials_file).expanduser()
+            if path.exists():
+                with open(path, 'r') as f:
+                    return json.load(f)
+            else:
+                logger.error("Credentials file does not exist")
 
-        # if credentials_file is set, attempt to copy the file
-        if self.credentials_file and os.path.isfile(self.credentials_file):
-            copyfile(self.credentials_file, secret_file)
-            return True
-        return False
+        raise ValueError('OAuth2 credentials could not be found')
 
-    def _execute_google_auth_flow(self, secret_file, scopes, code=None):
-        flow = Flow.from_client_secrets_file(secret_file, scopes)
+    def _execute_google_auth_flow(self, oauth_user_info: dict):
+        flow = Flow.from_client_config(
+            oauth_user_info,
+            scopes=self.scopes
+        )
 
         flow.redirect_uri = request.headers['ORIGIN'] + '/verify-auth'
 
-        if code:
-            flow.fetch_token(code=code)
+        if self.code:
+            flow.fetch_token(code=self.code)
             creds = flow.credentials
             return creds
         else:
             auth_url = flow.authorization_url()[0]
             raise AuthException(f'Authorisation required. Please follow the url: {auth_url}', auth_url=auth_url)
 
-    def _save_credentials_to_file(self, creds, file_path):
-        with open(file_path, 'w') as token:
-            data = self._convert_credentials_to_dict(creds)
-            token.write(json.dumps(data))
-
     def _convert_credentials_to_dict(self, credentials):
         return {
             'token': credentials.token,

mindsdb/integrations/utilities/pydantic_utils.py

@@ -0,0 +1,208 @@
+import pprint
+
+pydantic_schema_description = """## Understanding Pydantic Schemas for JSON Formatting
+
+Pydantic schemas provide a framework for defining the structure and validation rules for JSON output. Below is an overview of key components commonly found in a Pydantic schema:
+
+### Key Components
+
+Each object in the schema represents a Pydantic model in JSON format. Typical fields in a Pydantic model description include:
+
+- **`anyOf`**:
+  - A list describing possible values for a Pydantic model field.
+
+- **`additionalProperties`**:
+  - Describes the keys of a dictionary. Keys are always of type `string` due to this being a JSON Pydantic schema. The corresponding key types supported by Pydantic are:
+    - `string`: a text string
+    - `integer`: an integer number
+    - `number`: a floating-point number
+
+- **`items`**:
+  - Describes the items contained within an `array` (list).
+
+- **`type`**:
+  - Specifies the Pydantic type assigned to the field, defining the expected data type. Common types include:
+    - `string`: a text string
+    - `integer`: an integer number
+    - `number`: a floating-point number
+    - `array`: a list
+    - `object`: a dictionary
+    - `null`: the python null value None. Indicates the field is optional.
+
+- **`description`**:
+  - Provides a textual narrative explaining the purpose and details of the output JSON field.
+
+- **`title`**:
+  - A Pydantic-generated, human-readable title for the field.
+
+- **`default`**:
+  - The default value for this field if no value is provided by the user.
+
+### Schema
+
+Below is the Pydantic schema:
+
+{schema}
+
+### Examples
+
+Below is an example of well-formed output adhering to this schema.
+
+- Dummy text strings are represented as "lorem ipsum."
+
+{example}
+"""
+
+
+def get_dummy_value(field_value):
+    """A function to return a dummy value of a Pydantic model field."""
+    type_str = field_value["type"]
+    example_dict = {
+        "string": "lorem ipsum",
+        "int": 3,
+        "number": 42.0,
+        "null": None,
+        "object": {"lorem ipsum": "lorem_ipsum"},
+    }
+
+    if type_str in example_dict:
+        return example_dict[type_str]
+    else:
+        return None
+
+
+def get_dummy_array(field_value):
+    """A function to return a dummy array of a Pydantic model field."""
+    items = field_value["items"]
+
+    if "type" in items:
+        if items["type"] == "null":  # skip if null
+            pass
+        elif items["type"] == "array":  # is it an array?
+            array_value = get_dummy_array(items)
+        elif (
+            items["type"] == "object" and "additionalProperties" in items
+        ):  # is it a dict?
+            array_value = get_dummy_dict(items)
+        else:  # it is a regular value!
+            array_value = get_dummy_value(items)
+        return [array_value for _ in range(2)]
+
+    elif "AnyOf" in field_value["items"]:
+        array_value = get_any_of(field_value["items"])  # can be one of many types
+        return [array_value for _ in range(2)]
+
+    else:  # is it a pydantic class?
+        array_value = example_generator(items)
+        return [array_value for _ in range(2)]
+
+
+def get_dummy_dict(field_value):
+    """A function to return a dummy dictionary of a Pydantic model field."""
+    return get_dummy_value(field_value)
+
+
+def get_any_of(field_value):
+    """A function to return the first viable pydantic type of an Any() Pydantic model field."""
+    for any_of in field_value["anyOf"]:
+        if "type" in any_of:
+            if any_of["type"] == "null":  # skip if null
+                continue
+            elif any_of["type"] == "array":  # is it an array?
+                out = get_dummy_array(any_of)
+                return out
+            elif (
+                any_of["type"] == "object" and "additionalProperties" in any_of
+            ):  # is it a dict?
+                out = get_dummy_dict(any_of)
+                return out
+            else:  # it is a regular value!
+                out = get_dummy_value(any_of)
+                return out
+        else:  # is it a pydantic class?
+            out = example_generator(any_of)
+            return out
+
+
+def example_generator(pydantic_json_schema):
+    """dynamically parse a pydantic object and generate an example of it's formatting."""
+
+    example_dict = {}
+    for schema_name, schema in pydantic_json_schema.items():
+
+        for field_name, field_value in schema.items():
+            if "type" in field_value:
+
+                if field_value["type"] == "array":  # is it an array?
+                    example_dict[field_name] = get_dummy_array(field_value)
+
+                elif (
+                    field_value["type"] == "object"
+                    and "additionalProperties" in field_value
+                ):  # is it a dict?
+                    example_dict[field_name] = get_dummy_dict(field_value)
+
+                else:  # it is a regular value!
+                    example_dict[field_name] = get_dummy_value(field_value)
+
+            elif "anyOf" in field_value:
+                example_dict[field_name] = get_any_of(field_value)
+
+            else:  # it is a pydantic class
+                example_dict[field_name] = example_generator(field_value)
+    return example_dict
+
+
+def search_and_replace_refs(schema, defs, ref_skip={}, n=0):
+    """Dynamically substitute subclass references in a Pydantic object schema."""
+    for key, value in schema.items():
+        if key in ref_skip:
+            continue
+        if type(value) is dict:
+            if "$ref" in value:
+                definition_key = value["$ref"].split("/")[-1]
+                if definition_key in ref_skip:
+                    schema[key] = {"type": "null"}
+                else:
+                    schema[key] = {definition_key: defs[definition_key]["properties"]}
+            else:
+                search_and_replace_refs(value, defs, ref_skip, n + 1)
+        elif type(value) is list:
+            for val in value:
+                search_and_replace_refs(val, defs, ref_skip, n + 1)
+
+
+def remove_extraneous_fields(schema, ref_skip):
+    """Remove extraneous fields from object descriptions."""
+    reduced_schema = schema["properties"]
+
+    for ref in ref_skip.keys():
+        if ref in reduced_schema:
+            del reduced_schema[ref]
+
+    for key, value in reduced_schema.items():
+        if "title" in value:
+            del value["title"]
+        if "$defs" in value:
+            del value["$defs"]
+        if "required" in value:
+            del value["required"]
+
+    return reduced_schema
+
+
+def format_for_prompt(pydantic_object, ref_skip={}):
+    """Format a Pydantic object description for prompting an LLM."""
+    schema = {k: v for k, v in pydantic_object.schema().items()}
+
+    search_and_replace_refs(
+        schema=schema["properties"], defs=schema["$defs"], ref_skip=ref_skip, n=0
+    )
+
+    reduced_schema = remove_extraneous_fields(schema, ref_skip)
+
+    reduced_schema = {schema["title"]: reduced_schema}
+
+    out = pprint.pformat(reduced_schema)
+
+    return out, reduced_schema

mindsdb/integrations/utilities/rag/pipelines/rag.py

@@ -294,16 +294,23 @@ class LangChainRAGPipeline:
         retriever = SQLRetriever(
             fallback_retriever=vector_store_retriever,
             vector_store_handler=knowledge_base_table.get_vector_db(),
-            metadata_schemas=retriever_config.metadata_schemas,
-            examples=retriever_config.examples,
+            min_k=retriever_config.min_k,
+            max_filters=retriever_config.max_filters,
+            filter_threshold=retriever_config.filter_threshold,
+            database_schema=retriever_config.database_schema,
             embeddings_model=embeddings,
+            search_kwargs=config.search_kwargs,
             rewrite_prompt_template=retriever_config.rewrite_prompt_template,
-            metadata_filters_prompt_template=retriever_config.metadata_filters_prompt_template,
+            table_prompt_template=retriever_config.table_prompt_template,
+            column_prompt_template=retriever_config.column_prompt_template,
+            value_prompt_template=retriever_config.value_prompt_template,
+            boolean_system_prompt=retriever_config.boolean_system_prompt,
+            generative_system_prompt=retriever_config.generative_system_prompt,
             num_retries=retriever_config.num_retries,
             embeddings_table=knowledge_base_table._kb.vector_database_table,
             source_table=retriever_config.source_table,
+            source_id_column=retriever_config.source_id_column,
             distance_function=distance_function,
-            search_kwargs=config.search_kwargs,
             llm=sql_llm
         )
         return cls(