fix memory response type

agixt/graphqlendpoints/Extensions.py

@@ -1,14 +1,8 @@
 from typing import List, Dict, Any, Optional
 import strawberry
-from fastapi import Depends, HTTPException, Header
-from Models import (
-    CommandExecution,
-    CommandArgs,
-    ExtensionsModel,
-    ExtensionSettings,
-)
-from Extensions import Extensions
-from ApiClient import Agent, Conversations, verify_api_key, get_api_client, is_admin
+from fastapi import HTTPException
+from Models import CommandExecution
+from ApiClient import verify_api_key, is_admin
 from endpoints.Extension import (
     get_extension_settings as rest_get_extension_settings,
     get_command_args as rest_get_command_args,
@@ -18,20 +12,56 @@
 )
 
 
-# Convert Pydantic models to Strawberry types
-@strawberry.experimental.pydantic.type(model=ExtensionSettings)
-class ExtensionSettingsType:
-    extension_settings: Dict[str, Dict[str, Any]]
-
-
-@strawberry.experimental.pydantic.type(model=CommandArgs)
-class CommandArgsType:
-    command_args: Dict[str, Any]
-
-
-@strawberry.experimental.pydantic.type(model=ExtensionsModel)
-class ExtensionsModelType:
-    extensions: List[Dict[str, Any]]
+""" get_agent_extensions output:
+{"extensions": [{"extension_settings": {'sendgrid_email': {'SENDGRID_API_KEY': '', 'SENDGRID_EMAIL': ''},
+ 'postgres_database': {'POSTGRES_DATABASE_NAME': '',
+  'POSTGRES_DATABASE_HOST': '',
+  'POSTGRES_DATABASE_PORT': 5432,
+  'POSTGRES_DATABASE_USERNAME': '',
+  'POSTGRES_DATABASE_PASSWORD': ''},
+ 'google_search': {'GOOGLE_API_KEY': '', 'GOOGLE_SEARCH_ENGINE_ID': ''},
+ 'mysql_database': {'MYSQL_DATABASE_NAME': '',
+  'MYSQL_DATABASE_HOST': '',
+  'MYSQL_DATABASE_PORT': 3306,
+  'MYSQL_DATABASE_USERNAME': '',
+  'MYSQL_DATABASE_PASSWORD': ''},
+ 'oura': {'OURA_API_KEY': ''},
+ 'discord': {'DISCORD_API_KEY': '', 'DISCORD_COMMAND_PREFIX': '/AGiXT'},
+ 'github': {'GITHUB_USERNAME': '', 'GITHUB_API_KEY': ''},
+ 'AGiXT Chains': {'Think twice': {'user_input': ''}}]}
+"""
+
+""" get_extensions output:
+{"extensions":
+[{'extension_name': 'Long Term Memory',
+  'description': "The Long Term Memory extension enables AGiXT to create and manage persistent memory databases.\nIt provides commands for:\n- Creating specialized memory databases for different types of information\n- Organizing and storing memories in structured tables\n- Retrieving specific memories through SQL queries\n- Tracking metadata about stored knowledge\n- Managing the evolution of memory organization over time\n\nThis extension allows agents to maintain their own organized, searchable knowledge bases\nthat persist across conversations. This acts as the assistant's very long-term memory.",
+  'settings': [],
+  'commands': [{'friendly_name': 'Create Memory Database',
+    'description': 'Create a new memory database for storing and organizing information. This command should be used whenever the agent wants to:\n- Create a new category of memories or knowledge\n- Start tracking a new type of information\n- Organize related data in a structured way\n\nExamples of when to use this command:\n- Creating a database for learning progress in a specific subject\n- Starting a database for tracking project-related information\n- Creating a database for storing research findings\n- Making a database for conversation summaries\n- Creating specialized databases for different types of technical knowledge\n\nArgs:\nname (str): Name of the database (e.g., "russian_learning", "project_memories", "technical_docs")\ndescription (str): Detailed description of what this database stores and its purpose\n\nReturns:\nstr: Success message confirming database creation\n\nExample Usage:\n<execute>\n<name>Create Memory Database</name>\n<name>russian_vocabulary</name>\n<description>Database for storing Russian vocabulary words, phrases, and usage examples learned during conversations, including difficulty levels and practice timestamps.</description>\n</execute>',
+    'command_name': 'create_memory_database',
+    'command_args': {'name': '', 'description': ''}},
+   {'friendly_name': 'Remember This',
+    'description': "Store new information in the assistant's long-term memory. This command should be used when:\n- Learning new information that should be remembered later\n- Saving important facts, concepts, or insights\n- Recording structured information for future reference\n- Creating persistent knowledge that should be available across conversations\n- Building up knowledge bases for specific topics\n\nThe assistant will:\n- Analyze what type of information is being stored\n- Choose or create an appropriate memory database\n- Design or use existing table structures\n- Store the information with relevant metadata\n- Verify successful storage\n\nArgs:\ncontent (str): The information to remember (e.g., facts, concepts, structured data)\nmemory_type (str, optional): Category or type of memory to help with organization\n\nExample Usage:\n<execute>\n<name>Store in Long Term Memory</name>\n<content>The word 'полка' means 'shelf' in Russian. Common usage is 'Книга на полке' meaning 'The book is on the shelf'. This is a frequently used noun in household contexts.</content>\n<memory_type>russian_vocabulary</memory_type>\n</execute>\n\nReturns:\nstr: Confirmation of what was stored and where it can be found",
+    'command_name': 'store_memory',
+    'command_args': {'content': '', 'memory_type': ''}},
+   {'friendly_name': 'List Memory Databases',
+    'description': 'List all available memory databases and their descriptions. This command should be used when:\n- Deciding which database to store new information in\n- Looking for existing knowledge on a topic\n- Planning where to organize new information\n- Reviewing available knowledge categories\n- Checking when databases were last updated\n\nThe command returns a CSV formatted list containing:\n- Database names\n- Their descriptions\n- Creation dates\n- Last modification dates\n\nThis is particularly useful for:\n- Finding the right database for storing new information\n- Discovering existing knowledge bases\n- Maintaining organization of memories\n- Tracking knowledge evolution over time\n\nReturns:\nstr: CSV formatted list of all memory databases with their metadata\n\nExample Output:\n```csv\n"Database Name","Description","Created Date","Last Modified"\n"russian_vocabulary","Storage for Russian language learning progress","2024-01-01 12:00:00","2024-01-02 15:30:00"\n"project_notes","Technical documentation and decision history for current project","2024-01-01 09:00:00","2024-01-02 14:45:00"\n```',
+    'command_name': 'list_memory_databases',
+    'command_args': {}}]
+}
+"""
+
+""" get_commands output:
+{"commands":
+{'Create Memory Database': False,
+ 'Remember This': False,
+ 'List Memory Databases': False,
+ 'Update Memory Database Description': False,
+ 'Retrieve Memories': False,
+ 'Send Email with Sendgrid': False,
+ }
+ }
+"""
 
 
 # Input types

agixt/graphqlendpoints/Memories.py

@@ -1,20 +1,16 @@
 from typing import List, Dict, Any, Optional
 import strawberry
-from fastapi import Depends, HTTPException, Header
+from fastapi import HTTPException
 from Models import (
     AgentMemoryQuery,
     TextMemoryInput,
     FileInput,
     UrlInput,
-    ResponseMessage,
     Dataset,
     FinetuneAgentModel,
     ExternalSource,
     UserInput,
     FeedbackInput,
-    MemoryResponse,
-    MemoryCollectionResponse,
-    DPOResponse,
 )
 from endpoints.Memory import (
     query_memories as rest_query_memories,
@@ -35,23 +31,35 @@
 from ApiClient import verify_api_key
 
 
+@strawberry.type
+class Memory:
+    external_source_name: str
+    id: str
+    description: str
+    text: str
+    embedding: str
+    additional_metadata: str
+    key: str
+    timestamp: str
+
+
 # Convert Pydantic models to Strawberry types
-@strawberry.experimental.pydantic.type(model=MemoryResponse)
+@strawberry.type
 class MemoryResponseType:
-    memories: List[Dict[str, Any]]
+    memories: List[Memory]
 
 
-@strawberry.experimental.pydantic.type(model=MemoryCollectionResponse)
+@strawberry.type
 class MemoryCollectionResponseType:
     external_sources: List[str]
 
 
-@strawberry.experimental.pydantic.type(model=ResponseMessage)
+@strawberry.type
 class ResponseMessageType:
     message: str
 
 
-@strawberry.experimental.pydantic.type(model=DPOResponse)
+@strawberry.type
 class DPOResponseType:
     prompt: str
     chosen: str

agixt/graphqlendpoints/Schema.py

@@ -4,9 +4,10 @@
 # from graphqlendpoints.Auth import schema as auth_schema
 # from graphqlendpoints.Chains import schema as chains_schema
 # from graphqlendpoints.Completions import schema as completions_schema
-# from graphqlendpoints.Memories import schema as memories_schema
+# from graphqlendpoints.Extensions import schema as extensions_schema
+
+from graphqlendpoints.Memories import schema as memories_schema
 from graphqlendpoints.Conversations import schema as conversations_schema
-from graphqlendpoints.Extensions import schema as extensions_schema
 from graphqlendpoints.Prompts import schema as prompts_schema
 from graphqlendpoints.Providers import schema as providers_schema
 
@@ -17,9 +18,9 @@ class Query(
     # auth_schema.Query,
     # chains_schema.Query,
     # completions_schema.Query,
-    # memories_schema.Query,
+    memories_schema.Query,
     conversations_schema.Query,
-    extensions_schema.Query,
+    # extensions_schema.Query,
     prompts_schema.Query,
     providers_schema.Query,
 ):
@@ -32,9 +33,9 @@ class Mutation(
     # auth_schema.Mutation,
     # chains_schema.Mutation,
     # completions_schema.Mutation,
-    # memories_schema.Mutation,
+    memories_schema.Mutation,
     conversations_schema.Mutation,
-    extensions_schema.Mutation,
+    # extensions_schema.Mutation,
     prompts_schema.Mutation,
     providers_schema.Mutation,
 ):