(20241212) Reorganizing code to perform core actions in one place.

This commit is contained in:
2024-12-12 18:34:51 +05:30
parent 70c817b3b8
commit fb95e0b38c
50 changed files with 1259 additions and 3166 deletions
View File
View File
+182
View File
@@ -0,0 +1,182 @@
"""
AUTHOR:
Khushal P Soonderji
DATE:
Thursday, 5th Dec., 2024
OBJECTIVE:
To create an interface between OpenAI and our internal system to perform LLM-based activities.
REFERENCES:
N/A
DOWNLOADS:
N/A
"""
# *****************************************************************************************************************
# ***** ****
# *** IMPORT ***
# ***** ****
# *****************************************************************************************************************
# To make sibling directories accessible for imports:
import sys
sys.path.append(".")
sys.path.append("..")
# My async utils:
from utils_v2.database.async_mongo_v2 import AsyncMongo
# Base model:
from controllers.base import BaseModel
# Data Models:
from models.core.user import CoreUserInfoModel
from models.core.ai.llm import LLMInput, LLMOutput, LLMUsageTokens
# To work with LLMs:
from langchain_openai import ChatOpenAI
# *****************************************************************************************************************
# ***** ****
# *** MACROS / ONE-TIME INIT ***
# ***** ****
# *****************************************************************************************************************
# --- Nothing Yet
# *****************************************************************************************************************
# ***** ****
# *** VARIABLES ***
# ***** ****
# *****************************************************************************************************************
# --- Nothing Yet
# *****************************************************************************************************************
# ***** ****
# *** FUNCTIONS ***
# ***** ****
# *****************************************************************************************************************
# --- Nothing Yet
# *****************************************************************************************************************
# ***** ****
# *** CLASSES ***
# ***** ****
# *****************************************************************************************************************
class LLMController(BaseModel):
AI_USAGE_COLLECTION = "_aiUsage"
def __init__(
self,
llm_creds: dict,
cache = None,
alert_url = None,
http_client = None,
debug = True,
debug_prefix = "Model | ",
debug_only_errors = True
):
"""
This is the model that works with OpenAi's LLM to perform tasks like text completion.
:param llm_creds: The JSON that holds the credentials to access your OpenAI account. Should have the keys
'model', and 'openai_api_key'.
:param cache: The object to use for caching results from database calls.
:param alert_url: Which URL to call when something goes wrong.
:param http_client: The instance of an HTTP client to use when trying to send alerts and make other APIs.
:param debug: Whether, or not, you would like to print debugging messages:
:param debug_prefix: The prefix to print with the debugging messages.
:param debug_only_errors: Whether you would like to print only error messages or all messages.
:return: None.
"""
# Initialize the parent:
super().__init__(
cache = cache,
alert_url = alert_url,
http_client = http_client,
debug = debug,
debug_prefix = debug_prefix,
debug_only_errors = debug_only_errors
)
# Create the interface to the LLM:
self.__llm = ChatOpenAI(**llm_creds)
async def invoke(
self,
mongo_conn: AsyncMongo,
user_info: CoreUserInfoModel,
llm_input: LLMInput
) -> LLMOutput:
# Format the message as per the format of OpenAI:
prompt = [
{
"role": {"system": "system", "ai": "assistant", "human": "user"}[message.role],
"content": message.content
} for message in llm_input.messages
]
# Invoke the AI, and format the response:
llm_response = await self.__llm.ainvoke(prompt)
print("LLM RESPONSE:", llm_response)
print("INPUT MESSAGES:", llm_input.messages)
llm_response = LLMOutput(
messages = llm_input.messages,
output = llm_response.content,
client = "openai",
model = llm_response.response_metadata["model_name"],
tokens = LLMUsageTokens(
input = llm_response.usage_metadata["input_tokens"],
output = llm_response.usage_metadata["output_tokens"],
total = llm_response.usage_metadata["total_tokens"],
)
)
# Store this into MongoDB:
mongo_document = {"user": user_info.model_dump()}
for k, v in llm_response.model_dump().items(): mongo_document[k] = v
inserted_id = await mongo_conn.insert_one(
collection = self.AI_USAGE_COLLECTION,
document = mongo_document
)
if inserted_id: llm_response.invocationId = str(inserted_id)
# Done here:
return llm_response
# *****************************************************************************************************************
# ***** ****
# *** MAIN PROGRAM ***
# ***** ****
# *****************************************************************************************************************
if __name__ == "__main__":
pass
+314
View File
@@ -0,0 +1,314 @@
"""
AUTHOR:
Khushal P Soonderji
DATE:
Thursday, 12th Dec., 2024
OBJECTIVE:
To handle all auth-tokens from one place.
REFERENCES:
N/A
DOWNLOADS:
N/A
"""
# *****************************************************************************************************************
# ***** ****
# *** IMPORT ***
# ***** ****
# *****************************************************************************************************************
# To make sibling directories accessible for imports:
import sys
sys.path.append(".")
sys.path.append("..")
# For Quart:
from quart import current_app
# My async utils:
from utils_v2.string import json
from utils_v2.date_time import date_time
from utils_v2.database.async_mysql_v2 import AsyncMySQL
from utils_v2.database.async_mongo_v2 import AsyncMongo
# Base model:
from controllers.base import BaseModel
# Data models:
from models.core.auth_token import CoreAuthTokenModel
# To work with MongoDB:
from bson import ObjectId
# *****************************************************************************************************************
# ***** ****
# *** MACROS / ONE-TIME INIT ***
# ***** ****
# *****************************************************************************************************************
# --- Nothing Yet
# *****************************************************************************************************************
# ***** ****
# *** VARIABLES ***
# ***** ****
# *****************************************************************************************************************
# --- Nothing Yet
# *****************************************************************************************************************
# ***** ****
# *** FUNCTIONS ***
# ***** ****
# *****************************************************************************************************************
# --- Nothing Yet
# *****************************************************************************************************************
# ***** ****
# *** CLASSES ***
# ***** ****
# *****************************************************************************************************************
class AuthTokenController(BaseModel):
# ┏┓┓ ┓┏
# ┃ ┃┏┓┏┏ ┃┃┏┓┏┓┏
# ┗┛┗┗┻┛┛ ┗┛┗┻┛ ┛
# For MongoDB:
AUTH_COLLECTION = "_authTokens"
async def get_token_id(
self,
db_conn: AsyncMySQL,
mongo_conn: AsyncMongo,
auth_token: CoreAuthTokenModel,
token_notes: dict,
session_token: str = None,
) -> ObjectId:
"""
Stores params from the session info and gives an identifier to use in the authorization URL. Use this when the
user requests an authorization URL to link your service to another service (like GMail).
:param db_conn: The database connection (MariaDB) to use to perform the action.
:param mongo_conn: The database connection (MongoDB) to use to perform the action.
:param auth_token: An instance of the core auth-token model that holds data in the database.
:param token_notes: Any notes to feed into MariaDB with the token identifier.
:param session_token: The session token of the user who requested this service.
:return: An ObjectId to later store the granted tokens.
"""
# Note down the timestamp at which this event occurred:
request_ts = date_time.get_current_utc_date_time(as_string = False)
# Get the identifier from the database if it already exists, else create one.
# BE CAREFUL WITH THE KEYS HERE, THEY SHOULD MATCH THE FIELDS OF THE CORE AUTH-TOKEN MODEL:
mongo_json = await mongo_conn.find_one_and_update(
collection = self.AUTH_COLLECTION,
filter = mongo_conn.dict_to_dot_notation({
"serviceType": auth_token.serviceType,
"user": {
"entityId": auth_token.user.entityId,
"billingAccountId": auth_token.user.billingAccountId
},
"clientUserId": auth_token.clientUserId
}),
update = {
"$set": {
"lastRequestTs": auth_token.lastRequestTs,
"status": auth_token.status,
"syncFreq": auth_token.syncFreq
},
"$setOnInsert": {
"serviceType": auth_token.serviceType,
"client": auth_token.client,
"authType": auth_token.authType,
"user": auth_token.user.model_dump(),
"clientUserId": auth_token.clientUserId,
"auth": auth_token.auth,
"token": auth_token.token,
"firstRefreshTs": auth_token.firstRefreshTs,
"lastRefreshTs": auth_token.lastRefreshTs,
"firstRequestTs": auth_token.firstRequestTs or request_ts,
}
},
projection = {
"_id": True
},
upsert = True,
return_updated = True
)
# Tell MariaDB that an authorization request was initiated:
db_json = {}
if mongo_json is not None:
db_json = await self.call_procedure(
db_conn = db_conn,
proc_name = "entity_integration_save",
proc_args = (
auth_token.user.entityId, # ..................................... 'p_entity_id'
auth_token.client, # ............................................ 'p_provider'
auth_token.status, # ............................................ 'p_current_status'
"Auth Requested", # ............................................. 'p_last_action'
None, # ......................................................... 'p_display_name'
None, # ......................................................... 'p_display_picture'
str(mongo_json["_id"]), # ....................................... 'p_token_id'
json.to_string(python_data = token_notes, no_space = True), # ... 'p_notes'
auth_token.user.userId # ........................................ 'p_created_by'
),
session_token = session_token
)
# Done here:
return mongo_json["_id"] if mongo_json and db_json.get("status") == 1 else None
async def set_token(
self,
db_conn: AsyncMySQL,
mongo_conn: AsyncMongo,
token_id: ObjectId | str,
auth_token: CoreAuthTokenModel,
token_notes: dict,
session_token: str = None
) -> bool:
"""
This method is to be called when the end user authorizes your service to connect to his third-party account. For
example, when the end user allows you to access his GMail account. USE THIS FOR UPDATING (REFRESHING) TOKENS
ALSO.
:param db_conn: The database connection (MariaDB) to use to perform the action.
:param mongo_conn: The database connection (MongoDB) to use to perform the action.
:param token_id: The identifier granted by the 'get_token_id' method.
:param auth_token: The actual auth/token data to be saved to the database.
:param token_notes: Any notes to feed into MariaDB with the token identifier.
:param session_token: The session token of the user who requested this service.
:return: True if saved, False if failed.
"""
# Start by assuming failure:
token_saved = False
# Note down the timestamp at which this event occurred:
request_ts = date_time.get_current_utc_date_time(as_string = False)
# Save the token to MongoDB.
# BE CAREFUL WITH THE KEYS HERE, THEY SHOULD MATCH THE FIELDS OF THE CORE AUTH-TOKEN MODEL:
mongo_json = await mongo_conn.find_one_and_update(
collection = self.AUTH_COLLECTION,
filter = mongo_conn.dict_to_dot_notation({
"_id": ObjectId(token_id),
"clientUserId": auth_token.clientUserId
}),
update = [{
"$set": {
"auth": auth_token.auth,
"token": auth_token.token,
"status": auth_token.status,
"lastRefreshTs": request_ts,
"firstRefreshTs": {
"$cond": {
"if": {
"$or": [
{"$eq": ["$firstRefreshTs", None]},
{"$eq": [{"$type": "$firstRefreshTs"}, "missing"]}
]
},
"then": request_ts,
"else": "$firstRefreshTs"
}
}
}
}],
projection = {"token": False},
return_updated = True,
upsert = False
)
# Tell MariaDB that the token was saved:
if mongo_json is not None:
db_json = await self.call_procedure(
db_conn = db_conn,
proc_name = "entity_integration_save",
proc_args = (
mongo_json["user"]["entityId"], # ............................... 'p_entity_id'
mongo_json["client"], # ......................................... 'p_provider'
auth_token.status, # ............................................ 'p_current_status'
"Auth Granted", # ............................................... 'p_last_action'
auth_token.token.get("displayName"), # .......................... 'p_display_name'
auth_token.token.get("displayPictureUrl"), # .................... 'p_display_picture'
token_id, # ..................................................... 'p_token_id'
json.to_string(python_data = token_notes, no_space = True), # ... 'p_notes'
auth_token.user.userId # ........................................ 'p_created_by'
),
session_token = session_token
)
if db_json["status"] == 1: token_saved = True
# Done here:
return token_saved
async def get_token(
self,
mongo_conn: AsyncMongo,
token_id: ObjectId | str = None,
**kwargs
) -> CoreAuthTokenModel | None:
"""
To retrieve stored tokens from the database.
:param mongo_conn: The database connection (MongoDB) to use to perform the action.
:param token_id: The identifier granted by the 'get_token_id' method.
:param kwargs: Any set of key-value pairs to build custom search criteria. This could be things like the user
info, the client, the type of authentication used, or even the kind of service.
:return: The retrieved record that has the token, and information about the service and client if found, else
None when there is no matching record.
"""
# Build the filter:
filter_json = {k: v for k, v in kwargs.items()}
if token_id: filter_json["_id"] = ObjectId(token_id)
# If there is no search criteria, we exit with failure:
if not filter_json: return None
# If there is some filtering possible, we fetch the token:
token = await mongo_conn.find_one(
collection = self.AUTH_COLLECTION,
filter = filter_json,
)
# Done here:
return CoreAuthTokenModel(**token) if token else None
# *****************************************************************************************************************
# ***** ****
# *** MAIN PROGRAM ***
# ***** ****
# *****************************************************************************************************************
if __name__ == "__main__":
pass
+389
View File
@@ -0,0 +1,389 @@
"""
AUTHOR:
Khushal P Soonderji
DATE:
Thursday, 12th Dec., 2024
OBJECTIVE:
To handle all messages from one place.
REFERENCES:
N/A
DOWNLOADS:
N/A
"""
# *****************************************************************************************************************
# ***** ****
# *** IMPORT ***
# ***** ****
# *****************************************************************************************************************
# To make sibling directories accessible for imports:
import sys
from pyexpat.errors import messages
sys.path.append(".")
sys.path.append("..")
# For Quart:
from quart import current_app
# My async utils:
from utils_v2.string import json
from utils_v2.date_time import date_time
from utils_v2.database.async_mysql_v2 import AsyncMySQL
from utils_v2.database.async_mongo_v2 import AsyncMongo, AsyncMongoStorage
# Base model:
from controllers.base import BaseModel
# Data models:
from models.core.auth_token import CoreAuthTokenModel
from models.core.message import CoreMessageModel
from models.core.user import CoreUserInfoModel
# To work with MongoDB:
from bson import ObjectId
from pymongo import InsertOne, UpdateOne, ReplaceOne
# To work with datatypes:
from typing import Literal, List, Dict, Any
# To make deep-copies:
import copy
# To work with base-64 encoding:
import base64
# To work with date and time:
import datetime
# For asynchronous activities:
import asyncio
# *****************************************************************************************************************
# ***** ****
# *** MACROS / ONE-TIME INIT ***
# ***** ****
# *****************************************************************************************************************
# --- Nothing Yet
# *****************************************************************************************************************
# ***** ****
# *** VARIABLES ***
# ***** ****
# *****************************************************************************************************************
# --- Nothing Yet
# *****************************************************************************************************************
# ***** ****
# *** FUNCTIONS ***
# ***** ****
# *****************************************************************************************************************
# --- Nothing Yet
# *****************************************************************************************************************
# ***** ****
# *** CLASSES ***
# ***** ****
# *****************************************************************************************************************
class MessageController(BaseModel):
# ┏┓┓ ┓┏
# ┃ ┃┏┓┏┏ ┃┃┏┓┏┓┏
# ┗┛┗┗┻┛┛ ┗┛┗┻┛ ┛
# For MongoDB:
MESSAGES_COLLECTION = "_messages"
# ┏┓┳┓┳┳┳┓ ┏┓
# ┃ ┣┫┃┃┃┃ ━━ ┃ ┏┓┏┓┏┓╋┏┓
# ┗┛┛┗┗┛┻┛ ┗┛┛ ┗ ┗┻┗┗
async def insert(
self,
mongo_conn: AsyncMongo,
message: CoreMessageModel
) -> ObjectId:
# Simply insert the document:
return await mongo_conn.insert_one(
collection = self.MESSAGES_COLLECTION,
document = message,
raise_exception = True
)
async def bulk_write(
self,
mongo_conn: AsyncMongo,
mongo_operations
) -> int:
return await mongo_conn.bulk_write(
collection = self.MESSAGES_COLLECTION,
requests = mongo_operations
)
# ┏┓┳┓┳┳┳┓ ┳┓ •
# ┃ ┣┫┃┃┃┃ ━━ ┣┫┏┓╋┏┓┓┏┓┓┏┏┓
# ┗┛┛┗┗┛┻┛ ┛┗┗ ┗┛ ┗┗ ┗┛┗
async def count_messages(
self,
mongo_conn: AsyncMongo,
token_ids: List[ObjectId | str],
additional_filter: dict = None
) -> int:
"""
Just counts the no. of messages that match a given set of conditions.
:param mongo_conn: The instance of the database connector to use for the operation.
:param token_ids: The token ids of the accounts from which these messages must be fetched.
:param additional_filter: Any addition filters to use.
:return: The no. of messages that match the given conditions.
"""
# Prepare the filter:
if not isinstance(token_ids, list): token_ids = [token_ids]
token_ids = [ObjectId(t) for t in token_ids]
filter_json = {"tokenId": {"$in": token_ids}}
if additional_filter:
for k, v in additional_filter.items():
filter_json[k] = v
# Get the count of the documents that match the criteria:
count = await mongo_conn.count(
collection = self.MESSAGES_COLLECTION,
filter = filter_json,
raise_exception = True
)
# Done here:
return count
async def get_previews(
self,
mongo_conn: AsyncMongo,
token_ids: List[ObjectId | str],
limit: int = 100,
skip: int = 0,
additional_filter: dict = None
) -> List[CoreMessageModel] | None:
"""
Fetches many messages in one call, but leaves out the full payloads. This does not mark messages as read.
:param mongo_conn: The instance of the database connector to use for the operation.
:param token_ids: The token ids of the accounts from which these messages must be fetched.
:param limit: The max. no. of messages to retrieve in this call.
:param skip: The no. of initial messages to skip. Useful for pagination.
:param additional_filter: Any addition filters to use.
:return: The list of messages (as the message model). This list can be empty.
"""
# Prepare the filter:
if not isinstance(token_ids, list): token_ids = [token_ids]
token_ids = [ObjectId(t) for t in token_ids]
filter_json = {"tokenId": {"$in": token_ids}}
if additional_filter:
for k, v in additional_filter.items():
filter_json[k] = v
# We fetch the messages that are identified by a specific token id,
# with the specified fetching limits, while enforcing the sorting condition:
records = await mongo_conn.find_many(
collection = self.MESSAGES_COLLECTION,
filter = filter_json,
limit = limit,
skip = skip,
sort = {"ts": -1},
projection = {
"_id": True,
"ts": True,
"markedAsUnread": True,
"serviceType": True,
"client": True,
"clientMessageId": True,
"clientThreadId": True,
"isSent": True,
"isBroadcast": True,
"sentSuccessfully": True,
"aiSnippet": True,
"preview": True,
"message": {},
"tags": True,
"usedAi": True
},
raise_exception = True
)
# Convert the fetched records to instances of the data model and return:
return [CoreMessageModel(**record) for record in records]
async def get_messages(
self,
mongo_conn: AsyncMongo,
token_ids: List[ObjectId | str],
limit: int = 100,
skip: int = 0,
additional_filter: dict = None
) -> List[CoreMessageModel] | None:
"""
Fetches many full messages in one call.
:param mongo_conn: The instance of the database connector to use for the operation.
:param token_ids: The token ids of the accounts from which these messages must be fetched.
:param limit: The max. no. of messages to retrieve in this call.
:param skip: The no. of initial messages to skip. Useful for pagination.
:param additional_filter: Any addition filters to use.
:return: The list of messages (as the message model). This list can be empty.
"""
# Note down the timestamp at which this event occurred:
request_ts = date_time.get_current_utc_date_time(as_string = False)
# Prepare the filter:
if not isinstance(token_ids, list): token_ids = [token_ids]
token_ids = [ObjectId(t) for t in token_ids]
filter_json = {"tokenId": {"$in": token_ids}}
if additional_filter:
for k, v in additional_filter.items():
filter_json[k] = v
# We fetch the messages that are identified by a specific token id,
# with the specified fetching limits, while enforcing the sorting condition:
records = await mongo_conn.find_many(
collection = self.MESSAGES_COLLECTION,
filter = filter_json,
limit = limit,
skip = skip,
sort = {"ts": -1},
raise_exception = True
)
# We now mark these fetched messages as read through a bulk-write operation:
operations = [
UpdateOne(
filter = {"_id": record["_id"]},
update = [{
"$set": {
"readTs": {
"$cond": {
"if": {
"$or": [
{"$eq": ["$readTs", None]},
{"$eq": [{"$type": "$readTs"}, "missing"]}
]
},
"then": request_ts,
"else": "$readTs"
}
}
}
}],
upsert = False
) for record in records
]
updated_count = await mongo_conn.bulk_write(
collection = self.MESSAGES_COLLECTION,
requests = operations,
raise_exception = True
)
# Convert the fetched records to instances of the data model and return:
return [CoreMessageModel(**record) for record in records]
async def get_message(
self,
mongo_conn: AsyncMongo,
message_id: ObjectId | str,
) -> CoreMessageModel | None:
"""
Gets one message if you know its message id. Marks that message as read.
:param mongo_conn:
:param message_id:
:return:
"""
# Note down the timestamp at which this event occurred:
request_ts = date_time.get_current_utc_date_time(as_string = False)
# We fetch the whole payload of that one message
# while also marking it as read if not already marked:
record = await mongo_conn.find_one_and_update(
collection = self.MESSAGES_COLLECTION,
filter = {"_id": ObjectId(message_id)},
update = [{
"$set": {
"readTs": {
"$cond": {
"if": {
"$or": [
{"$eq": ["$readTs", None]},
{"$eq": [{"$type": "$readTs"}, "missing"]}
]
},
"then": request_ts,
"else": "$readTs"
}
}
}
}],
raise_exception = True
)
# If no such message was found:
if record is None: return None
# If a record was found,
# we return it as our data model:
return CoreMessageModel(**record)
# ┏┓┳┓┳┳┳┓ ┳┳ ┓
# ┃ ┣┫┃┃┃┃ ━━ ┃┃┏┓┏┫┏┓╋┏┓
# ┗┛┛┗┗┛┻┛ ┗┛┣┛┗┻┗┻┗┗
# ┛
# We don't support updating messages themselves,
# but we will allow updating fields like tags, marking as read or unread, etc.
# ┏┓┳┓┳┳┳┓ ┳┓ ┓
# ┃ ┣┫┃┃┃┃ ━━ ┃┃┏┓┃┏┓╋┏┓
# ┗┛┛┗┗┛┻┛ ┻┛┗ ┗┗ ┗┗
# No support whatsoever for deleting messages.
# *****************************************************************************************************************
# ***** ****
# *** MAIN PROGRAM ***
# ***** ****
# *****************************************************************************************************************
if __name__ == "__main__":
pass