made the bot async, fixed the bug where it goes offline

.gitignore

@@ -1,2 +1,3 @@
 .env
-venv/
+venv/
+learn.py

app.py

@@ -1,8 +1,7 @@
 import discord
+import aiohttp
 
 import os
-import json
-import requests
 import threading
 
 
@@ -34,7 +33,7 @@ async def help(ctx: discord.ApplicationContext):
                       \n2)Those that tag the bot.")
 
 
-def llm_output(question: str, context: str) -> str:
+async def llm_output(question: str, context: str) -> str:
     """
     Returns output from the LLM using the given user-question and retrived context
     """
@@ -54,14 +53,15 @@ def llm_output(question: str, context: str) -> str:
             'prompt': prompt,
             'stream': False
         }
-
-    response = requests.post(URL_LLM + "/api/generate", json=obj)
-    response_json = json.loads(response.text)
-
+    
+    async with aiohttp.ClientSession() as session:
+        async with session.post(URL_LLM + "/api/generate", json=obj) as response:
+            response_json = await response.json()
+    
     return response_json['response']
 
 
-def embedding_output(message: str) -> list:
+async def embedding_output(message: str) -> list:
     """
     Returns embeddings for the given message
 
@@ -70,13 +70,14 @@ def embedding_output(message: str) -> list:
 
     URL_EMBEDDING = 'https://robinroy03-fury-embeddings-endpoint.hf.space'
 
-    response = requests.post(URL_EMBEDDING + "/embedding", json={"text": message})
-    response_json = json.loads(response.text)
-
+    async with aiohttp.ClientSession() as session:
+        async with session.post(URL_EMBEDDING + "/embedding", json={"text": message}) as response:
+            response_json = await response.json()
+    
     return response_json['output']
 
 
-def db_output(embedding: list) -> dict:
+async def db_output(embedding: list) -> dict:
     """
     Returns the KNN results.
 
@@ -85,9 +86,10 @@ def db_output(embedding: list) -> dict:
 
     URL_DB = 'https://robinroy03-fury-db-endpoint.hf.space'
     
-    response = requests.post(URL_DB + "/query", json={"embeddings": embedding})
-    response_json = json.loads(response.text)
-
+    async with aiohttp.ClientSession() as session:
+        async with session.post(URL_DB + "/query", json={"embeddings": embedding}) as response:
+            response_json = await response.json()
+        
     return response_json
 
 
@@ -104,9 +106,10 @@ async def on_message(message):
     await message.reply(content="Your message was received, it'll take around 30 seconds for FURY to process an answer.")
 
     question = message.content.replace("<@1243428204124045385>", "")
-    embedding: list = embedding_output(question)
-    db_knn: dict = db_output(embedding)
-    llm_answer: str = llm_output(question, db_knn['matches'][0]['metadata']['text'])      # for the highest knn result (for the test only right now)  TODO: make this better
+    
+    embedding: list = await embedding_output(question)
+    db_knn: dict = await db_output(embedding)
+    llm_answer: str = await llm_output(question, db_knn['matches'][0]['metadata']['text'])      # for the highest knn result (for the test only right now)  TODO: make this better
 
     try:
         await message.reply(content=llm_answer[:1990], view=Like_Dislike())             # TODO: handle large responses (>2000)