open-source-mirrors
/
GPT3Discord
mirror of https://github.com/Kav-K/GPT3Discord
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Stabilize pinecone conversations

Browse Source
Kaveen Kumarasinghe 1 year ago
parent 7fa80f5f0b
commit 6bff3b2cb5
5 changed files with 131 additions and 96 deletions
Show Stats Download Patch File Download Diff File

206
cogs/gpt_3_commands_and_converser.py
Unescape View File

@ -580,17 +580,17 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
summarized_text = response["choices"][0]["text"]
new_conversation_history = []
new_conversation_history.append(self.CONVERSATION_STARTER_TEXT)
new_conversation_history.append(EmbeddedConversationItem(self.CONVERSATION_STARTER_TEXT,0))
new_conversation_history.append(
"\nThis conversation has some context from earlier, which has been summarized as follows: "
EmbeddedConversationItem("\nThis conversation has some context from earlier, which has been summarized as follows: ",0)
)
new_conversation_history.append(summarized_text)
new_conversation_history.append(EmbeddedConversationItem(summarized_text,0))
new_conversation_history.append(
"\nContinue the conversation, paying very close attention to things <username> told you, such as their name, and personal details.\n"
EmbeddedConversationItem("\nContinue the conversation, paying very close attention to things <username> told you, such as their name, and personal details.\n",0)
)
# Get the last entry from the thread's conversation history
new_conversation_history.append(
self.conversation_threads[message.channel.id].history[-1] + "\n"
EmbeddedConversationItem(self.conversation_threads[message.channel.id].history[-1] + "\n",0)
)
self.conversation_threads[message.channel.id].history = new_conversation_history
@ -628,22 +628,29 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
self.conversation_threads[
after.channel.id
].history = self.conversation_threads[after.channel.id].history[:-2]
self.conversation_threads[after.channel.id].history.append(
f"\n{after.author.display_name}: {after.content}<|endofstatement|>\n"
)
edited_content = "".join(
self.conversation_threads[after.channel.id].history
)
pinecone_dont_reinsert = None
if not self.pinecone_service:
self.conversation_threads[after.channel.id].history.append(
EmbeddedConversationItem(f"\n{after.author.display_name}: {after.content}<|endofstatement|>\n",0)
)
self.conversation_threads[after.channel.id].count += 1
print("-------------------------- Conversation POINT 1")
print(self.conversation_threads[ctx.channel.id].history)
print("---------------------------- END Conersation POINT 1")
await self.encapsulated_send(
id=after.channel.id,
prompt=edited_content,
ctx=ctx,
response_message=response_message,
edited_request=True,
)
self.redo_users[after.author.id].prompt = after.content
if not self.pinecone_service:
self.redo_users[after.author.id].prompt = edited_content
async def check_and_launch_moderations(self, guild_id, alert_channel_override=None):
# Create the moderations service.
@ -756,7 +763,7 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
if not self.pinecone_service:
self.conversation_threads[message.channel.id].history.append(
f"\n'{message.author.display_name}': {prompt} <|endofstatement|>\n"
EmbeddedConversationItem(f"\n'{message.author.display_name}': {prompt} <|endofstatement|>\n",0)
)
# increment the conversation counter for the user
@ -771,7 +778,7 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
primary_prompt = prompt
else:
primary_prompt = "".join(
self.conversation_threads[message.channel.id].history
[item.text for item in self.conversation_threads[message.channel.id].history]
)
await self.encapsulated_send(
@ -801,6 +808,8 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
presence_penalty_override=None,
from_g_command=False,
custom_api_key=None,
edited_request=False,
redo_request=False,
):
new_prompt = prompt + "\nGPTie: " if not from_g_command else prompt
@ -812,6 +821,13 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
# Pinecone is enabled, we will create embeddings for this conversation.
if self.pinecone_service and ctx.channel.id in self.conversation_threads:
# Delete "GPTie: <|endofstatement|>" from the user's conversation history if it exists
# check if the text attribute for any object inside self.conversation_threads[converation_id].history
# contains ""GPTie: <|endofstatement|>"", if so, delete
for item in self.conversation_threads[ctx.channel.id].history:
if item.text.strip() == "GPTie:<|endofstatement|>":
self.conversation_threads[ctx.channel.id].history.remove(item)
# The conversation_id is the id of the thread
conversation_id = ctx.channel.id
@ -830,82 +846,90 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
str(datetime.datetime.now().timestamp()).replace(".", "")
)
starter_conversation_item = EmbeddedConversationItem(
str(self.conversation_threads[ctx.channel.id].history[0]), 0
)
self.conversation_threads[ctx.channel.id].history[
0
] = starter_conversation_item
new_prompt_item = EmbeddedConversationItem(new_prompt, timestamp)
self.conversation_threads[conversation_id].history.append(
new_prompt_item
)
# Create and upsert the embedding for the conversation id, prompt, timestamp
await self.pinecone_service.upsert_conversation_embedding(
self.model,
conversation_id,
new_prompt,
timestamp,
custom_api_key=custom_api_key,
)
if not redo_request:
self.conversation_threads[conversation_id].history.append(
new_prompt_item
)
embedding_prompt_less_author = await self.model.send_embedding_request(
prompt_less_author, custom_api_key=custom_api_key
) # Use the version of the prompt without the author's name for better clarity on retrieval.
print("-------------------------- Conversation POINT 2")
print(self.conversation_threads[ctx.channel.id].history)
print("---------------------------- END Conersation POINT 2")
if edited_request:
new_prompt = "".join([item.text for item in self.conversation_threads[ctx.channel.id].history])
self.redo_users[ctx.author.id].prompt = new_prompt
else:
# Create and upsert the embedding for the conversation id, prompt, timestamp
await self.pinecone_service.upsert_conversation_embedding(
self.model,
conversation_id,
new_prompt,
timestamp,
custom_api_key=custom_api_key,
)
# Now, build the new prompt by getting the X most similar with pinecone
similar_prompts = self.pinecone_service.get_n_similar(
conversation_id,
embedding_prompt_less_author,
n=self.model.num_conversation_lookback,
)
embedding_prompt_less_author = await self.model.send_embedding_request(
prompt_less_author, custom_api_key=custom_api_key
) # Use the version of the prompt without the author's name for better clarity on retrieval.
# When we are in embeddings mode, only the pre-text is contained in self.conversation_threads[message.channel.id].history, so we
# can use that as a base to build our new prompt
prompt_with_history = [
self.conversation_threads[ctx.channel.id].history[0]
]
# Now, build the new prompt by getting the X most similar with pinecone
similar_prompts = self.pinecone_service.get_n_similar(
conversation_id,
embedding_prompt_less_author,
n=self.model.num_conversation_lookback,
)
# Append the similar prompts to the prompt with history
prompt_with_history += [
EmbeddedConversationItem(prompt, timestamp)
for prompt, timestamp in similar_prompts
]
# When we are in embeddings mode, only the pre-text is contained in self.conversation_threads[message.channel.id].history, so we
# can use that as a base to build our new prompt
prompt_with_history = [
self.conversation_threads[ctx.channel.id].history[0]
]
# Append the similar prompts to the prompt with history
prompt_with_history += [
EmbeddedConversationItem(prompt, timestamp)
for prompt, timestamp in similar_prompts
]
print("-------------------------- Conversation POINT 3")
print(self.conversation_threads[ctx.channel.id].history)
print("---------------------------- END Conersation POINT 3")
# iterate UP TO the last X prompts in the history
for i in range(
1,
min(
len(self.conversation_threads[ctx.channel.id].history),
self.model.num_static_conversation_items,
),
):
prompt_with_history.append(
self.conversation_threads[ctx.channel.id].history[-i]
)
# iterate UP TO the last X prompts in the history
for i in range(
1,
min(
len(self.conversation_threads[ctx.channel.id].history),
self.model.num_static_conversation_items,
),
):
prompt_with_history.append(
self.conversation_threads[ctx.channel.id].history[-i]
)
# remove duplicates from prompt_with_history and set the conversation history
prompt_with_history = list(dict.fromkeys(prompt_with_history))
self.conversation_threads[ctx.channel.id].history = prompt_with_history
# remove duplicates from prompt_with_history
prompt_with_history = list(dict.fromkeys(prompt_with_history))
# Sort the prompt_with_history by increasing timestamp if pinecone is enabled
if self.pinecone_service:
prompt_with_history.sort(key=lambda x: x.timestamp)
# Sort the prompt_with_history by increasing timestamp
prompt_with_history.sort(key=lambda x: x.timestamp)
# Ensure that the last prompt in this list is the prompt we just sent (new_prompt_item)
if prompt_with_history[-1] != new_prompt_item:
try:
prompt_with_history.remove(new_prompt_item)
except ValueError:
pass
prompt_with_history.append(new_prompt_item)
# Ensure that the last prompt in this list is the prompt we just sent (new_prompt_item)
if prompt_with_history[-1] != new_prompt_item:
try:
prompt_with_history.remove(new_prompt_item)
except ValueError:
pass
prompt_with_history.append(new_prompt_item)
prompt_with_history = "".join(
[item.text for item in prompt_with_history]
)
prompt_with_history = "".join(
[item.text for item in prompt_with_history]
)
new_prompt = prompt_with_history + "\nGPTie: "
new_prompt = prompt_with_history
tokens = self.usage_service.count_tokens(new_prompt)
@ -929,7 +953,7 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
# Check again if the prompt is about to go past the token limit
new_prompt = (
"".join(self.conversation_threads[id].history) + "\nGPTie: "
"".join([item.text for item in self.conversation_threads[id].history]) + "\nGPTie: "
)
tokens = self.usage_service.count_tokens(new_prompt)
@ -950,6 +974,10 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
await self.end_conversation(ctx)
return
print("-------------------------- BEFORE MODEL REQUEST")
print(self.conversation_threads[ctx.channel.id].history)
print("---------------------------- BEFORE MODEL REQUEST")
# Send the request to the model
response = await self.model.send_request(
new_prompt,
@ -982,9 +1010,10 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
and not from_g_command
and not self.pinecone_service
):
self.conversation_threads[id].history.append(
"\nGPTie: " + str(response_text) + "<|endofstatement|>\n"
)
if not redo_request:
self.conversation_threads[id].history.append(
EmbeddedConversationItem("\nGPTie: " + str(response_text) + "<|endofstatement|>\n",0)
)
# Embeddings case!
elif (
@ -1050,7 +1079,7 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
)
self.redo_users[ctx.author.id] = RedoUser(
prompt, ctx, ctx, actual_response_message
new_prompt, ctx, ctx, actual_response_message
)
self.redo_users[ctx.author.id].add_interaction(
actual_response_message.id
@ -1070,6 +1099,8 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
if ctx.channel.id in self.awaiting_thread_responses:
self.awaiting_thread_responses.remove(ctx.channel.id)
# Error catching for OpenAI model value errors
except ValueError as e:
if from_context:
@ -1280,11 +1311,11 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
# Append the starter text for gpt3 to the user's history so it gets concatenated with the prompt later
if minimal or opener_file:
self.conversation_threads[thread.id].history.append(
self.CONVERSATION_STARTER_TEXT_MINIMAL
EmbeddedConversationItem(self.CONVERSATION_STARTER_TEXT_MINIMAL,0)
)
elif not minimal:
self.conversation_threads[thread.id].history.append(
self.CONVERSATION_STARTER_TEXT
EmbeddedConversationItem(self.CONVERSATION_STARTER_TEXT,0)
)
await thread.send(
@ -1302,7 +1333,7 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
if not self.pinecone_service:
self.conversation_threads[thread.id].history.append(
f"\n'{ctx.author.display_name}': {opener} <|endofstatement|>\n"
EmbeddedConversationItem(f"\n'{ctx.author.display_name}': {opener} <|endofstatement|>\n",0)
)
self.conversation_threads[thread.id].count += 1
@ -1311,7 +1342,7 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
thread.id,
opener
if thread.id not in self.conversation_threads or self.pinecone_service
else "".join(self.conversation_threads[thread.id].history),
else "".join([item.text for item in self.conversation_threads[thread.id].history]),
thread_message,
custom_api_key=user_api_key,
)
@ -1608,6 +1639,7 @@ class RedoButton(discord.ui.Button["ConversationView"]):
ctx=ctx,
response_message=response_message,
custom_api_key=self.custom_api_key,
redo_request=True,
)
else:
await interaction.response.send_message(

2
gpt3discord.py
Unescape View File

@ -24,7 +24,7 @@ from models.openai_model import Model
from models.usage_service_model import UsageService
from models.env_service_model import EnvService
__version__ = "5.1.1"
__version__ = "5.2"
"""
The pinecone service is used to store and retrieve conversation embeddings.

10
models/openai_model.py
Unescape View File

@ -56,8 +56,8 @@ class Model:
self._summarize_threshold = 2500
self.model_max_tokens = 4024
self._welcome_message_enabled = True
self._num_static_conversation_items = 6
self._num_conversation_lookback = 10
self._num_static_conversation_items = 8
self._num_conversation_lookback = 6
try:
self.IMAGE_SAVE_PATH = os.environ["IMAGE_SAVE_PATH"]
@ -425,7 +425,7 @@ class Model:
await self.valid_text_request(response)
print(response["choices"][0]["text"])
#print(response["choices"][0]["text"])
return response
@ -482,9 +482,9 @@ class Model:
) as resp:
response = await resp.json()
# print(f"Payload -> {payload}")
# print(f"Response -> {response}")
# Parse the total tokens used for this request and response pair from the response
await self.valid_text_request(response)
print(f"Response -> {response}")
return response
@ -559,7 +559,7 @@ class Model:
) as resp:
response = await resp.json()
print(response)
#print(response)
image_urls = []
for result in response["data"]:

4
models/pinecone_service_model.py
Unescape View File

@ -23,8 +23,6 @@ class PineconeService:
# Split the text into 512 character chunks
chunks = [text[i : i + 500] for i in range(0, len(text), 500)]
for chunk in chunks:
print("The split chunk is ", chunk)
# Create an embedding for the split chunk
embedding = await model.send_embedding_request(
chunk, custom_api_key=custom_api_key
@ -61,7 +59,7 @@ class PineconeService:
include_metadata=True,
filter={"conversation_id": conversation_id},
)
print(response)
#print(response)
relevant_phrases = [
(match["id"], match["metadata"]["timestamp"])
for match in response["matches"]

5
models/user_model.py
Unescape View File

@ -105,3 +105,8 @@ class EmbeddedConversationItem:
def __ne__(self, other):
return not self.__eq__(other)
# Make it such that if there is an arry with these EmbeddedConversationItems, if we "".join the array, each item will
# return the .text attribute
def __format__(self, format_spec):
return self.text
Write Preview
Loading…
Cancel
Save