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

Merge remote-tracking branch 'Remote_Origin/main' into override_threads

Browse Source
Rene Teigen 1 year ago
parent b74e0bf80e e489b20edd
commit 6b99f57fdf
7 changed files with 214 additions and 116 deletions
Show Stats Download Patch File Download Diff File

1
.gitignore vendored
Unescape View File

@ -5,6 +5,7 @@ __pycache__
#user files
.env
.vscode
*.sqlite
bot.pid
usage.txt
/dalleimages

297
cogs/gpt_3_commands_and_converser.py
Unescape View File

@ -581,17 +581,27 @@ 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(
"\nThis conversation has some context from earlier, which has been summarized as follows: "
EmbeddedConversationItem(self.CONVERSATION_STARTER_TEXT, 0)
)
new_conversation_history.append(summarized_text)
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(
"\nThis conversation has some context from earlier, which has been summarized as follows: ",
0,
)
)
new_conversation_history.append(EmbeddedConversationItem(summarized_text, 0))
new_conversation_history.append(
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
@ -622,19 +632,23 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
ctx = self.redo_users[after.author.id].ctx
await response_message.edit(content="Redoing prompt 🔄...")
edited_content = after.content
edited_content = await self.mention_to_username(after, after.content)
if after.channel.id in self.conversation_threads:
# Remove the last two elements from the history array and add the new <username>: prompt
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
overrides = self.conversation_threads[after.channel.id].get_overrides()
@ -648,9 +662,11 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
top_p_override=overrides["top_p"],
frequency_penalty_override=overrides["frequency_penalty"],
presence_penalty_override=overrides["presence_penalty"],
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.
@ -713,7 +729,7 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
if not user_api_key:
return
prompt = content
prompt = await self.mention_to_username(message, content)
await self.check_conversation_limit(message)
@ -763,7 +779,10 @@ 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
@ -778,7 +797,12 @@ 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
]
)
#set conversation overrides
@ -802,6 +826,20 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
response_text = response_text.replace("<|endofstatement|>", "")
return response_text
async def mention_to_username(self, ctx, message):
if not discord.utils.raw_mentions(message):
return message
else:
for mention in discord.utils.raw_mentions(message):
try:
user = await discord.utils.get_or_fetch(
ctx.guild, "member", mention
)
message = message.replace(f"<@{str(mention)}>", user.display_name)
except:
pass
return message
# ctx can be of type AppContext(interaction) or Message
async def encapsulated_send(
self,
@ -815,6 +853,8 @@ 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
@ -826,6 +866,13 @@ 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
@ -844,82 +891,91 @@ 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,
)
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.
# 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,
)
# 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
]
if not redo_request:
self.conversation_threads[conversation_id].history.append(
new_prompt_item
)
# 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]
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,
)
# remove duplicates from prompt_with_history
prompt_with_history = list(dict.fromkeys(prompt_with_history))
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.
# Sort the prompt_with_history by increasing timestamp
prompt_with_history.sort(key=lambda x: x.timestamp)
# 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,
)
# 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)
# 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
]
# 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]
)
prompt_with_history = "".join(
[item.text for item in prompt_with_history]
)
# 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
# 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)
# 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]
)
new_prompt = prompt_with_history
new_prompt = prompt_with_history + "\nGPTie: "
tokens = self.usage_service.count_tokens(new_prompt)
@ -943,7 +999,13 @@ 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)
@ -983,12 +1045,8 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
response_text = response_text.strip()
response_text = f"***{prompt}***\n\n{response_text}"
# If GPT3 tries to ping somebody, don't let it happen
if re.search(r"<@!?\d+>|<@&\d+>|<#\d+>", str(response_text)):
message = "I'm sorry, I can't mention users, roles, or channels."
await ctx.send_followup(message) if from_context else await ctx.reply(
message
)
# If gpt3 tries writing a user mention try to replace it with their name
response_text = await self.mention_to_username(ctx, response_text)
# If the user is conversing, add the GPT response to their conversation history.
if (
@ -996,9 +1054,12 @@ 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 (
@ -1032,9 +1093,12 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
custom_api_key=custom_api_key,
)
# Cleanse
# Cleanse again
response_text = self.cleanse_response(response_text)
# escape any other mentions like @here or @everyone
response_text = discord.utils.escape_mentions(response_text)
# If we don't have a response message, we are not doing a redo, send as a new message(s)
if not response_message:
if len(response_text) > self.TEXT_CUTOFF:
@ -1064,7 +1128,7 @@ 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
@ -1169,7 +1233,7 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
presence_penalty: float,
):
user = ctx.user
prompt = prompt.strip()
prompt = await self.mention_to_username(ctx, prompt.strip())
user_api_key = None
if USER_INPUT_API_KEYS:
@ -1250,6 +1314,33 @@ class GPT3ComCon(discord.Cog, name="GPT3ComCon"):
)
return
if opener:
opener = await self.mention_to_username(ctx, opener)
if not opener and not opener_file:
user_id_normalized = user.id
else:
user_id_normalized = ctx.author.id
if (
opener_file
): # only load in files if it's included in the command, if not pass on as normal
if opener_file.endswith(".txt"):
# Load the file and read it into opener
opener_file = EnvService.find_shared_file(
f"openers{separator}{opener_file}"
)
opener_file = await self.load_file(opener_file, ctx)
if (
not opener
): # if we only use opener_file then only pass on opener_file for the opening prompt
opener = opener_file
else:
opener = opener_file + opener
if not opener_file:
return
else:
pass
if private:
await ctx.respond(user.name + "'s private conversation with GPT3")
thread = await ctx.channel.create_thread(
@ -1307,11 +1398,11 @@ 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)
)
# Set user as thread owner before sending anything that can error and leave the thread unowned
@ -1335,7 +1426,10 @@ 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
@ -1344,7 +1438,9 @@ 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,
temp_override=overrides["temperature"],
top_p_override=overrides["top_p"],
@ -1643,6 +1739,7 @@ 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(

11
cogs/image_prompt_optimizer.py
Unescape View File

@ -74,7 +74,8 @@ class ImgPromptOptimizer(discord.Cog, name="ImgPromptOptimizer"):
user = ctx.user
final_prompt = self.OPTIMIZER_PRETEXT
final_prompt += prompt
# replace mentions with nicknames for the prompt
final_prompt += await self.converser_cog.replace_mention(ctx, prompt)
# If the prompt doesn't end in a period, terminate it.
if not final_prompt.endswith("."):
@ -100,12 +101,8 @@ class ImgPromptOptimizer(discord.Cog, name="ImgPromptOptimizer"):
# also relatively cost-effective
response_text = response["choices"][0]["text"]
if re.search(r"<@!?\d+>|<@&\d+>|<#\d+>", response_text):
await ctx.respond(
"I'm sorry, I can't mention users, roles, or channels."
)
return
# escape any mentions
response_text = discord.utils.escape_mentions(response_text)
# If the response_message is > 75 words, concatenate to the last 70th word
# TODO Temporary workaround until prompt is adjusted to make the optimized prompts shorter.

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

@ -123,3 +123,8 @@ 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