aljaz/chatgpt-telegram-bot
1
1
Fork 0
mirror of https://github.com/aljazceru/chatgpt-telegram-bot.git synced 2025-12-20 14:14:52 +01:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'main' into add-balance-function

Browse Source
This commit is contained in:
AlexHTW
2023-03-24 11:32:16 +01:00
parent 29ecd1723c 5d2c288789
commit 17028b8e57
3 changed files with 47 additions and 127 deletions
Download Patch File Download Diff File Expand all files Collapse all files

65
.env.example
View File

@@ -6,68 +6,3 @@ TELEGRAM_BOT_TOKEN="XXX"
# Comma separated list of telegram user IDs, or * to allow all. First USER_ID has access to admin commands. # Comma separated list of telegram user IDs, or * to allow all. First USER_ID has access to admin commands.
ALLOWED_TELEGRAM_USER_IDS="USER_ID_1,USER_ID_2" ALLOWED_TELEGRAM_USER_IDS="USER_ID_1,USER_ID_2"
# Comma separated list of user budgets, USD limit per month, or * to allow all users unlimited usage
# same order of users as in ALLOWED_TELEGRAM_USER_IDS
MONTHLY_USER_BUDGETS="100.0,100.0"
# Guest Budget, USD limit per month for non-user requests in group chats with users
MONTHLY_GUEST_BUDGET="20.0"
# Proxy to be used for OpenAI and Telegram bot
PROXY="http://localhost:8080"
# Proxy to be used for OpenAI and Telegram bot
OPENAI_MODEL="gpt-3.5-turbo"
# A system message that sets the tone and controls the behavior of the assistant
ASSISTANT_PROMPT="You are a helpful assistant."
# Whether to show OpenAI token usage information after each response
SHOW_USAGE=false
# Upper bound on how many tokens the ChatGPT API will return
MAX_TOKENS=1200
# Max number of messages to keep in memory, after which the conversation will be summarised
MAX_HISTORY_SIZE=10
# Max minutes a conversation will live, after which the conversation will be reset to avoid excessive token usage
MAX_CONVERSATION_AGE_MINUTES=180
# Whether to answer to voice messages with the transcript or with a ChatGPT response of the transcript
VOICE_REPLY_WITH_TRANSCRIPT_ONLY=false
# How many chat completion choices to generate for each input message
N_CHOICES=1
# Number between 0 and 2. Higher values like 0.8 will make the output more random,
# while lower values like 0.2 will make it more focused and deterministic
TEMPERATURE=1.0
# Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far,
# increasing the model's likelihood to talk about new topics
PRESENCE_PENALTY=0
# Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far,
# decreasing the model's likelihood to repeat the same line verbatim
FREQUENCY_PENALTY=0
# The DALL·E generated image size
IMAGE_SIZE="512x512"
# Group trigger keyword, if set, the bot will only respond to messages that start with this keyword
# Useful for bots added to groups with privacy mode disabled
GROUP_TRIGGER_KEYWORD=""
# Whether transcriptions should be ignored in group chats
IGNORE_GROUP_TRANSCRIPTIONS=true
# USD-price per 1000 tokens for cost information in usage statistics, see https://openai.com/pricing
TOKEN_PRICE=0.002
# USD-prices per image for the sizes 256x256,512x512,1024x1024 for cost information in usage statistics
IMAGE_PRICES="0.016,0.018,0.02"
# USD-price for 1 minute of audio transcription for cost information in usage statistics
TRANSCRIPTION_PRICE=0.006

80
README.md
View File

@@ -41,57 +41,37 @@ PRs are always welcome!
## Getting started ## Getting started
### Configuration ### Configuration
Customize the configuration by copying `.env.example` and renaming it to `.env`, then editing the parameters as desired: Customize the configuration by copying `.env.example` and renaming it to `.env`, then editing the required parameters as desired:
```bash
OPENAI_API_KEY="YOUR_OPENAI_API_KEY"
TELEGRAM_BOT_TOKEN="YOUR_TELEGRAM_BOT_TOKEN"
# Optional parameters | Parameter | Description |
ALLOWED_TELEGRAM_USER_IDS="USER_ID_1,USER_ID_2,..." # Defaults to "*" (everyone) |-----------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
MONTHLY_USER_BUDGETS="BUDGET_USER_ID_1,BUDGET_USER_ID_2,..." # Defaults to "*" (no restrictions) | `OPENAI_API_KEY` | Your OpenAI API key, you can get it from [here](https://platform.openai.com/account/api-keys) |
MONTHLY_GUEST_BUDGET="20.0" # Defaults to 100.0 | `TELEGRAM_BOT_TOKEN` | Your Telegram bot's token, obtained using [BotFather](http://t.me/botfather) (see [tutorial](https://core.telegram.org/bots/tutorial#obtain-your-bot-token)) |
PROXY="YOUR_PROXY" # e.g. "http://localhost:8080", defaults to none | `ALLOWED_TELEGRAM_USER_IDS` | A comma-separated list of Telegram user IDs that are allowed to interact with the bot (use [getidsbot](https://t.me/getidsbot) to find your user ID). First USER_ID has access to admin commands. **Note**: by default, *everyone* is allowed and has access to admin commands (`*`)|
OPENAI_MODEL="gpt-3.5-turbo" # Defaults to gpt-3.5-turbo
ASSISTANT_PROMPT="Custom prompt" # Defaults to "You are a helpful assistant." ### Optional configuration
SHOW_USAGE=true # Defaults to false | Parameter | Description | Default value |
MAX_TOKENS=2000 # Defaults to 1200 |------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------|
MAX_HISTORY_SIZE=15 # Defaults to 10 | `MONTHLY_USER_BUDGETS` | A comma-separated list of $-amounts per user from list `ALLOWED_TELEGRAM_USER_IDS` to set custom usage limit of OpenAI API costs for each. **Note**: by default, *no limits* for anyone (`*`) | `*` |
MAX_CONVERSATION_AGE_MINUTES=120 # Defaults to 180 (3h) | `MONTHLY_GUEST_BUDGET` | $-amount as usage limit for all guest users. Guest users are users in group chats that are not in the `ALLOWED_TELEGRAM_USER_IDS` list. Value is ignored if no usage limits are set in user budgets (`MONTHLY_USER_BUDGETS`="*") | `100.0` |
VOICE_REPLY_WITH_TRANSCRIPT_ONLY=false # Defaults to true | `PROXY` | Proxy to be used for OpenAI and Telegram bot (e.g. `http://localhost:8080`) | - |
N_CHOICES=1 # Defaults to 1 | `OPENAI_MODEL` | The OpenAI model to use for generating responses | `gpt-3.5-turbo` |
TEMPERATURE=1.0 # Defaults to 1.0 | `ASSISTANT_PROMPT` | A system message that sets the tone and controls the behavior of the assistant | `You are a helpful assistant.` |
PRESENCE_PENALTY=0 # Defaults to 0 | `SHOW_USAGE` | Whether to show OpenAI token usage information after each response | false |
FREQUENCY_PENALTY=0 # Defaults to 0 | `MAX_TOKENS` | Upper bound on how many tokens the ChatGPT API will return | 2000 |
IMAGE_SIZE="256x256" # Defaults to 512x512 | `MAX_HISTORY_SIZE` | Max number of messages to keep in memory, after which the conversation will be summarised to avoid excessive token usage | 10 |
GROUP_TRIGGER_KEYWORD="@bot" # Defaults to "" (no keyword required) | `MAX_CONVERSATION_AGE_MINUTES` | Maximum number of minutes a conversation should live, after which the conversation will be reset | 120 |
IGNORE_GROUP_TRANSCRIPTIONS=true # Whether transcriptions should be ignored in group chats. Defaults to true | `VOICE_REPLY_WITH_TRANSCRIPT_ONLY` | Whether to answer to voice messages with the transcript only or with a ChatGPT response of the transcript | false |
TOKEN_PRICE=0.002 # Defaults to 0.002, current price: https://openai.com/pricing | `N_CHOICES` | Number of answers to generate for each input message | 1 |
IMAGE_PRICES="0.016,0.018,0.02" # Defaults to OpenAI Dall-E pricing for sizes 256x256,512x512,1024x1024 | `TEMPERATURE` | Number between 0 and 2. Higher values like 0.8 will make the output more random | 1.0 |
TRANSCRIPTION_PRICE=0.006 # Defaults to minute price of OpenAI Whisper of 0.006 | `PRESENCE_PENALTY` | Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far | 0 |
``` | `FREQUENCY_PENALTY` | Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far | 0 |
* `OPENAI_API_KEY`: Your OpenAI API key, you can get it from [here](https://platform.openai.com/account/api-keys) | `IMAGE_SIZE` | The DALL·E generated image size. Allowed values: "256x256", "512x512", or "1024x1024" | "256x256" |
* `TELEGRAM_BOT_TOKEN`: Your Telegram bot's token, obtained using [BotFather](http://t.me/botfather) (see [tutorial](https://core.telegram.org/bots/tutorial#obtain-your-bot-token)) | `GROUP_TRIGGER_KEYWORD` | If set, the bot in group chats will only respond to messages that start with this keyword | "" |
* `ALLOWED_TELEGRAM_USER_IDS`: A comma-separated list of Telegram user IDs that are allowed to interact with the bot (use [getidsbot](https://t.me/getidsbot) to find your user ID). First USER_ID has access to admin commands. **Note**: by default, *everyone* is allowed (`*`) and has access to admin commands. | `IGNORE_GROUP_TRANSCRIPTIONS` | If set to true, the bot will not process transcriptions in group chats | true |
* `MONTHLY_USER_BUDGETS`: A comma-separated list of $-amounts per user from list `ALLOWED_TELEGRAM_USER_IDS` to set custom usage limit of OpenAI API costs for each. **Note**: by default, *no limits* for anyone (`*`) | `TOKEN_PRICE` | $-price per 1000 tokens used to compute cost information in usage statistics (https://openai.com/pricing) | 0.002 |
* `MONTHLY_GUEST_BUDGET`: $-amount as usage limit for all guest users. Guest users are users in group chats that are not in the `ALLOWED_TELEGRAM_USER_IDS` list. Value is ignored if no usage limits are set in user budgets (`MONTHLY_USER_BUDGETS`="*") | `IMAGE_PRICES` | A comma-separated list with 3 elements of prices for the different image sizes: 256x256, 512x512 and 1024x1024 | "0.016,0.018,0.02" |
* `PROXY`: Proxy to be used for OpenAI and Telegram bot | `TRANSCRIPTION_PRICE` | USD-price for one minute of audio transcription | 0.006 |
* `OPENAI_MODEL`: Define which OpenAI model to use (default is `gpt-3.5-turbo`)
* `ASSISTANT_PROMPT`: A system message that sets the tone and controls the behavior of the assistant
* `SHOW_USAGE`: Whether to show OpenAI token usage information after each response
* `MAX_TOKENS`: Upper bound on how many tokens the ChatGPT API will return
* `MAX_HISTORY_SIZE`: Max number of messages to keep in memory, after which the conversation will be summarised to avoid excessive token usage ([#34](https://github.com/n3d1117/chatgpt-telegram-bot/issues/34))
* `MAX_CONVERSATION_AGE_MINUTES`: Maximum number of minutes a conversation should live, after which the conversation will be reset to avoid excessive token usage
* `VOICE_REPLY_WITH_TRANSCRIPT_ONLY`: Whether to answer to voice messages with the transcript only or with a ChatGPT response of the transcript ([#38](https://github.com/n3d1117/chatgpt-telegram-bot/issues/38))
* `N_CHOICES`: Number of answers to generate for each input message
* `TEMPERATURE`: Number between 0 and 2. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic
* `PRESENCE_PENALTY`: Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far, increasing the model's likelihood to talk about new topics
* `FREQUENCY_PENALTY`: Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far, decreasing the model's likelihood to repeat the same line verbatim
* `IMAGE_SIZE`: The DALL·E generated image size. Allowed values: "256x256", "512x512", or "1024x1024"
* `GROUP_TRIGGER_KEYWORD`: If set, the bot will only respond to messages that start with this keyword. This is useful for bots added to groups with privacy mode disabled. **Note**: by default, *no keyword* is required (`""`)
* `IGNORE_GROUP_TRANSCRIPTIONS`: If set to true, the bot will not process transcriptions in group chats
* `TOKEN_PRICE`: USD-price per 1000 tokens for cost information in usage statistics. Defaults to [OpenAI price](https://openai.com/pricing) for gpt-3.5-turbo
* `IMAGE_PRICES`: A comma-separated list with 3 elements of prices for the different image sizes 256x256, 512x512 and 1024x1024. Defaults to [OpenAI prices](https://openai.com/pricing) for Dall-E.
* `TRANSCRIPTION_PRICE`: USD-price for one minute of audio transcription. Defaults to [OpenAI price](https://openai.com/pricing) for Whisper
Check out the [official API reference](https://platform.openai.com/docs/api-reference/chat) for more details. Check out the [official API reference](https://platform.openai.com/docs/api-reference/chat) for more details.

19
bot/openai_helper.py
View File

@@ -6,7 +6,6 @@ import tiktoken
import openai import openai
import requests import requests
import json import json
from datetime import date from datetime import date
@@ -215,18 +214,24 @@ class OpenAIHelper:
encoding = tiktoken.encoding_for_model(model) encoding = tiktoken.encoding_for_model(model)
except KeyError: except KeyError:
encoding = tiktoken.get_encoding("gpt-3.5-turbo") encoding = tiktoken.get_encoding("gpt-3.5-turbo")
if model in GPT_ALL_MODELS:
if model in GPT_3_MODELS:
tokens_per_message = 4 # every message follows <im_start>{role/name}\n{content}<im_end>\n
tokens_per_name = -1 # if there's a name, the role is omitted
elif model in GPT_4_MODELS + GPT_4_32K_MODELS:
tokens_per_message = 3
tokens_per_name = 1
else:
raise NotImplementedError(f"""num_tokens_from_messages() is not implemented for model {model}.""")
num_tokens = 0 num_tokens = 0
for message in messages: for message in messages:
num_tokens += 4 # every message follows <im_start>{role/name}\n{content}<im_end>\n num_tokens += tokens_per_message
for key, value in message.items(): for key, value in message.items():
num_tokens += len(encoding.encode(value)) num_tokens += len(encoding.encode(value))
if key == "name": # if there's a name, the role is omitted if key == "name":
num_tokens += -1 # role is always required and always 1 token num_tokens += tokens_per_name
num_tokens += 2 # every reply is primed with <im_start>assistant num_tokens += 2 # every reply is primed with <im_start>assistant
return num_tokens return num_tokens
else:
raise NotImplementedError(f"__count_tokens() is not presently implemented for model {model}")
def get_grant_balance(self): def get_grant_balance(self):
"""Gets remaining grant balance for new users from OpenAI API. """Gets remaining grant balance for new users from OpenAI API.