<div align="center">
<img style="display: block;" align="center" src="https://github.com/maxvfischer/askai/blob/main/images/logo.png?raw=True"/>
</div>
`askai` is a CLI integration with OpenAI's GPT3, enabling you to ask questions and
receive the answers straight in your terminal.
| ❗ **Other model integrations** ❗ |
|-------------------------------------------------------------------------------------------------------------------------------------------------------------|
| Currently, `askai` only supports integration with OpenAI. But as soon as other NLP API-endpoints start popping up, I will work on integrating them as well. |
## Installation
You can either install it through pip
```bash
pip install askai
```
or directly from the repo
```
git clone git@github.com:maxvfischer/askai.git
cd askai
pip install .
```
## Initialize askai
`askai` needs to be initialized to set up the default config and to connect your
OpenAI API key. Note that the key is only stored locally in `~/.askai/key`.
```bash
askai init
```
## Create OpenAI API-key
For `askai` to work, you need to add an OpenAI API-key. When creating an OpenAI account,
they give you $18 to use for free. After that you need to set up a paid account to
continue using their service.
During the development and testing of this CLI, I used $0.67.
An OpenAI API-key can be created by:
1. Creating an account on OpenAI: https://openai.com/api/
2. Logging in and click on `New API keys`
3. Click `Create new secret key`
## How to use
### Simple question
Ask a question using your saved config.
```
askai "<QUESTION>"
```
### Override config
It's possible to override the default config by using arguments:
```
askai "<QUESTION>" --num-answers <INT> --model <MODEL_STRING> --temperature <FLOAT> --top_p <FLOAT> --max-tokens <INT> --frequency-penalty <FLOAT> --presence_penalty <FLOAT>
```
| **Argument** | **Allowed values** | **Description** |
|---------------------|----------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------|
| --num-answers or -n | \>0 | Number of answers to generate. Note that more answers consume more tokens |
| --model or -m | See list below | Which model to use. See list of available models below. |
| --temperature or -t | 0.0 <= t <= 1.0 | What sampling temperature to use. Higher value makes the model more "creative". Do not use at the same time as `top-p`. |
| --top-p | 0.0 <= top_p <= 1.0 | What sampling nucleus to use. The model considers the results of the tokens with top_p probability mass. Do not use at the same time as `temperature`. |
| --max-tokens | \>0 | Maximum number of tokens used per question (incl. question + answer) |
| --frequency-penalty | -2.0 <= frequency_penalty <= 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 | -2.0 <= presence_penalty <= 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. |
## Update config
If you find yourself overriding the config a lot when asking questions, you can update the default config instead.
### Update all config values
```bash
askai config update all
```
### Update individual config values
```bash
askai config update num-answers
askai config update model
askai config update temperature
askai config update top-p
askai config update max-tokens
askai config update frequency-penalty
askai config update presence-penalty
```
### Reset to default config
```bash
askai config reset
```
### See current config
```
askai config show
```
## Update API-key
It's possible to update the API-key without re-initializing the CLI.
### Overwrite current API-key
```
askai key add
```
### Remove current API-key
```
askai key remove
```
## Available models
This list was updated from OpenAI's website on 2022-12-20 and might go out of date at any time. Please
check [here](https://beta.openai.com/docs/models) to see an accurate list.
#### Text-generating models
| --model | Description | Max tokens |
|------------------|----------------------------------------------|------------|
| text-davinci-003 | The most capable GPT3 model | 4000 |
| text-curie-001 | Worse than davinci. Still capable and faster | 2048 |
| text-babbage-001 | Can do straight forward tasks. Very fast | 2048 |
| text-ada-001 | Capable of very simple tasks. Very fast | 2048 |
#### Code-generating models
| --model | Description | Max tokens |
|------------------|-------------------------------------------|------------|
| code-davinci-002 | Most capable code-generating model. | 8000 |
| code-cushman-001 | Almost as capable as davinci, but faster. | 2048 |
## Important notes
Note that the answers generated by OpenAI and shown by `askai` is by no means a "truth".
You always need to be skeptical. And of course, don't execute generated commands or code
without verifying that it does what you asked for.
Raw data
{
"_id": null,
"home_page": "https://github.com/maxvfischer/askai",
"name": "askai",
"maintainer": "",
"docs_url": null,
"requires_python": "",
"maintainer_email": "",
"keywords": "",
"author": "Max Fischer",
"author_email": "",
"download_url": "https://files.pythonhosted.org/packages/7f/1f/f3ac4f0badcb8c4a8d63943b0408b1351444967b51d2cbcb13f51026bfbf/askai-1.0.5.tar.gz",
"platform": null,
"description": "<div align=\"center\">\n <img style=\"display: block;\" align=\"center\" src=\"https://github.com/maxvfischer/askai/blob/main/images/logo.png?raw=True\"/>\n</div>\n\n`askai` is a CLI integration with OpenAI's GPT3, enabling you to ask questions and \nreceive the answers straight in your terminal.\n\n\n\n\n| \u2757 **Other model integrations** \u2757 |\n|-------------------------------------------------------------------------------------------------------------------------------------------------------------|\n| Currently, `askai` only supports integration with OpenAI. But as soon as other NLP API-endpoints start popping up, I will work on integrating them as well. |\n\n\n## Installation\n\nYou can either install it through pip\n\n```bash\npip install askai\n```\n\nor directly from the repo\n\n```\ngit clone git@github.com:maxvfischer/askai.git\ncd askai\npip install .\n```\n\n## Initialize askai\n\n`askai` needs to be initialized to set up the default config and to connect your \nOpenAI API key. Note that the key is only stored locally in `~/.askai/key`.\n\n```bash\naskai init\n```\n\n\n\n## Create OpenAI API-key\n\nFor `askai` to work, you need to add an OpenAI API-key. When creating an OpenAI account, \nthey give you $18 to use for free. After that you need to set up a paid account to \ncontinue using their service. \n\nDuring the development and testing of this CLI, I used $0.67.\n\nAn OpenAI API-key can be created by:\n\n1. Creating an account on OpenAI: https://openai.com/api/\n2. Logging in and click on `New API keys`\n3. Click `Create new secret key`\n\n## How to use\n\n\n### Simple question\nAsk a question using your saved config.\n\n```\naskai \"<QUESTION>\"\n```\n\n\n\n### Override config\nIt's possible to override the default config by using arguments:\n\n```\naskai \"<QUESTION>\" --num-answers <INT> --model <MODEL_STRING> --temperature <FLOAT> --top_p <FLOAT> --max-tokens <INT> --frequency-penalty <FLOAT> --presence_penalty <FLOAT>\n```\n\n\n| **Argument** | **Allowed values** | **Description** |\n|---------------------|----------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------|\n| --num-answers or -n | \\>0 | Number of answers to generate. Note that more answers consume more tokens |\n| --model or -m | See list below | Which model to use. See list of available models below. |\n| --temperature or -t | 0.0 <= t <= 1.0 | What sampling temperature to use. Higher value makes the model more \"creative\". Do not use at the same time as `top-p`. |\n| --top-p | 0.0 <= top_p <= 1.0 | What sampling nucleus to use. The model considers the results of the tokens with top_p probability mass. Do not use at the same time as `temperature`. |\n| --max-tokens | \\>0 | Maximum number of tokens used per question (incl. question + answer) |\n| --frequency-penalty | -2.0 <= frequency_penalty <= 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. |\n| --presence-penalty | -2.0 <= presence_penalty <= 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. |\n\n## Update config\nIf you find yourself overriding the config a lot when asking questions, you can update the default config instead.\n\n### Update all config values\n\n```bash\naskai config update all\n```\n\n### Update individual config values\n\n```bash\naskai config update num-answers\naskai config update model\naskai config update temperature\naskai config update top-p\naskai config update max-tokens\naskai config update frequency-penalty\naskai config update presence-penalty\n```\n\n### Reset to default config\n```bash\naskai config reset\n```\n\n### See current config\n```\naskai config show\n```\n\n## Update API-key\n\nIt's possible to update the API-key without re-initializing the CLI.\n\n### Overwrite current API-key\n```\naskai key add\n```\n\n### Remove current API-key\n```\naskai key remove\n```\n\n## Available models\n\nThis list was updated from OpenAI's website on 2022-12-20 and might go out of date at any time. Please\ncheck [here](https://beta.openai.com/docs/models) to see an accurate list.\n\n#### Text-generating models\n| --model | Description | Max tokens |\n|------------------|----------------------------------------------|------------|\n| text-davinci-003 | The most capable GPT3 model | 4000 |\n| text-curie-001 | Worse than davinci. Still capable and faster | 2048 |\n| text-babbage-001 | Can do straight forward tasks. Very fast | 2048 |\n| text-ada-001 | Capable of very simple tasks. Very fast | 2048 |\n\n#### Code-generating models\n| --model | Description | Max tokens |\n|------------------|-------------------------------------------|------------|\n| code-davinci-002 | Most capable code-generating model. | 8000 |\n| code-cushman-001 | Almost as capable as davinci, but faster. | 2048 |\n\n## Important notes\n\nNote that the answers generated by OpenAI and shown by `askai` is by no means a \"truth\". \nYou always need to be skeptical. And of course, don't execute generated commands or code\nwithout verifying that it does what you asked for.\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "Your simple terminal helper",
"version": "1.0.5",
"split_keywords": [],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "7f1ff3ac4f0badcb8c4a8d63943b0408b1351444967b51d2cbcb13f51026bfbf",
"md5": "8d47a276ce31ca81f59fdbbfaf29f45f",
"sha256": "91a6097f28816413d91619f558d497c89896317bd6a9c76b48ef4df6a0250542"
},
"downloads": -1,
"filename": "askai-1.0.5.tar.gz",
"has_sig": false,
"md5_digest": "8d47a276ce31ca81f59fdbbfaf29f45f",
"packagetype": "sdist",
"python_version": "source",
"requires_python": null,
"size": 12848,
"upload_time": "2023-01-18T18:56:37",
"upload_time_iso_8601": "2023-01-18T18:56:37.898425Z",
"url": "https://files.pythonhosted.org/packages/7f/1f/f3ac4f0badcb8c4a8d63943b0408b1351444967b51d2cbcb13f51026bfbf/askai-1.0.5.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2023-01-18 18:56:37",
"github": true,
"gitlab": false,
"bitbucket": false,
"github_user": "maxvfischer",
"github_project": "askai",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"lcname": "askai"
}
JSON
