Command line tool to ask AI for help.
Askai is a tool with the command line interface to ask questions to AI provided by OpenAI and Cohere via REST API. It can read the user's prompt from its command line parameter or/and from stdin. The answer is printed to stdout. The question can be directed to one AI provider (OpenAI or Cohere) or both providers simultaneously. It is capable of processing of text inputs that are larger than maximum number of tokens supported by the given model. This is done by splitting the prompt into multiple text segments and making summary for each of them.
Prerequisites:
- go 1.20 or later
- make
Just run make command in the root directory of the project.
ilia:~/Projects/askai$ make
go build -o bin/askai
The binary can be found in bin directory.
To install the binary:
ilia:~/Projects/askai$ make install
Getting help.
ilia:~/Projects/askai/bin$ ./askai --help
Usage of ./askai:
-b Batch mode, do not ask for prompt if stdin is empty
-e string
AI engine to use (default "cohere")
-ea
Use all supported AI engines
-nostdin
Skip reading prompt from stdin
-p string
Prompt to AI
-pe
Print engine name in output
-pp
Print prompt in output
Asking a question.
ilia:~/Projects/askai/bin$ ./askai "What does Cohere mean?"
Cohere is an adjective that means sticking together firmly as parts.
If the prompt is empty, it will be requested from stdin.
ilia:~/Projects/askai/bin$ ./askai
Enter prompt to AI:
How are you?
I'm doing well, thank you. How about you?
You can apply some prompt to the text redirected to stdin.
ilia:~/Projects/askai/bin$ man ls | ./askai "Summary"
ls is a command line tool for displaying information about files and directories in the Linux operating system. It can display information such as file size, modification time, and file type. It can also be used to sort files by size, time, or name.
If you want to use command line parameters, you must use -p parameter for the prompt.
ilia:~/Projects/askai/bin$ ./askai -p "What do you think about OpenAI's models?" -e cohere
I think OpenAI's models are impressive. They have achieved human-level performance in a variety of tasks, including language translation and summarization, and they have the potential to revolutionize many industries. However, I also think that there are potential risks associated with these models. For example, they could be used to automate harmful tasks, such as warfare or mass surveillance. It is important to carefully consider the potential consequences of these models and to ensure that they are used in a responsible and ethical manner.
If you have installed the binary using "make install" then you can run askai from any directory.
ilia:~$ askai "Who am I?"
You are a unique individual with your own thoughts, feelings, and experiences. You are a person with your own goals, interests, and beliefs. You are a member of a larger community and society, and your actions and words have an impact on those around you.
The program's configuraion is stored in user's home directory: ~/.askai/config/askai.json
Example of configuration:
{
"apikeys": {
"cohere": "",
"openai": ""
},
"engine": "cohere",
"summarizeprompt": "Summarize:",
"providermodel": {
"cohere": "command-xlarge-nightly",
"openai": "gpt-3.5-turbo"
},
"printaiengine": "#%s#",
"loglevel": "trace",
"logdir": "~/.askai/log",
"logformat": ""
}- section "apikeys" contains API keys for Cohere and OpenAI. You can fill this information in configuration file or it will be asked on the first run.
- parameter "engine" is used to specify the default engine to use (openai or cohere).
- parameter "summarizeprompt" is used to specify the prompt to summarize the text input.
- section "providermodel" is used to specify the default provider model to use for each AI provider.
- parameter "printaiengine" is used to specify print template to print AI engine name in output.
- parameter "loglevel" is used to specify the default log level. It can be trace, debug, info, warn, error, fatal.
- parameter "logdir" is used to specify the default log directory.
- parameter "logformat" is used to specify the default log format.
The project is distributed under the terms of the MIT license.