Human

If you want to get the user to participate in chain logic, or to substitute in for an LLM, you can do so using the Human LLM class:

from langchain_contrib.llms.human import Human

llm = Human()
llm("Enter something: ")

Enter something:  Something

'Something'

By default, the LLM returns the very first thing the user enters. If you want multiline output from the human, you can enforce stops:

human_response = llm("Write me a Python script\n\n```\n", stop=["\n```"])
print(f"The human entered in:\n\n{human_response}")

Write me a Python script

```
 #!/usr/bin/env python3
 print('Hi')
 ```

The human entered in:

#!/usr/bin/env python3
print('Hi')

BaseHuman

The basic Human commandline input functionality is provided by the BaseHuman class. You can override its _call function to customize how the input is handled.

For example, Human itself overrides BaseHuman._call to present a menu when encountering prompts from ChoicePromptTemplate. To check this functionality out, run the following snippet in your terminal:

from langchain_contrib.llms.human import Human
from langchain_contrib.prompts import ChoicePromptTemplate

llm = Human()
choice = ChoicePromptTemplate.from_template(
    """
The colors available are: {choices}

Pick a color: """.lstrip(),
    choices=["Red", "Orange", "Blue"],
)
llm(choice.format())

(this can’t be run in the notebook due to a lack of /dev/tty)