JSON THAT

Installation

pip install jsonthat

Start by setting up your LLM providers and enter your API keys when prompted:

jt --setup

Display usage

jt -h

Usage

Here are various ways to use the JSON THAT CLI tool:

Basic Usage

Process text into JSON using the default provider:

$ echo "My name is Jay and I'm 30 years old" | jt

{
  "name": "Jay",
  "age": 30
}

Using a plain TEXT Schema

Guide the output using a JSON schema:

$ cat person_schema.txt

  name: string
  occupation: string
  yearsOfExperience: number
  

$ echo "Alice is a software engineer with 5 years of experience" | jt --schema person_schema.txt

{
  "name": "Alice",
  "occupation": "software engineer",
  "yearsOfExperience": 5
}

Using a JSON Schema

Guide the output using a JSON schema:

$ cat person_schema.json

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "type": "object",
  "properties": {
    "name": {
      "type": "string"
    },
    "occupation": {
      "type": "string"
    },
    "yearsOfExperience": {
      "type": "number"
    }
  },
  "required": ["name", "occupation", "yearsOfExperience"]
}

$ echo "Alice is a software engineer with 5 years of experience" | jt --schema person_schema.json

{
  "name": "Alice",
  "occupation": "software engineer",
  "yearsOfExperience": 5
}

Processing a File

Convert the contents of a file to JSON:

$ cat input.txt | jt

{
  "content": "This is the content of input.txt converted to JSON format."
}

Using a Specific Provider

Specify a provider for a single command:

$ echo "Convert this sentence into JSON" | jt --provider openai

{
  "action": "Convert",
  "subject": "sentence",
  "target_format": "JSON"
}

Combining Options

Use multiple options together:

$ echo "Transform this using Ollama and a custom schema" | jt --provider ollama --schema custom_schema.json

{
  "transformed_text": "This has been transformed using Ollama and a custom schema",
  "provider_used": "ollama",
  "schema_applied": true
}

Stream output

Stream the output of the LLM in real-time:

$ echo "Explain the process of photosynthesis" | jt --stream

This will display the JSON output as it's being generated, allowing you to see the results in real-time.

Features

• Supports json schema to guide outputs
• Pipe content directly
• Multi-LLM provider support: OpenAI, Claude, Mistral, Ollama
• Local LLM support with Ollama
• Store multiple provider configurations
• Set a default provider
• Choose provider via command-line flag
• Set model name per provider
• Stream output in real-time

Configuration

Config file

Configurations are managed through:

~/.config/jsonthat/config.yaml

Command to display configuration:

$ jt --config

The config file stores multiple provider configurations and the default provider.

default_provider: openai
providers:
  claude:
    api_key: *****
  mistral:
    api_key: *****
    model: open-mistral-nemo
  ollama:
    api_url: http://localhost:11434
    model: llama3.1
  openai:
    api_key: *****
    model: gpt-4o-mini

Environment Variables

To configure the CLI tool, you can set the following environment variables:

export LLM_PROVIDER='openai'
export LLM_API_KEY='your_api_key_here'
export LLM_MODEL='gpt-4o-mini'
export OLLAMA_API_URL='http://127.0.0.1:11434'
export OLLAMA_MODEL='llama3'

Specifies the large language model provider (e.g., 'openai', 'claude', 'mistral' or 'ollama'), API key for the specified LLM provider, and Ollama-specific configurations.

Command-line Provider Selection

You can specify the provider to use for a single command:

echo "Your sample text" | jt --provider ollama

Feedback

Share your use case, something to improve or a question, join the discord: