llm-prettier

A simple, generic JSON prettier and interactive HTML viewer with a killer feature for developers working with LLM APIs: it automatically detects and parses embedded stringified JSON, rendering it as a full interactive tree rather than a raw escaped string.

Features

• 🌲 Interactive tree view — Collapsible nodes for any JSON structure
• 🔍 Search & filter — Ctrl+F search that highlights matching keys/values with Enter/Shift+Enter navigation
• 📋 Copy to clipboard — Hover any value or key to copy it; click any key to copy its full JSON path
• 🗺️ JSON path breadcrumb — Displays the full path (e.g. $.choices[0].message.content) on key click
• 📊 Stats bar — Shows total key count, max depth, and data size at a glance
• 🔢 Depth controls — Collapse to depth D2, D3, D4 in addition to expand/collapse all
• 🪡 Embedded JSON parsing — Automatically detects and renders stringified JSON, including JSON inside Markdown code blocks (json ... ) and XML tags (<tool_call> ... </tool_call>)
• 💡 Fully local — No external dependencies; everything runs in-browser from a single HTML file

Installation

Using pip:

pip install llm-prettier

Using uv (faster):

uv pip install llm-prettier

After installing, the llm-prettier CLI will be available.

Usage

CLI Tool

# Prettify a JSON file (saves as viewer_<name>.html)
llm-prettier path/to/response.json

# Open in browser immediately via local server
llm-prettier path/to/response.json --serve

# Specify custom output file
llm-prettier path/to/response.json -o my_output.html

# Read from stdin
cat response.json | llm-prettier -
curl https://api.example.com/data | llm-prettier - --serve

# Custom server port
llm-prettier response.json --serve --port 8080

# Print version
llm-prettier --version

# Run it without installing globally
uv run llm-prettier response.json --serve

Python API

import llm_prettier

my_data = {
    "model": "gemini-flash",
    "message": {
        "content": "{\"tool_call\": {\"temperature\": 0.8}}"
    }
}

# Generate HTML string
html_str = llm_prettier.render_html(my_data, title="My LLM Response")

# Save to file; returns the absolute path on success
out_path = llm_prettier.show(my_data, out_name="viewer_mydata.html")

Viewer Keyboard Shortcuts

License

MIT