jtt

Pretty-print JSON chat logs as left/right terminal bubbles with a built-in pager.

What is this?

jtt is a terminal viewer for JSON conversation files. It renders each message as a Unicode-bordered bubble — user messages float to the right, assistant messages to the left, system messages span the full width — and opens the result in a keyboard-driven, full-screen pager so you can scroll through long conversations without losing context.

It pairs naturally with raw-llm, which stores every conversation as a plain JSON file in exactly the format jtt expects.

Installation

From crates.io

cargo install jtt

From source

git clone https://github.com/rodolfovillaruz/jtt.git
cd jtt
cargo install --path .

Usage

jtt path/to/conversation.json

jtt reads the file, wraps every message to 75 % of your terminal width, and opens the built-in pager. When stdout is redirected to a pipe or file the pager is skipped and the rendered text is written directly.

# Pipe the rendered output to a file
jtt conversation.json > conversation.txt

Pager key bindings

Key	Action
q / Q / Ctrl-C	Quit
↓ / j / Enter	One line down
↑ / k	One line up
PgDn / Space / f	One page down
PgUp / b	One page up
g / Home	Jump to top
G / End	Jump to bottom

Conversation format

jtt reads JSON files that are either a single message object or an array of message objects. Each object needs a role and a content field:

[
  { "role": "system",    "content": "You are a helpful assistant." },
  { "role": "user",      "content": "What is context engineering?" },
  { "role": "assistant", "content": "Context engineering is the practice of …" }
]

role	Bubble position
user	Right (75 % width)
assistant	Left (75 % width)
system	Full terminal width
anything else	Left (75 % width)

Using with raw-llm

raw-llm saves every conversation as a JSON array in the same format jtt expects. The two tools compose naturally over a shared file.

Start a conversation, then view it

# Start a new conversation with Claude
echo "Explain monads in one paragraph" | claude .prompt/monads.json

# View the conversation as chat bubbles
jtt .prompt/monads.json

Continue a conversation, then view it

# Add another turn to an existing conversation
echo "Give me a concrete Haskell example" | claude .prompt/monads.json

# Review the full exchange
jtt .prompt/monads.json

Pipe the rendered output

# Render to plain text for sharing or archiving
jtt .prompt/monads.json > monads.txt

Typical workflow

# 1 · Ask a question
echo "Refactor this function" | cat main.rs - | claude .prompt/refactor.json

# 2 · Review the answer in the pager
jtt .prompt/refactor.json

# 3 · Edit the conversation file to steer context, then continue
$EDITOR .prompt/refactor.json
echo "Now add error handling" | claude .prompt/refactor.json

# 4 · Review the updated exchange
jtt .prompt/refactor.json

License

MIT