jdx — JSON Data eXplorer

An interactive, AI-augmented terminal tool for exploring JSON data. Think jq meets a code editor — with fuzzy search, tree navigation, schema inspection, and natural language querying.

---

Install

From crates.io

cargo install jdx

From Homebrew (macOS / Linux)

brew install eladbash/jdx/jdx

Pre-built Binaries

Download from GitHub Releases — available for Linux, macOS, and Windows (x86_64 and aarch64).

Build from Source

git clone https://github.com/eladbash/jdx.git
cd jdx
cargo build --release
# Binary at: target/release/jdx

---

Quick Start

# Configure AI providers interactively
jdx init

# Pipe JSON from stdin
echo '{"name": "Alice", "age": 30}' | jdx

# Explore an API response
curl -s https://api.github.com/repos/rust-lang/rust | jdx

# Open a file directly
jdx data.json

# YAML, TOML, CSV are auto-detected
cat config.yaml | jdx

---

Features

Interactive Dot-Notation Query

Type a path like .users[0].name to drill into JSON data in real-time. Every keystroke updates the filtered view instantly.

Fuzzy Tab-Completion

Press Tab to see available keys at the current depth. Fuzzy matching ranks candidates so you can type "us" and match "users". Ghost text shows the most likely completion inline (like fish shell).

Collapsible Tree View

Press Ctrl+S for a split panel with a tree navigator on the left. Expand and collapse nodes with arrow keys. The query input and tree stay synchronized.

Schema Inspector

Press S to toggle schema view. Infers the shape of your JSON data — types, optional fields, value ranges, and array lengths — perfect for exploring unfamiliar API responses.

{
  users: [object]  # array of 3
  count: number  # 3
  next_page: string | null
}

Filter Predicates

Filter arrays inline using bracket predicates:

.store.books[price < 10]              # books cheaper than $10
.users[role == "admin"]               # admin users only
.users[age >= 30]                     # users 30 and older
.items[status != "deleted"]           # exclude deleted items
.users[active == true]                # only active users
.items[deleted == null]               # items without a deleted field

Supported operators: ==, !=, <, >, <=, >=

Values can be numbers (10, 3.5), quoted strings ("admin"), booleans (true/false), or null.

Compound filters with && (AND) and || (OR):

.items[price > 5 && price < 20]      # price in range
.users[role == "admin" || role == "mod"]  # multiple roles
.items[featured == true || price < 10 && stock > 0]  # mixed (|| has lower precedence)

Filters can be combined with path navigation and transforms:

.store.books[price < 10].title        # titles of cheap books
.store.books[price < 15] :pick title,price :sort price
.users[role == "admin"] :count        # count admin users

Inline Data Transforms

Chain transforms after your query using : commands:

Command	Description	Example
:keys	Object keys as array	.store :keys
:values	Object values as array	.store :values
:count	Count elements	.users :count
:flatten	Flatten nested arrays	.tags :flatten
:pick	Select fields	.users :pick name,email
:omit	Exclude fields	.users :omit metadata
:sort	Sort by field (asc/desc)	.users :sort age desc
:uniq	Deduplicate	.tags :uniq
:group_by	Group by field	.users :group_by role
:filter	Filter by predicate	.users :filter age > 30
:sum	Sum numeric values	.orders :sum total
:avg	Average numeric values	.scores :avg value
:min	Minimum value	.products :min price
:max	Maximum value	.products :max price
:reverse	Reverse array or string	.users :reverse
:upper	Uppercase strings	.names :upper
:lower	Lowercase strings	.names :lower
:split	Split string by delimiter	.date :split -
:join	Join array with separator	.tags :join ,

The :filter transform supports compound expressions with && and ||:

.users :filter age > 20 && age < 40
.users :filter role == "admin" || role == "moderator"

Aggregate commands (:sum, :avg, :min, :max) work on arrays of numbers or on a specific field from objects:

.store.books :filter price < 10 :pick title,price
.users :pick name,age :filter age > 25 :sort age desc
.store.books :sum price                    # total of all book prices
.store.books :filter price < 15 :avg price # average of cheap books
.scores :min                               # minimum of a numeric array

String transforms can be chained together:

.date :split - :reverse :join /            # "2024-01-15" → "15/01/2024"
.name :upper :split , :reverse :join ,     # chain multiple string ops

Natural Language AI Querying

Press / to switch to AI mode. Ask questions in plain English:

"what is the total price of all books?" "which books cost less than $10?" "who are the admin users?"

The AI answers your question directly in natural language and optionally suggests a jdx query you can apply by pressing Enter. It sees the actual data, so it can compute totals, averages, find specific items, and more. Supports OpenAI, Anthropic, and local Ollama models.

Multi-Format Input/Output

Auto-detects JSON, YAML, TOML, CSV, and NDJSON input. Output in any format with --output:

cat config.yaml | jdx                  # YAML auto-detected
cat data.csv | jdx --input csv         # CSV -> JSON
jdx data.json --output yaml            # Output as YAML

Streaming NDJSON

Pipe a streaming source and jdx launches immediately with the initial data, then updates live as new lines arrive. The status bar shows "(streaming...)" while data is still flowing.

tail -f logs.jsonl | jdx --input ndjson
docker logs -f myapp | jdx --input ndjson

Clipboard Integration

• Ctrl+Y — Copy current value to clipboard
• Ctrl+D — Bookmark the current path

Persistent History

Query history is saved across sessions. Press Ctrl+R to search through past queries.

---

Keybindings

Query Mode (default)

Key	Action
Tab	Complete / cycle candidates
Shift+Tab	Cycle candidates backward
Enter	Confirm and output result
Esc / Ctrl+C	Quit without output
Ctrl+L	Toggle key-only mode
Ctrl+U	Clear query
Ctrl+W	Delete word backward
Ctrl+A / Home	Cursor to start
Ctrl+E / End	Cursor to end
Ctrl+J	Scroll down
Ctrl+K	Scroll up
Ctrl+N	Page down
Ctrl+P	Page up
Ctrl+T	Scroll to top
Ctrl+G	Scroll to bottom
Ctrl+Y	Copy current value to clipboard
Ctrl+R	Search query history
Ctrl+D	Bookmark current path
Ctrl+S	Toggle split view (tree + JSON)
S	Toggle schema view
/	Switch to AI query mode
?	Show help overlay

Tree Mode

Key	Action
Up/Down	Navigate tree nodes
Right / Enter	Expand node
Left	Collapse node
Esc / q	Back to query mode

AI Mode

Key	Action
Type freely	Enter natural language question
Enter	Send question to AI / Apply suggested query
Esc	Back to query mode

---

Configuration

Configuration lives at ~/.config/jdx/config.toml:

[ai]
provider = "ollama"       # "ollama", "openai", "anthropic", or "none"
model = "llama3.2"        # Model name
api_key = ""              # Required for cloud providers
endpoint = ""             # Custom API endpoint (optional)

[display]
monochrome = false        # Disable colors
max_candidates = 20       # Max items in autocomplete popup
schema_max_samples = 10   # Array elements to sample for schema inference

AI Setup

Local (Ollama):

# Install Ollama: https://ollama.ai
ollama pull llama3.2
# jdx auto-connects to localhost:11434

Cloud (OpenAI):

[ai]
provider = "openai"
model = "gpt-4o-mini"
api_key = "sk-..."

---

CLI Options

Usage: jdx [OPTIONS] [COMMAND] [FILE]

Commands:
  init  Run the interactive setup wizard to configure AI providers

Arguments:
  [FILE]  File to read JSON from (reads stdin if omitted)

Options:
  -Q, --query <QUERY>     Initial query (e.g., ".users[0]")
  -q, --query-output      Output the query string instead of the result
  -i, --input <FORMAT>    Input format: json, yaml, toml, csv, ndjson
  -o, --output <FORMAT>   Output format: json, yaml, toml, csv, ndjson
  -M, --monochrome        Disable colors
  -p, --pretty            Pretty-print output (default: true)
      --non-interactive   Evaluate query and print result without TUI
  -h, --help              Print help
  -V, --version           Print version

---

Comparison

Feature	jdx	jid	jq	fx	jless
Interactive TUI	Yes	Yes	No	Yes	Yes
Fuzzy completion	Yes	Prefix only	No	No	No
Tree navigation	Yes	No	No	No	Yes
Schema inspector	Yes	No	No	No	No
AI queries	Yes	No	No	No	No
Inline transforms	Yes	No	Yes (pipe)	Yes (JS)	No
Filter predicates	Yes	No	Yes	Yes (JS)	No
Multi-format input	Yes	No	No	Yes	No
Streaming NDJSON	Yes	No	Yes	Yes	No
Clipboard copy	Yes	No	No	No	No
Query history	Yes	No	No	No	No
Single binary	Yes	Yes	Yes	Yes	Yes
Written in	Rust	Go	C	Go	Rust

---

Contributing

# Clone and build
git clone https://github.com/eladbash/jdx.git
cd jdx
cargo build

# Run tests
cargo test

# Run with clippy
cargo clippy --all-targets --all-features -- -D warnings

# Format code
cargo fmt --all

---

License

MIT