AnyRepair

A Rust crate for repairing malformed structured data across 10 text formats (JSON, YAML, Markdown, XML, TOML, CSV, INI, Diff, Java properties, and .env) and 6 binary media formats (PNG, JPEG, WAV, MP3, MP4, MOV).

Quick Start

Installation

[dependencies]
anyrepair = "0.3.0"

Basic Usage

use anyrepair::repair;

// Auto-detect format and repair
let malformed = r#"{"name": "John", age: 30,}"#;
let repaired = repair(malformed)?;
println!("{}", repaired); // {"name": "John", "age": 30}

CLI

# Install
cargo install anyrepair

# Auto-detect and repair
anyrepair repair input.json

# Format-specific repair
anyrepair repair input.json --format json
anyrepair repair input.yaml --format yaml
anyrepair repair config.ini --format ini
anyrepair repair app.properties --format properties
anyrepair repair .env --format env

# Show confidence score
anyrepair repair input.json --format json --confidence

# Preview changes as a diff (no output written)
anyrepair repair input.json --format json --diff --dry-run

# Dry-run without writing output
anyrepair repair input.json --format json --dry-run

# Machine-readable JSON output for CI
anyrepair repair input.json --format json --json --dry-run

# Require minimum confidence (exit with error if below threshold)
anyrepair repair input.json --format json --min-confidence 0.8

# Show which repair strategies were applied
anyrepair repair input.json --format json --explain --dry-run

# Generate shell completions
anyrepair completions bash > /etc/bash_completion.d/anyrepair
anyrepair completions zsh > _anyrepair
anyrepair completions fish > ~/.config/fish/completions/anyrepair.fish

# Batch process multiple files
anyrepair batch --input ./data --output ./repaired --recursive

# Stream large files
anyrepair stream --input large_file.json --output repaired.json --format json

# Repair binary/media files
anyrepair repair-binary damaged.png --output fixed.png
anyrepair repair-binary video.mp4 --format mp4 --dry-run
anyrepair repair-binary corrupted.wav --output fixed.wav

# Validation without repair
anyrepair validate --input input.json --format json

Docker

docker build -t anyrepair .
docker run --rm -i anyrepair repair --format json <<'EOF'
{"name": "John", age: 30,}
EOF

# MCP server
docker run --rm -i anyrepair anyrepair-mcp

What's New

v0.3.0 — Current

• Media file repair — PNG, JPEG, WAV, MP3, MP4, MOV binary repair via BinaryRepair trait
• Binary format detection — magic bytes auto-detection for 6 media formats with confidence scores
• CLI repair-binary subcommand — anyrepair repair-binary <file> [--format png] [--dry-run]
• API: repair_binary(), repair_binary_with_format(), create_binary_repairer(), detect_binary_format()
• 513 tests, all passing

v0.2.9

• detect_format_with_confidence — format detection returns a confidence score (DetectionResult)
• Docker image — multi-stage Dockerfile for CLI and MCP binaries
• MCP version — anyrepair-mcp reports CARGO_PKG_VERSION (no more stale hardcoded version)
• Integration tests for smart quotes, boolean variants, and prose extraction
• 450 tests, all passing

v0.2.8

• CLI: --diff, --dry-run, --json, --min-confidence, --explain, --color
• Shell completions: anyrepair completions <shell>
• LLM JSON strategies: smart quotes, boolean variants (yes/no/on/off), prose/preamble extraction
• Optional strict feature for serde_json-backed JSON validation
• Golden master + properties/env integration tests

v0.2.7

• Auto-detect properties & env — detect_format() now recognizes .properties and .env files
• Heuristic validator fixes — XML / CSV edge cases
• Criterion benchmarks — All 10 formats + format detection + large-document throughput

v0.2.6

• Minimal dependencies and json_util without serde
• Heuristic validators for XML, TOML, CSV, and YAML

v0.2.0–0.2.5 (highlights)

• Properties/.env formats, centralized registry, Python-compatible jsonrepair(), MCP, streaming

See CHANGELOG.md for full version history.

Why AnyRepair?

Structured data from LLMs, APIs, or manual editing is often malformed. AnyRepair fixes common issues:

• JSON: Missing quotes, trailing commas, syntax errors
• YAML: Indentation, missing colons
• Markdown: Headers, links, fences
• XML / TOML / CSV / INI / Diff: Format-specific repairs
• Properties / .env: Key=value lines, sections, escaping

Key features:

• Auto-detects format for all 10 text formats + 6 binary media formats
• Deterministic heuristic repairs (no network, no ML)
• Small dependency footprint (four runtime crates)
• MCP server for Claude and other MCP clients
• Streaming for large files
• Binary media repair (PNG, JPEG, WAV, MP3, MP4, MOV)
• Python-compatible JSON API
• Optional strict feature for full serde_json parser validation
• 578 tests (cargo test)

Dependencies

Parsing and validation for JSON, XML, TOML, CSV, and YAML use in-crate heuristics and json_util rather than heavyweight parser dependencies. Enable the strict feature for serde_json-backed JSON validation:

cargo build --features strict
cargo test --features strict

Usage Examples

Multi-Format Auto-Detection

use anyrepair::repair;

let json = repair(r#"{"key": value,}"#)?;
let yaml = repair("name: John\nage: 30")?;
let markdown = repair("# Header\n[link](https://github.com/yingkitw/anyrepair/blob/main/url")?;

Explicit key-value formats

use anyrepair::repair_with_format;

let props = repair_with_format("db.url jdbc:postgresql://localhost", "properties")?;
let env = repair_with_format("API_KEY = secret", "env")?;

Python-Compatible JSON API

use anyrepair::{jsonrepair, JsonRepair};

let repaired = jsonrepair(r#"{"name": "John", age: 30,}"#)?;

let mut jr = JsonRepair::new();
let repaired = jr.jsonrepair(r#"{name: "John"}"#)?;

Format-Specific Repairers

use anyrepair::{create_repairer, repair_with_format, traits::Repair};

let mut repairer = create_repairer("json")?;
let repaired = repairer.repair(malformed_json)?;
let confidence = repairer.confidence(&repaired);

let repaired = repair_with_format(malformed_yaml, "yaml")?;

Streaming Large Files

use anyrepair::StreamingRepair;
use std::fs::File;
use std::io::BufReader;

let input = BufReader::new(File::open("large_file.json")?);
let mut output = File::create("repaired.json")?;

let processor = StreamingRepair::with_buffer_size(65536);
processor.process(input, &mut output, Some("json"))?;

Binary Media Repair

use anyrepair::repair_binary;

// Auto-detect and repair
let repaired = repair_binary(&damaged_png_bytes)?;

// Explicit format
use anyrepair::repair_binary_with_format;
let repaired = repair_binary_with_format(&damaged_mp4_bytes, "mp4")?;

// Using the BinaryRepair trait
use anyrepair::{create_binary_repairer, BinaryRepair};
let mut repairer = create_binary_repairer("png")?;
let repaired = repairer.repair(&damaged_png_bytes)?;
let confidence = repairer.confidence(&damaged_png_bytes);

MCP Server Integration

cargo install anyrepair
anyrepair-mcp

claude_desktop_config.json:

{
  "mcpServers": {
    "anyrepair": {
      "command": "anyrepair-mcp"
    }
  }
}

Tools: repair, repair_json, repair_yaml, repair_markdown, repair_xml, repair_toml, repair_csv, repair_ini, repair_diff, repair_properties, repair_env, validate.

See MCP_SERVER.md for setup details.

Supported Formats

Performance

• Strategy pipeline with priority ordering
• Release profile optimized for size (opt-level = "z", LTO, strip)
• Streaming for files larger than RAM
• Fewer transitive dependencies for faster builds and smaller binaries

cargo build --release
cargo build --profile dist   # distribution profile in Cargo.toml

Testing

cargo test

See TEST_SUMMARY.md for more detail.

Comparison

Documentation

Links: GitHub Issues · crates.io · docs.rs

Examples

cargo run --example mcp_repair_json
cargo run --example mcp_multi_format

See examples/ and examples/README.md.

Roadmap

See TODO.md for planned work (WebP, FLAC, WebM, Protobuf, CLI polish, web/API, format-preserving repairs, and more).

License

Apache-2.0

Repository

If AnyRepair helps you, consider starring the repo on GitHub — it helps others find the project.

https://github.com/yingkitw/anyrepair