fix-engine

Generic fix engine for applying pattern-based and LLM-assisted code migration fixes.

Takes Konveyor analysis output and applies automated fixes to your codebase — deterministic pattern-based renames, attribute removals, import rewrites, and LLM-assisted structural transformations.

Quick Start

# Build
cargo build --release

# Apply fixes (default behavior — writes to disk)
fix-engine fix ./my-project -i analysis.yaml

# Preview changes as a diff without writing
fix-engine fix ./my-project -i analysis.yaml --dry-run

# With fix strategies
fix-engine fix ./my-project -i analysis.yaml \
  --strategies rules/fix-strategies.json \
  --strategies output/fix-strategies.json

Installation

cargo install --path .

Cross-compilation

See CROSS-COMPILATION.md for building Linux binaries on macOS.

Usage

Basic

# Apply all fixes to the project
fix-engine fix ./project -i konveyor-output.yaml

# Preview what would change (unified diff)
fix-engine fix ./project -i konveyor-output.yaml --dry-run

With LLM-assisted fixes

# Use local goose CLI for complex fixes
fix-engine fix ./project -i output.yaml --llm-provider goose

# Use a remote OpenAI-compatible endpoint
fix-engine fix ./project -i output.yaml \
  --llm-provider openai \
  --llm-endpoint https://api.openai.com/v1/chat/completions

With fix strategies

Strategy files are JSON maps of rule ID to fix strategy. They can come from two sources:

• Hand-written — bundled alongside rule YAML files
• Auto-generated — produced by semver-analyzer's konveyor command

Pass multiple files with --strategies; later files override earlier ones on key conflicts:

fix-engine fix ./project -i output.yaml \
  --strategies rules/fix-strategies.json \
  --strategies output/fix-strategies.json

CI/CD integration

# JSON output for machine parsing
fix-engine fix ./project -i output.yaml --dry-run --output-format json

# Quiet mode (errors + summary only)
fix-engine fix ./project -i output.yaml -q

# Disable colors for log files
fix-engine fix ./project -i output.yaml --color never

Shell completions

# Generate and install (zsh example)
fix-engine completions zsh > ~/.zsh/completions/_fix-engine

# Other shells: bash, fish, elvish, powershell
fix-engine completions bash > /etc/bash_completion.d/fix-engine

CLI Reference

fix-engine fix [OPTIONS] --input <INPUT> <PROJECT>

Flag	Short	Description
<PROJECT>		Path to the project to fix (positional, required)
--input	-i	Path to Konveyor analysis output — YAML or JSON (required)
--dry-run		Preview changes as a unified diff without writing to disk
--llm-provider		LLM provider: goose (local CLI) or openai (remote endpoint)
--llm-endpoint		LLM endpoint URL (required when --llm-provider=openai)
--strategies		Fix strategies JSON file(s) — can be repeated, later overrides earlier
--log-dir		Directory to save goose prompts/responses for debugging
--verbose	-v	Show detailed output
--quiet	-q	Suppress progress, show only errors and summary
--output-format		Output format: text (default) or json
--color		Color mode: auto (default), always, or never

Environment variables

Variable	Description
RUST_LOG	Control log verbosity (e.g., RUST_LOG=debug). Default: info. Log output is routed through the progress display so it won't clobber active spinners.

Architecture

The project is a Cargo workspace with four crates:

Crate	Description
fix-engine-cli	CLI binary — argument parsing, progress reporting, orchestration
fix-engine-core	Core types — text edits, fix plans, strategies, fix result
fix-engine	Engine — pattern-based planning/applying, goose/LLM clients
fix-engine-js-fix	JS/TS language provider — prop removal, import dedup, lockfile management

Fix pipeline

1. Load — Parse Konveyor analysis output (YAML or JSON)
2. Plan — Map each violation to a fix strategy (pattern, LLM, or manual)
3. Apply patterns — Deterministic renames, removals, import rewrites
4. Apply LLM fixes — Send complex structural changes to goose or OpenAI
5. Report — Show results, manual review items, and summary

License

Apache-2.0