cchain

Overview

cchain is a modern CLI automation tool built in Rust that allows you to chain commands together in a structured, retry-aware workflow. Define your automation pipelines in JSON, and let cchain handle execution, error recovery, and dynamic input generation using language models (LLMs). Whether you're automating local development tasks, CI/CD pipelines, or complex workflows, cchain is designed to replace brittle shell scripts with a declarative, developer-friendly approach.

Features

• Command Chaining: Execute a series of commands with configurable arguments, retries, and environment variables.
• Retry Logic: Automatically retry failed commands with a specified number of attempts.
• Dynamic Input Generation: Use LLMs to generate command inputs dynamically (e.g., AI-generated commit messages).
• Environment Variable Management: Override environment variables per-command and pass outputs between steps.
• Interpreter Agnostic: Run commands in sh, bash, or any interpreter of your choice.
• Bookmarking: Save frequently used command chains for quick access.
• Lightweight & Fast: Built in Rust for performance and reliability—no startup lag or dependency hell.

Why cchain?

• Replace Bash Scripts: Stop debugging flaky shell scripts. Define workflows in JSON for version control, reusability, and auditability.
• AI-Powered Automation: Integrate LLMs into your workflows—generate commit messages, summarize logs, or categorize outputs on the fly.
• Local & CI Consistency: Test pipeline steps locally before pushing to CI. Ensure identical behavior across environments.
• Cross-Platform: Single binary deployment works on Linux, macOS, and Windows.

Comparison to Alternatives

Installation

Cargo

Install cchain using Cargo:

cargo install cchain

Building from Source

Clone the repository and build it using Cargo:

git clone https://github.com/aspadax/cchain.git
cd cchain
cargo build --release

Usage

Basic Example

Create a JSON configuration file to define your command chain. For example:

[
    {
        "command": "echo",
        "arguments": ["Hello, world!"],
        "retry": 3
    },
    {
        "command": "ls",
        "arguments": ["-la"],
        "retry": 1
    }
]

You may also run this command to generate a template file to work with:

cchain new your_file_name

The generated file will be automatically named as: cchain_your_file_name.json. Notice that cchain will only recognize files that has a cchain_ prefix along with a .json extension.

Run cchain with a chain file:

cchain run path/to/cchain_your_file.json

Failure Handling

cchain not only supports retry but also additional steps to take when a command line fails to execute. For example, if the LLM call failed when trying to create a commit message, and yo u would like to git reset and then perform a new commit after fixing the LLM issue. Then you probably will need this:

[
  {
    "command": "git",
    "arguments": ["add", "--all"],
    "retry": 0
  },
  {
    "command": "git",
    "arguments": ["commit", "-m", "llm_generate('based on the below git diff output, summarize the changes in a few sentences.', 'git --no-pager diff --staged')"],
    "failure_handling_options": {
      "exit_on_failure": true,
      "remedy_command_line": {
        "command": "git",
        "arguments": ["reset"]
      }
    },
    "retry": 0
  },
  {
    "command": "git",
    "arguments": ["push"],
    "retry": 0
  }
]

In the second program of this chain, the key failure_handling_options provides two options when this program fails to execute. The exit_on_failure specifies whether the chain needs to exit entirely when this program fails. The remedy_command_line specifies a command to run when the program fails.

In this case, we need cchain to execute git reset if the commit fails. And then, we will need to chain to stop running the subsequential programs to prevent a pointless push.

Environment Variables

Override environment variables per-command and capture command outputs for reuse:

[
    {
        "command": "echo",
        "arguments": ["$<<env_var_name>>"],
        "environment_variables_override": {
            "hello": "world"
        },
        "stdout_stored_to": "<<env_var_output>>",
        "interpreter": "sh",
        "retry": 0
    },
    {
        "command": "echo",
        "arguments": ["$<<env_var_output>>"],
        "retry": 0
    }
]

AI-Powered Workflows

Use LLMs to generate dynamic inputs. For example, generate a commit message based on git diff:

[
    {
        "command": "git",
        "arguments": ["commit", "-m", "llm_generate('generate a commit message', 'git --no-pager diff')"],
        "retry": 1
    }
]

Configure your LLM by setting these environment variables:

export CCHAIN_OPENAI_API_BASE="http://localhost:11434/v1"
export CCHAIN_OPENAI_API_KEY="test_api_key"
export CCHAIN_OPENAI_MODEL="mistral"

Bookmarking

Bookmark frequently used command chains for quick access:

cchain add path/to/cchain_your_file.json

Or, you may want to add a directory of cchain files

cchain add path/to/a/directory

Look up existing bookmarked chains and their indexes

cchain list

Run a bookmarked chain by index:

cchain run 0

Delete a bookmark by index:

cchain remove 0

Reset the entire bookmark if needed:

cchain remove --reset

Syntax Check

You may want to make sure that your chain won't crash unexpected. The syntax check offers a mechanism that looks over your chain and then pick up potential issues. This is still an experimental feature, and may not be accurate.

You can check your chain bu running:

cchain check ./path/to/your/chain.json

or, simply pass in an index of a bookmarked chain

cchain check 0

Example Use Cases

• Git Automation: Automate git add, git commit (with AI-generated messages), and git push in one workflow.
• CI/CD Prep: Run linters, build artifacts, and generate changelogs locally before pushing to CI.
• Onboarding: Set up dev environments with a single command—clone repos, install dependencies, and configure tools.
• AI-Augmented Debugging: Use LLMs to analyze logs, categorize errors, or suggest fixes.

Contributing

Contributions are welcome! Check out the Contributing Guidelines to get started.

License

This project is licensed under the MIT License. See the LICENSE file for details.