ci-copilot

README

CI Copilot

Your AI-powered assistant for GitLab CI/CD pipelines.

CI Copilot helps you analyze, fix, and understand your .gitlab-ci.yml files using large language models (LLMs).

It works both as a command-line tool and as a local HTTP API.

🚀 Features

CI Copilot supports 4 powerful AI-driven modes:

---

• 🔍 Suggest
  Analyze your pipeline and get clear recommendations to improve reliability, structure, or efficiency.

• 🛠 Fix
  Automatically fix common issues in your .gitlab-ci.yml file. You can preview or apply the changes.

• 🔐 Audit
  Perform a focused audit for security, speed, or cleanliness. Get prioritized recommendations.

• 📘 Explain
  Understand what your pipeline is doing. The AI explains your .gitlab-ci.yml in simple language.

⚙️ Usage (CLI + API)

CI Copilot can be used as a command-line tool or as a local API.

---

🖥 CLI usage

# Get suggestions for improvements
ci-copilot suggest --in .gitlab-ci.yml

# Automatically fix and save to a new file
ci-copilot fix --in .gitlab-ci.yml --out .gitlab-ci.fixed.yml

# Audit the pipeline for security issues
ci-copilot audit --in .gitlab-ci.yml --focus security

# Get a simple explanation of the pipeline
ci-copilot explain --in .gitlab-ci.yml

🧩 API usage

You can use CI Copilot as a local HTTP API by running the Go server.

Start the server:

go run ./cmd/server/main.go

By default, the API runs at http://localhost:8080

POST /suggest
POST /fix
POST /audit
POST /explain

Example request (using curl)

Request:

curl -X POST http://localhost:8080/suggest \
  -H "Content-Type: application/json" \
  -d '{"message": "your .gitlab-ci.yml content here"}'

Response:

{
  "summary": "The pipeline is mostly correct but can be improved.",
  "recommendations": [
    {
      "issue": "Job 'build' is missing retry policy",
      "suggestion": "Add 'retry: 2' to ensure resilience",
      "priority": "medium"
    }
  ]
}

🛠 Installation

1. Clone the repository

git clone https://github.com/genov8/ci-copilot.git
cd ci-copilot

2. Alternatively, if you only need a single CLI command:

You can also install individual CLI tools:

go install github.com/genov8/ci-copilot/cmd/suggest@latest
go install github.com/genov8/ci-copilot/cmd/fix@latest
go install github.com/genov8/ci-copilot/cmd/audit@latest
go install github.com/genov8/ci-copilot/cmd/explain@latest

🧪 Tests

Basic API behavior is covered with integration tests.

Since core functionality relies on AI model responses, we recommend manual review for output accuracy.

Run all tests with:

go test ./...

📁 Project Structure

The project is organized into clearly separated modules:

ci-copilot/
├── cmd/ # CLI entry points (suggest, fix, audit, explain)
├── internal/
│ ├── client/ # Main logic for interacting with the AI service
│ ├── config/ # Centralized configuration (e.g. endpoint URLs)
│ ├── server/ # HTTP API server for local AI access
│ └── ui/ # Optional pretty output for CLI users
├── prompts/ # All prompt templates used by AI interactions
├── tests/ # Test cases (optional)
└── ai-service/ # Optional local AI service using Gemini via FastAPI

🧠 How It Works

CI Copilot combines CLI tools written in Go with a local AI-powered analysis service.

Here’s the flow:

1. The CLI reads your .gitlab-ci.yml file.
2. It loads a specific prompt (stored in the prompts/ folder).
3. The content and prompt are combined and sent via HTTP to a local FastAPI AI service.
4. The AI model (e.g., Gemini) generates a structured response.
5. The CLI tool formats and prints the result — or applies changes if needed.

The architecture is modular:

• Prompts are decoupled and easy to edit.
• You can replace the AI backend without changing the CLI logic.

💡 Potential Extensions

This project is designed to be extensible. Below are directions where CI Copilot could evolve:_

• 🧠 Custom LLMs for CI/CD
  Train or fine-tune your own language models that specialize in DevOps and .gitlab-ci.yml pipelines.

• 🔁 Pluggable AI Backends
  Support multiple AI providers — OpenAI, Gemini, or local LLMs — by switching via simple configuration.

• 🧪 GitLab-native validation
  Add automatic .gitlab-ci.yml checks during push via GitLab Webhooks or server-side hooks — no changes to pipeline files needed.

• 🛠 Custom rule engines
  Allow teams to define internal rules or style guides that the assistant must enforce.

• 🧩 IDE plugins
  Integrate with VS Code or JetBrains IDEs to provide in-editor suggestions and validation.

• 🌐 Web dashboard
  Visualize suggestions, track history, and monitor best practice adoption across projects.

• 🛡 Enterprise mode
  Add support for self-hosted models, offline operation, and team-level controls in secure environments.

📜 License

This project is licensed under the MIT License.