ciwi

Simple, portable, single-binary CI/CD server + agent.

WIP. NOT SUITABLE FOR PUBLIC SERVERS. ONLY FOR PRIVATE NETWORKS AND HOMELAB STYLE PROJECTS.

Background

ciwi started as a practical replacement for Jenkins/TeamCity for private projects.

Quick Start

1. Install server/agent with scripts in docs/installation.md.
2. Open UI at http://127.0.0.1:8112/.
3. Import a project that contains ciwi-project.yaml.
4. Run pipeline and inspect jobs.

Detailed guide: docs/getting-started.md.

Note on agent workdir:

• CIWI_AGENT_WORKDIR is optional. If unset, the agent defaults to .ciwi-agent/work relative to its working directory.
• Installer-based deployments set an absolute workdir (Linux: /var/lib/ciwi-agent/work, macOS: $HOME/.ciwi-agent/work, Windows: %ProgramData%\\ciwi-agent\\work).
• The agent normalizes the workdir to an absolute path at runtime.

Note on artifact log verbosity:

• Agent artifact upload logs are compact by default (CIWI_ARTIFACT_LOG_LEVEL=summary).
• Use CIWI_ARTIFACT_LOG_LEVEL=verbose and optionally CIWI_ARTIFACT_LOG_MAX_INCLUDE_LINES for per-file include lines.
• Dependency artifact restore logs are also compact by default (CIWI_DEP_ARTIFACT_LOG_LEVEL=summary), with optional CIWI_DEP_ARTIFACT_LOG_MAX_RESTORED_LINES in verbose mode.
• See docs/configuration.md for details.

Documentation Map

• Getting started: docs/getting-started.md
• Installation scripts: docs/installation.md
• Env vars, prerequisites, tool requirements: docs/configuration.md
• Pipeline config and runtime model: docs/pipelines.md
• Backend API reference (grouped by consumer): docs/api.md
• Vault/AppRole integration: docs/vault.md
• Operations (update policy, maintenance, troubleshooting): docs/operations.md
• Architecture and flows: docs/architecture.md
• Domain terminology: terminology.md

Design Philosophy

ciwi intentionally avoids fragile behavior that depends on parsing human-readable logs.

• ciwi is designed around explicit API contracts and structured payloads between server, agent, and UI.
• Features should use dedicated fields/endpoints instead of scraping job output text.
• Job output remains for humans; machine behavior should rely on typed data.

Examples

Example ciwi-project.yaml files:

• ciwi-project.yaml for building/publishing cupuacu.
• ciwi-project.yaml for building ciwi itself.

Screenshots