Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

15 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

CrawDevAi

CrawDevAi is the deployment, update, status, validation, and recovery toolkit for the OpenCraw-based Ai server stack.

v0.1-beta Scope

This release is limited to the foundational Dual-G/A architecture:

  • OpenCraw Primary: the normal gateway/assistant control surface.
  • OpenCraw Rescue: the backup gateway/assistant and intended recovery surface.
  • CrawDevAi install/status/validation/recovery scripts.
  • CrawDevAi staged deployment/preflight workflow.
  • systemd service templates for Primary, Rescue, and Primary recovery.
  • LiteLLM infrastructure service templates for loopback model routing behind OpenCraw.
  • configuration, policy, state, workspace, and log boundaries.

No worker services are installed in v0.1-beta. CrewAI, LangGraph, CrawCode, OpenHands, SQLite task memory, and scheduled work are documented as future integration points only. LiteLLM is added only as an infrastructure model-router service behind OpenCraw.

Referenced Projects

OpenCraw and CrawCode are branded stack components used by CrawDevAi. Public CrawDevAi installation does not require local source checkouts, GitHub repo access, upstream remotes, or repo-maintenance workflows for these projects. Upstream attribution and license obligations remain with their source projects.

Public Install Model

The public installer configures approved stack components, service templates, live configuration paths, policy templates, runtime links, logs, and recovery boundaries. It expects an approved OpenCraw runtime to be available at the runtime link before services are started; it does not require a local OpenCraw Git repo.

Default public layout:

  • CrawDevAi install link: /opt/crawdevai/current
  • OpenCraw runtime link: /opt/opencraw/current
  • OpenCraw config: /etc/opencraw/{primary,rescue}
  • OpenCraw policies: /etc/opencraw/policies
  • CrawDevAi config: /etc/crawdevai/crawdevai.env
  • LiteLLM config: /etc/litellm/config.yaml
  • LiteLLM env/secrets: /etc/litellm/litellm.env
  • OpenCraw state: /var/lib/opencraw/{primary,rescue}
  • LiteLLM state: /var/lib/litellm
  • OpenCraw workspaces: /srv/opencraw/workspace/{primary,rescue}
  • Logs: /var/log/opencraw/{primary,rescue}, /var/log/litellm, and /var/log/crawdevai

Optional U/D development-server layout may keep source repos under /srv/opencraw/repos, including /srv/opencraw/repos/CrawDevAi and /srv/opencraw/repos/OpenCraw. That source-repo workflow is not the public install default.

Commands

Public install checks and framework install:

scripts/validate.sh
scripts/preflight.sh --dry-run
scripts/install.sh --dry-run
scripts/install-litellm-runtime.sh --dry-run
scripts/install.sh --enable
scripts/status.sh
scripts/recover-primary.sh

Optional repo-based development source build:

OPENCRAW_SOURCE_DIR=/srv/opencraw/repos/OpenCraw CRAWDEVAI_OPENCRAW_SOURCE_MODE=1 scripts/install.sh --build --enable

scripts/install.sh writes system files and therefore must be run as root during live deployment. The default install does not start services unless --start is provided. --build is optional development behavior and requires OPENCRAW_SOURCE_DIR; public installs should use an approved OpenCraw runtime at /opt/opencraw/current instead of requiring source-repo metadata.

Service Units

  • opencraw-primary.service
  • opencraw-rescue.service
  • litellm.service
  • crawdevai-primary-recovery.service
  • crawdevai-primary-recovery.timer

Primary and Rescue both use systemd restart policy. Rescue is the intended recovery controller for normal repair work. The CrawDevAi recovery timer is an optional root-level systemd watchdog fallback, disabled by default, that keeps Rescue available, checks Primary health, and restarts Primary only through systemd when emergency fallback recovery is needed.

LiteLLM is a loopback-only infrastructure service. It is installed inactive and disabled until the runtime, LITELLM_MASTER_KEY, and any deployment-specific provider credentials needed for enabled aliases are configured.

Documentation

About

Linux Native Ai Inference Gateway/Assistant Interface and Agentic/Orchestration Stack Installation

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages