Setup
WorldFork has two setup paths:
Agent-guided setup, which is the preferred onboarding path.
Complete manual setup, for operators who want every command.
Agent-Guided Setup
Paste this prompt into your agent:
Run this command to install the WorldFork skill, then use it to set up WorldFork on this computer:
npx skills add Hilo-Hilo/WorldFork/skills/worldfork --all
After installing it, use the setup module in the skill to guide me through prerequisites,
.env configuration with OPENROUTER_API_KEY, CLI installation, Docker Compose
startup, migrations, seeding, readiness verification, a short explanation of
Big Bangs, multiverses, ticks, branches, endpoint ledgers, and reports, and the
onboarding demo if I confirm I want to spend live API credits. Be welcoming and
proactive: explain what each setup phase proves, use `worldfork setup` to show
provider options, ask which providers I want to configure, recommend the
fast-cheap cohort plus stronger governance/report split for Atlas, and narrate
what Atlas is doing while it runs.
Use the configured split for live API-credit runs: OpenRouter `deepseek/deepseek-v4-flash` for fast cohort/hero/action/event-summary/single-report work, and OpenAI Codex `gpt-5.4` for initializer, God-review, endpoint-ledger, report fallback, and final multiverse report synthesis.
Complete Manual Setup
Use this path when you want to run each command yourself.
Prerequisites
Docker Desktop or another Docker Compose runtime
Python 3.11 or newer
uvNode.js 20 or newer for
npx skillsAn OpenRouter API key
Configure Environment
Copy the example environment:
cp .env.example .env
Set OPENROUTER_API_KEY in .env.
WorldFork can route cohort, hero, action, event-summary, predicate, and single-report work to deepseek/deepseek-v4-flash through OpenRouter. Initializer, God-review, endpoint-ledger, report fallback, and final multiverse report routes use OpenAI Codex gpt-5.4 by default unless you intentionally choose another configured strong model.
Install The CLI
Install the CLI package from the monorepo:
python3.11 -m pip install -e ./cli
worldfork --help
If worldfork resolves to an old global shim, reinstall the editable CLI and
check which -a worldfork. While repairing a shim, use the source-checkout
fallback from cli/:
uv run --extra dev worldfork --help
The CLI chooses its backend target in this order:
--base-urlWORLD_FORK_API_BASEBACKEND_API_BASEhttp://127.0.0.1:8003
After the CLI is installed, configure the strong governance/report route. For OpenAI Codex OAuth:
worldfork settings openai-codex-login
This writes the default auth file under ~/.worldfork/. Operators may instead
set OPENAI_CODEX_OAUTH_TOKEN or OPENAI_CODEX_AUTH_FILE in .env.
Start The Stack
make build
make up
make migrate
make seed
Use make logs if the API or worker fails to start.
If startup or readiness fails, check Docker Desktop first, then check for host
port conflicts on 8003, 5433, and 6379.
Update Later
Use the CLI updater for normal code refreshes:
npx skills add Hilo-Hilo/WorldFork/skills/worldfork --all
worldfork update --dry-run
worldfork update --yes
Refresh the worldfork skill first so the agent has the latest update runbook.
The updater only fast-forwards the source checkout. It does not overwrite .env, local Docker overrides, run folders, artifacts, or database state, and it does not run migrations unless a future command explicitly adds that behavior.
Verify Readiness
worldfork status
worldfork query GET /readyz --no-api-prefix
Readiness should show the database and Redis checks as healthy. OpenRouter is healthy when OPENROUTER_API_KEY is set, and OpenAI Codex is healthy when OAuth auth is present if that provider is part of the configured route policy.
Confirm the effective model and provider routes before spending live API credits:
worldfork setup
worldfork settings llm
worldfork setup is the compact provider-choice helper for agents and first-time
operators. It lists available provider options, shows the recommended Atlas
profile, and tells the agent what to ask before configuring API keys or model
routes. Use worldfork setup --include-patch only when the agent needs the full
Atlas routing PATCH payload. Local OpenAI-compatible providers such as Ollama,
vLLM, LM Studio, and LocalAI are supported through provider rows with
api_key_env: "none" and can be routed to any audited agent type; use
host.docker.internal in the base URL when the backend runs inside Docker on
macOS. If a strict local proxy rejects bearer headers, set
payload.omit_auth_header: true on that provider row.
For actor calls, prefer the audited route names cohort_agent and hero_agent.
The internal worker job for one actor decision is actor_deliberation_call.
First Big Bang
worldfork init \
--name "Atlas onboarding" \
--scenario-file examples/test-big-bang.md \
--max-ticks 4 \
--tick-duration-minutes 720
The command waits for backend initialization to return and then prints the initialized workspace, actors, traits, graph baseline, sociology baseline, and emotion baseline.
This verifies setup and initialization. Use worldfork demo atlas for a full
tick, branch, endpoint-ledger, and report demonstration.
Inspect the initialized workspace:
worldfork watch big-bang <big-bang-id> --once
worldfork runs workspace <big-bang-id>
worldfork runs cost <big-bang-id>
worldfork ledgers list <big-bang-id>
worldfork logs list --status failed
For the detailed initializer, tick, queue, timing, and ledger flow behind this first run, see Runtime Walkthrough.
Stop Or Reset
Stop services without deleting data:
make down
Stop services and remove Docker volumes:
make clean-data