urish

AI Cost Tracking

🤖 LLM usage: $0.0295 (4 commits)
👤 Human dev: ~$150 (1.5h @ $100/h, 30min dedup)

Generated on 2026-06-15 using openrouter/qwen/qwen3-coder-next

Unified URI shell facade: one command (uri) over uri3, uri2run, uri2ops, hypervisor, urigen.

uri call python://uri2voice.stt:transcribe --payload '{"text":"test"}'
uri call python://uri2voice.stt:transcribe --payload @payload.json
uri explain weather://forecast/Gdansk/14/html
uri agent health weather-map-agent.local
uri doctor
uri doctor --strict
uri ask "stworz agenta pogodowego z healthcheckiem"
uri ask "stwórz web UI agenta hypervisor-dashboard do pokazywania procesów"
uri dashboard create hypervisor-dashboard --plan-only
uri watch health://agent/weather-map-agent.local
uri stream sse://localhost:8791/events
uri shell   # same as running `uri` with no arguments

Running uri or urish with no arguments opens the interactive shell.

Semantic model everywhere:

URI + payload + context + policy → explain → execute → ServiceResult envelope

Subcommands

Command	Backend
`call`, `explain`, `plan`	uri3 + uri2run
`run`	uri2ops / uri2flow
`logs`, `watch`	uri3 logs
`stream`	SSE/WS
`ask`, `nl`	nl2uri + urigen
`select`	envelope piping
`agent`	hypervisor
`dashboard`	urigen + hypervisor dashboard agent
`repair`	repair supervisor
`ticket`	planfile
`evolve`, `proposal`	evolution
`ecosystem`	urigen
`context`	config/urish/contexts

Policy

uri call repair://agent/demo/apply --dry-run
uri call repair://agent/demo/apply --approve --policy prod
uri run workflow://flow/demo --approve

Exit codes: 0=ok, 1=failed, 2=execution, 3=validation, 4=policy blocked, 5=not found, 6=dependency missing.

Legacy CLIs remain as backends; uri is the user-facing facade.

Learning curve: start with docs/GETTING_STARTED.md — one shell, one lifecycle, profiles instead of package names.

Dashboard Agent Workflow

urish creates the dashboard as a controlled URI ecosystem, not as an ad-hoc web app:

uri ecosystem plan "stwórz web UI agenta hypervisor-dashboard do pokazywania procesów" \
  --profile dashboard-agent \
  --out output/proposals/hypervisor-dashboard.ecosystem.proposal.yaml
uri ecosystem generate output/proposals/hypervisor-dashboard.ecosystem.proposal.yaml \
  --out output/ecosystems/hypervisor-dashboard
uri ecosystem verify output/ecosystems/hypervisor-dashboard/ecosystem.yaml
uri ecosystem apply output/ecosystems/hypervisor-dashboard/ecosystem.yaml --plan
uri agent run hypervisor-dashboard.local --wait-healthy --approve

Shortcut:

uri dashboard create hypervisor-dashboard --plan-only
uri agent create-dashboard hypervisor-dashboard --dry-run

Read-only dashboard views stay behind resource://dashboard/... contracts and view://... capabilities. Mutating actions such as repair://.../apply remain approval-gated.

The generated ecosystem also carries the dashboard app source under app/ (hypervisor_dashboard_agent.main:app) so the workflow artifact contains the system agent implementation, not only registrations.

Ticket → Evolution → Ecosystem

When a planfile task describes a dashboard agent, urish chains the ticket flow into the ecosystem pipeline:

uri ticket show ticket://feature/PL-10
uri evolve from-ticket ticket://feature/PL-10
uri proposal verify evolution/proposals/proposal_from_ticket_PL-10.yaml
uri proposal apply evolution/proposals/proposal_from_ticket_PL-10.yaml --sandbox
uri ecosystem generate output/proposals/hypervisor-dashboard.ecosystem.proposal.yaml \
  --out output/ecosystems/hypervisor-dashboard
uri ecosystem apply output/ecosystems/hypervisor-dashboard/ecosystem.yaml --approve

uri doctor --strict additionally validates incidents, tickets, evolution proposals, and artifact lifecycle envelope coverage.

Internal Shape

urish.cli only wires the Typer app and high-level command groups. Runtime commands (call, explain, plan, run, logs, watch, stream) are registered from urish.commands.runtime; backend work stays in urish.backends.*.

Install

pip install -e .
uri --help

License

Licensed under Apache-2.0.