How long does the integration take?

Change one base URL in an OpenAI-compatible client, or install the Python or TypeScript SDK; the quickstart routes a first request in a few minutes.

Will this lock us into another vendor?

No. The router is Apache-2.0 open source, your provider keys stay yours, and the API is OpenAI-compatible, so leaving is a change of base URL.

What happens when a provider fails?

The router falls back to the next model your policy allows and records the fallback in the trace, so an outage costs a retry, not an incident.

Aria-Agent - Skill - OpenSmartRoute

Imported from ashishkumar006/Aria-Agent (AGENTS.md). Install upstream with npx skills add ashishkumar006/Aria-Agent. Copyright stays with the author.

Aria — agent instructions

Layout

S9SharedCode/code/ — agent server (agent_server.py), skills, executor (flow.py), React console (console-frontend/). Run agent commands from this dir.
llm_gatewayV9/ — LLM gateway (own venv). Run gateway commands from this dir.
agent-desktop/ — Electron shell. build_desktop.ps1 (repo root) builds the portable app at agent-desktop/dist/Aria/Aria.exe and its shortcuts; the shell drives the same :8500 console, it is not a second backend.

Services (restart to pick up backend changes)

Agent: http://127.0.0.1:8500 — start with uv run agent_server.py in S9SharedCode/code
Gateway: http://127.0.0.1:8109 — start with uv run main.py in llm_gatewayV9
Shortcut: powershell -ExecutionPolicy Bypass -File start_aria.ps1 starts both (re-runs reuse whatever is already listening, waits for health, opens the console; -NoBrowser skips the browser). stop_aria.ps1 stops the listeners on both ports.
To restart: kill the listener on the port, then start fresh. Never assume a restart happened — verify with /api/health (agent) and /v1/providers (gateway).

Verify loop (run after every backend/frontend change)

Agent tests: uv run python -m pytest tests/ -q from S9SharedCode/code (NOT bare uv run pytest — the shim resolves the wrong interpreter and fails on numpy). Full suite ≈ 5–6 min; targeted files are fine.
Gateway tests: uv run python -m pytest -q from llm_gatewayV9 — NOT tests/. 14 of the 16 adaptor suites, plus policy/ and integrations/, live outside tests/, so pytest tests/ -q passes happily while skipping 344 tests. Full suite ≈ 1.5 min.
Frontend: npx tsc --noEmit then npm run build from console-frontend/.
Live: GET /api/health must show gateway_up: true.

Standing rules

NEVER read .env files (enforced by .opencode/plugins/env-protection.js). Flip flags with blind byte-replacements and verify with a single-line grep.
PowerShell 5.1: ; chains commands (no &&); quote paths with spaces; > writes UTF-16 — prefer Set-Content/atomic writes via Python.
Inline python -c breaks on quoting — write temp scripts under the system temp dir instead, delete when done.
Scratch E2E/visual scripts (*.cjs, temp .py) live outside the repo and are deleted after the run they verify.

Install Aria-Agent and it becomes one more candidate the router can pick - when it fits.

A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.

Aria-Agent

About

Aria — agent instructions

Layout

Services (restart to pick up backend changes)

Verify loop (run after every backend/frontend change)

Standing rules

Install Aria-Agent and it becomes one more candidate the router can pick - when it fits.

What it costs to adopt - in engineering time.

One base URL

No vendor lock-in

Failover built in

Use it

Manifest

Reviews