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.

run-deep - Prompt - OpenSmartRoute

Name: run-deep
Author: albegali

Prompt file imported from albegali/agentic-portal (.codex/prompts/run-deep.md). Fill in {{arguments}} before use. Copyright stays with the author.

Recommended model: gpt-5 (codex -p deep) (tier deep, see models.toml). Switch to it before running this command.

Arguments: {{arguments}}.

ID only, no RUN or RUN in Draft without strategy → follow workflow/steps/run-strategy.md.
approve (typed by the human) → on the story branch (create it now if missing, delivery/guide/git.md): record the approval in RUN §2 (Approved, reviewer = project.toml owner, today's date), status Strategy Approved, Spec status In Progress; commit these two files alone, with the trailers Approved-by: <owner> (via /run approve) and Approved-at: <ISO datetime>; then follow workflow/steps/execute.md. If the human already ran make approve … GATE=strategy, just execute. Quick track (delivery/guide/quick.md): this one approval also covers the Spec, which may still be Draft.
revise: … → update only the strategy (untick Stopped in §2, status Draft, note the request in §2) and stop again.
ID only with an approved strategy → resume following workflow/steps/execute.md.
tuning: … → in-scope fix recorded in RUN §5 (see workflow/steps/tuning.md); a Persisted or Reviewed RUN goes back to Tuning (then /persist again).
approve-deviation <n> → fill Approved by (project.toml owner) and Date of RUN §6 Deviation #<n>; if the Spec is not In Progress or Ready, stop and say what the Spec still needs. Otherwise apply its plan impact to §1 as new steps, RUN Executing, Spec In Progress, resume following workflow/steps/execute.md.
cancel: <reason> → the owner dropped the story: status = "cancelled" on the story in backlog/stories.toml (reason in notes); RUN (if any) Cancelled with the reason in §9 and code changes reverted or listed there; Spec (if any) Cancelled; Brief Superseded. No new RUN.

L/XL stories run through /run-deep <ID> (same procedure, deep tier); otherwise use the deep tier only if the user asked for it.

Install run-deep 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.

> Recommended model: **gpt-5** (`codex -p deep`) (tier `deep`, see models.toml). Switch to it before running this command. Arguments: {{arguments}}. - ID only, no RUN or RUN in Draft without strategy → follow `workflow/steps/run-strategy.md`. - `approve` (typed by the human) → on the story branch (create it now if missing, `delivery/guide/git.md`): record the approval in RUN §2 (Approved, reviewer = `project.toml` owner, today's date), status `Strategy Approved`, Spec status `In Progress`; commit these two files alone, with the trailers `Approved-by: <owner> (via /run approve)` and `Approved-at: <ISO datetime>`; then follow `workflow/steps/execute.md`. If the human already ran `make approve … GATE=strategy`, just execute. Quick track (`delivery/guide/quick.md`): this one approval also covers the Spec, which may still be `Draft`. - `revise: …` → update only the strategy (untick `Stopped` in §2, status `Draft`, note the request in §2) and stop again. - ID only with an approved strategy → resume following `workflow/steps/execute.md`. - `tuning: …` → in-scope fix recorded in RUN §5 (see `workflow/steps/tuning.md`); a `Persisted` or `Reviewed` RUN goes back to `Tuning` (then `/persist` again). - `approve-deviation <n>` → fill `Approved by` (`project.toml` owner) and `Date` of RUN §6 `Deviation #<n>`; if the Spec is not `In Progress` or `Ready`, stop and say what the Spec still needs. Otherwise apply its plan impact to §1 as new steps, RUN `Executing`, Spec `In Progress`, resume following `workflow/steps/execute.md`. - `cancel: <reason>` → the owner dropped the story: `status = "cancelled"` on the story in `backlog/stories.toml` (reason in `notes`); RUN (if any) `Cancelled` with the reason in §9 and code changes reverted or listed there; Spec (if any) `Cancelled`; Brief `Superseded`. No new RUN. L/XL stories run through `/run-deep <ID>` (same procedure, deep tier); otherwise use the deep tier only if the user asked for it.

{ "prompt": "\n> Recommended model: **gpt-5** (`codex -p deep`) (tier `deep`, see models.toml). Switch to it before running this command.\n\nArguments: {{arguments}}.\n\n- ID only, no RUN or RUN in Draft without strategy → follow `workflow/steps/run-strategy.md`.\n- `approve` (typed by the human) → on the story branch (create it now if missing, `delivery/guide/git.md`): record the approval in RUN §2 (Approved, reviewer = `project.toml` owner, today's date), status `Strategy Approved`, Spec status `In Progress`; commit these two files alone, with the trailers `Approved-by: <owner> (via /run approve)` and `Approved-at: <ISO datetime>`; then follow `workflow/steps/execute.md`. If the human already ran `make approve … GATE=strategy`, just execute. Quick track (`delivery/guide/quick.md`): this one approval also covers the Spec, which may still be `Draft`.\n- `revise: …` → update only the strategy (untick `Stopped` in §2, status `Draft`, note the request in §2) and stop again.\n- ID only with an approved strategy → resume following `workflow/steps/execute.md`.\n- `tuning: …` → in-scope fix recorded in RUN §5 (see `workflow/steps/tuning.md`); a `Persisted` or `Reviewed` RUN goes back to `Tuning` (then `/persist` again).\n- `approve-deviation <n>` → fill `Approved by` (`project.toml` owner) and `Date` of RUN §6 `Deviation #<n>`; if the Spec is not `In Progress` or `Ready`, stop and say what the Spec still needs. Otherwise apply its plan impact to §1 as new steps, RUN `Executing`, Spec `In Progress`, resume following `workflow/steps/execute.md`.\n- `cancel: <reason>` → the owner dropped the story: `status = \"cancelled\"` on the story in `backlog/stories.toml` (reason in `notes`); RUN (if any) `Cancelled` with the reason in §9 and code changes reverted or listed there; Spec (if any) `Cancelled`; Brief `Superseded`. No new RUN.\n\nL/XL stories run through `/run-deep <ID>` (same procedure, deep tier); otherwise use the deep tier only if the user asked for it.", "variables": [ "arguments" ], "notes": "Developer mode — /run on the deep tier, for L/XL stories (same arguments and procedure) Arguments: <ID> [approve | revise: … | tuning: … | approve-deviation <n> | cancel: …].", "metadata": { "source": { "provider": "github-codex-prompts", "repository": "https://github.com/albegali/agentic-portal", "path": ".codex/prompts/run-deep.md", "ref": "0694fcfa6f9f742ee5403ca5c19a53b6c55c6d0c", "url": "https://github.com/albegali/agentic-portal/blob/0694fcfa6f9f742ee5403ca5c19a53b6c55c6d0c/.codex/prompts/run-deep.md", "key": "albegali/agentic-portal/.codex/prompts/run-deep.md" } } }

run-deep

About

Install run-deep 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