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.

publish-inbox - Prompt - OpenSmartRoute

Name: publish-inbox
Author: lcilliers

Prompt file imported from lcilliers/learning4comfort (.github/prompts/publish-inbox.prompt.md). Copyright stays with the author.

Publish whatever is waiting in publication-inbox. Follow the workflow in README.md and the rules in SITE-DESIGN.md. Run Python as .venv\Scripts\python.

Invoking this command is the author's approval to publish. Site-wide legal and privacy compliance is handled separately; do not raise it for book updates.

Check the inbox. List the files in publication-inbox\<book>\ for every book folder. If nothing is there, say so and stop. Read each book's content\books\<book>\README.md and apply its rules and presentation overrides. Do not infer rules that are not written.
Convert.
- the_inner_being: run python scripts\publish_inner_being.py for a dry run and review the list of changed pages. Then run it with --write. It converts the Markdown, drops the 00-index file and any "Held" working notes, and rebuilds chapter contents, previous/next links and the opening-page contents. Chapters not in the inbox keep their published text.
- Any other book has no converter yet. Stop and ask the author how it should be converted. Do not guess a structure.
Verify. Spot-check that new or changed pages exist, that text matches the source, that heading anchors match the contents list, and that no "Held" or working-note text was published. Run python -m unittest discover -s tests.
Search indexes. The deploy workflow runs scripts\build_book_search_indexes.py itself, so nothing is committed for search. To check locally, copy site\ and content\ into a temporary folder, remove the README files, run python scripts\build_book_search_indexes.py --site <folder>, confirm the book's search-index.json includes the new pages, and delete the temporary folder.
Archive. Copy each inbox file to publication-archive\<book>\<yyyy-mm-dd>\. Delete the inbox copy only after its SHA-256 hash matches. Never commit the inbox or archive; both are git-ignored.
Track. Add a row to PUBLICATION-TRACKER.md with the date, "Approved and published (update)", the book, the source and destination, the publish method, and a short note. Say which chapters are new and which changed.
Publish. Stage only the book's content\books\<book> folder, PUBLICATION-TRACKER.md, and any script changes made in this run. Leave unrelated working-tree changes alone and mention them. Commit with a clear message and the trailer Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>, then git push. Watch the "Deploy Learning4Comfort site" run until it succeeds, then confirm that a new or changed page on https://learning4comfort.uk returns 200.
Report briefly: what was new or changed, the commit, the deploy result, and anything left unpublished or needing the author's decision.

Stop and ask the author, rather than continuing, when a file name or format is not recognised, when a converter would change a published page in a way the source does not explain, or when a check fails.

Install publish-inbox 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.

Publish whatever is waiting in `publication-inbox`. Follow the workflow in `README.md` and the rules in `SITE-DESIGN.md`. Run Python as `.venv\Scripts\python`. Invoking this command is the author's approval to publish. Site-wide legal and privacy compliance is handled separately; do not raise it for book updates. 1. **Check the inbox.** List the files in `publication-inbox\<book>\` for every book folder. If nothing is there, say so and stop. Read each book's `content\books\<book>\README.md` and apply its rules and presentation overrides. Do not infer rules that are not written. 2. **Convert.** - `the_inner_being`: run `python scripts\publish_inner_being.py` for a dry run and review the list of changed pages. Then run it with `--write`. It converts the Markdown, drops the `00-index` file and any "Held" working notes, and rebuilds chapter contents, previous/next links and the opening-page contents. Chapters not in the inbox keep their published text. - Any other book has no converter yet. Stop and ask the author how it should be converted. Do not guess a structure. 3. **Verify.** Spot-check that new or changed pages exist, that text matches the source, that heading anchors match the contents list, and that no "Held" or working-note text was published. Run `python -m unittest discover -s tests`. 4. **Search indexes.** The deploy workflow runs `scripts\build_book_search_indexes.py` itself, so nothing is committed for search. To check locally, copy `site\` and `content\` into a temporary folder, remove the README files, run `python scripts\build_book_search_indexes.py --site <folder>`, confirm the book's `search-index.json` includes the new pages, and delete the temporary folder. 5. **Archive.** Copy each inbox file to `publication-archive\<book>\<yyyy-mm-dd>\`. Delete the inbox copy only after its SHA-256 hash matches. Never commit the inbox or archive; both are git-ignored. 6. **Track.** Add a row to `PUBLICATION-TRACKER.md` with the date, "Approved and published (update)", the book, the source and destination, the publish method, and a short note. Say which chapters are new and which changed. 7. **Publish.** Stage only the book's `content\books\<book>` folder, `PUBLICATION-TRACKER.md`, and any script changes made in this run. Leave unrelated working-tree changes alone and mention them. Commit with a clear message and the trailer `Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>`, then `git push`. Watch the "Deploy Learning4Comfort site" run until it succeeds, then confirm that a new or changed page on https://learning4comfort.uk returns 200. 8. **Report briefly:** what was new or changed, the commit, the deploy result, and anything left unpublished or needing the author's decision. Stop and ask the author, rather than continuing, when a file name or format is not recognised, when a converter would change a published page in a way the source does not explain, or when a check fails.

{ "prompt": "Publish whatever is waiting in `publication-inbox`. Follow the workflow in `README.md` and the rules in `SITE-DESIGN.md`. Run Python as `.venv\\Scripts\\python`.\n\nInvoking this command is the author's approval to publish. Site-wide legal and privacy compliance is handled separately; do not raise it for book updates.\n\n1. **Check the inbox.** List the files in `publication-inbox\\<book>\\` for every book folder. If nothing is there, say so and stop. Read each book's `content\\books\\<book>\\README.md` and apply its rules and presentation overrides. Do not infer rules that are not written.\n\n2. **Convert.**\n - `the_inner_being`: run `python scripts\\publish_inner_being.py` for a dry run and review the list of changed pages. Then run it with `--write`. It converts the Markdown, drops the `00-index` file and any \"Held\" working notes, and rebuilds chapter contents, previous/next links and the opening-page contents. Chapters not in the inbox keep their published text.\n - Any other book has no converter yet. Stop and ask the author how it should be converted. Do not guess a structure.\n\n3. **Verify.** Spot-check that new or changed pages exist, that text matches the source, that heading anchors match the contents list, and that no \"Held\" or working-note text was published. Run `python -m unittest discover -s tests`.\n\n4. **Search indexes.** The deploy workflow runs `scripts\\build_book_search_indexes.py` itself, so nothing is committed for search. To check locally, copy `site\\` and `content\\` into a temporary folder, remove the README files, run `python scripts\\build_book_search_indexes.py --site <folder>`, confirm the book's `search-index.json` includes the new pages, and delete the temporary folder.\n\n5. **Archive.** Copy each inbox file to `publication-archive\\<book>\\<yyyy-mm-dd>\\`. Delete the inbox copy only after its SHA-256 hash matches. Never commit the inbox or archive; both are git-ignored.\n\n6. **Track.** Add a row to `PUBLICATION-TRACKER.md` with the date, \"Approved and published (update)\", the book, the source and destination, the publish method, and a short note. Say which chapters are new and which changed.\n\n7. **Publish.** Stage only the book's `content\\books\\<book>` folder, `PUBLICATION-TRACKER.md`, and any script changes made in this run. Leave unrelated working-tree changes alone and mention them. Commit with a clear message and the trailer `Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>`, then `git push`. Watch the \"Deploy Learning4Comfort site\" run until it succeeds, then confirm that a new or changed page on https://learning4comfort.uk returns 200.\n\n8. **Report briefly:** what was new or changed, the commit, the deploy result, and anything left unpublished or needing the author's decision.\n\nStop and ask the author, rather than continuing, when a file name or format is not recognised, when a converter would change a published page in a way the source does not explain, or when a check fails.", "variables": [], "notes": "Process new files in publication-inbox into book pages, rebuild search, archive, and publish the site", "metadata": { "source": { "provider": "github-prompts", "repository": "https://github.com/lcilliers/learning4comfort", "path": ".github/prompts/publish-inbox.prompt.md", "ref": "4ffc9a7284ec6817da1bfc25b2e4d1f7d711e6e4", "url": "https://github.com/lcilliers/learning4comfort/blob/4ffc9a7284ec6817da1bfc25b2e4d1f7d711e6e4/.github/prompts/publish-inbox.prompt.md", "key": "lcilliers/learning4comfort/.github/prompts/publish-inbox.prompt.md" } } }

publish-inbox

About

Install publish-inbox 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