@ai-ecoverse/slicc-bios 1.44.2 → 1.44.4

package/README.md

@@ -637,10 +637,10 @@ The newest release is live at <https://seven.sliccy.ai/>, and every pull request
 
 The hosts are served by the [sliccy-ai](https://github.com/ai-ecoverse/sliccy-ai) worker from the R2 bucket `slicc-bios`. This repo only publishes into that bucket:
 
-- [`edge/publish.mjs`](edge/publish.mjs) `<branch>` uploads `src/` to `seven/` for `main` and to `branches/<label>/` for every other branch, and deletes what the previous upload had and this one doesn't. A manifest next to it (`seven.json`, `branches/<label>.json`) records which branch owns the label, so two branches that map to the same label can't overwrite each other.
+- [`edge/publish.mjs`](edge/publish.mjs) `main` uploads `src/` to `seven/`, and `preview` to `branches/<label>/`, and each deletes what the previous upload had and this one doesn't. A manifest next to it (`seven.json`, `branches/<label>.json`) records which branch owns the label, so two branches that map to the same label can't overwrite each other.
 - The bucket expires everything under `branches/` 30 days after upload. Every push uploads all files again, so a branch stays up while it's active and disappears a month after its last push. Nothing needs to be cleaned up when a branch is deleted.
 - **seven** is published only by a release. On a push to `main`, the pipeline's release job runs semantic-release in the `production` environment, and when that tags a new `v<version>` its `after-release` command runs `edge/publish.mjs main`, which uploads `seven/`. A push that releases nothing (`chore:`, `ci:`, `docs:`, `test:` and the like) publishes nothing, so a change that only matters to seven needs a `fix:` or `feat:` commit to go live. A release job is never cancelled by a newer one.
-  - The upload uses `SLICC_BIOS_DEPLOY_TOKEN`, an account token with Workers R2 Storage Write, handed to wrangler as `CLOUDFLARE_API_TOKEN`. That token could overwrite `seven/`, so it lives only in the `production` environment, whose deployment branch policy allows `main` alone, and `ci.yml` gives it only to the release job (`release-secrets`). The pipeline keeps its own `CLOUDFLARE_API_TOKEN`, the reviewers' screenshot token, out of `after-release`, hence the separate name.
+  - The upload goes over R2's S3-compatible API with `SLICC_BIOS_DEPLOY_TOKEN`, a Cloudflare API token with only Workers R2 Storage Bucket Item Read and Write on the `slicc-bios` bucket (Cloudflare honours bucket-scoped tokens on the S3 API only, not the REST API that `wrangler r2 object` uses). `edge/publish.mjs` derives the S3 credentials in the release job, in memory: it asks Cloudflare's token verify endpoint for the token's id (the access key id) and hashes the token with SHA-256 (the secret access key, as Cloudflare defines it), and signs each request with SigV4 through [aws4fetch](https://github.com/mhart/aws4fetch). Neither is printed or written anywhere. Files go up eight at a time, then deletions, then the manifest. That token could overwrite `seven/`, so it lives only in the `production` environment, whose deployment branch policy allows `main` alone, and `ci.yml` gives it only to the release job (`release-secrets`). The pipeline keeps its own `CLOUDFLARE_API_TOKEN`, the reviewers' screenshot token, out of `after-release`, hence the separate name.
 - **A preview** is published by the pipeline's `deploy` job on every pull request, in the `staging` environment, with `id-token: write`. It runs `edge/publish.mjs preview`, which publishes the pull request's merge commit when its branch (`GITHUB_HEAD_REF`) is a `bb/**` branch, writes the URL to the job's `url` output, and skips everything else: other branches, and a pull request from a fork, which gets no OIDC token. A publish that fails (the route refuses it or can't be reached) writes `preview failed: <reason>` to the job summary and sets no URL, so the tests and reviews still run; none of them needs `STAGING_URL`. A push to a branch without a pull request publishes nothing. A newer push cancels the running CI, its publish included.
   - **A preview** holds no secret: `ci.yml` sets no `deploy-secrets`, and the pipeline (from v2.12.0) passes the deploy job only the secrets that input names, so it gets none. `edge/publish.mjs` asks GitHub for the job's OIDC token (audience `https://sliccy.ai`) and publishes through the sliccy-ai worker's `/api/publish/*` route. The route checks the token's signature, issuer, audience, expiry, repository, the `staging` environment and the `refs/pull/<n>/merge` ref, then derives `branches/<label>/` from the token's `head_ref`. A branch can only ever write its own prefix, and never `seven/`. It first writes a manifest listing both the previous upload's files and this one's, so a publish cancelled midway leaves every file it uploaded on the list for the next publish to delete. Then files go up eight at a time, then deletions, then the final manifest.
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-ecoverse/slicc-bios",
-  "version": "1.44.2",
+  "version": "1.44.4",
   "description": "Slim loader that boots SLICC into OPFS and serves the UI from there",
   "type": "module",
   "license": "Apache-2.0",
@@ -50,13 +50,13 @@
     "@ai-ecoverse/slicc-spectrum": "1.104.1",
     "@ai-ecoverse/slicc-vault": "1.16.4",
     "@biomejs/biome": "2.5.15",
+    "aws4fetch": "1.0.20",
     "es-module-lexer": "3.0.4",
     "fake-indexeddb": "6.2.6",
     "lefthook": "2.2.2",
     "modern-tar": "0.8.6",
     "monocart-coverage-reports": "2.13.2",
     "semantic-release": "25.0.9",
-    "wrangler": "4.151.1",
     "ws": "8.22.0"
   }
 }