release: promote develop → main (v0.11.0)

.changeset/ci-889-coverage-floors.md

@@ -0,0 +1,6 @@
+---
+"ornn-api": patch
+"ornn-web": patch
+---
+
+Enforce per-package line-coverage floors (api 75%, web 12% ratchet) and split the pooled Codecov flag (#889)

.changeset/empty-772-todo.md

@@ -0,0 +1,2 @@
+---
+---

.changeset/fix-608-chat-completion-tool-calls.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Chat-completion providers now surface structured `execute_in_sandbox` tool calls in the Playground. `parseChatCompletionStream` accumulates streamed `choices[].delta.tool_calls[]` fragments per index and flushes each completed call as a `response.output_item.done` function_call event (the shape the playground consumer reads); the non-streamed `complete()` path maps `message.tool_calls[]` the same way; and the chat-completion request body now sends `tool_choice: "auto"` whenever tools are supplied. Previously tool calls were silently dropped so the sandbox never ran for `apiFormat=chat-completion` providers (#608)

.changeset/fix-632-zipbomb-chokepoint.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Server-side ZIP-bomb guards (cumulative/per-entry uncompressed caps, file-count, compression-ratio) are now enforced at the skill-ingestion chokepoint — covering both direct upload and the GitHub pull/refresh paths that previously bypassed the route-layer guard. Caps are env-overridable (#632)

.changeset/fix-750-version-write-guid.md

@@ -0,0 +1,5 @@
+---
+"ornn-web": patch
+---
+
+Fix version delete/deprecation 404 when Skill Detail is opened by name: the GUID-only version-write routes now always receive the skill GUID on the wire while cache invalidation stays keyed on idOrName (#750)

.changeset/fix-751-bell-refetch-on-open.md

@@ -0,0 +1,5 @@
+---
+"ornn-web": patch
+---
+
+Notification bell dropdown now refetches the list when opened, so a new broadcast reflected in the badge also appears in the list without a manual refresh (#751)

.changeset/fix-752-skill-timestamp-locale.md

@@ -0,0 +1,5 @@
+---
+"ornn-web": patch
+---
+
+skill timestamps (registry card, version list, detail page) now follow the active UI language — Chinese mode no longer shows English month names (#752)

.changeset/fix-757-redemption-code-randomint.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Redemption-code generation now draws each character with crypto.randomInt for an unbiased uniform draw over the alphabet, clearing CodeQL js/biased-cryptographic-random (#757)

.changeset/fix-757-sdk-baseurl-redos.md

@@ -0,0 +1,5 @@
+---
+"@chronoai/ornn-sdk": patch
+---
+
+TS SDK baseUrl normalization strips trailing slashes with a linear loop instead of a regex, removing a polynomial ReDoS vector on pathological all-slash inputs (#757)

.changeset/fix-766-abort-after-billable-charge.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Playground chat: a client abort (or error) after billable output now commits the reserved quota slot instead of refunding it, closing the abort-after-first-token free-usage path (#766)

.changeset/fix-888-lint-warnings.md

@@ -0,0 +1,6 @@
+---
+"ornn-api": patch
+"ornn-web": patch
+---
+
+Resolve all 65 ESLint warnings with proper type/effect/ref fixes; compiler-only rules off pending react-compiler (#888)

.changeset/fix-908-sort-dir-400.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Admin users list: invalid sort/dir query params now return 400 invalid_sort/invalid_dir instead of 500 (#908)

.changeset/fix-920-nan-limit.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Malformed or empty ?limit on the notifications feed now falls back to the default page size instead of erroring (#920)

.changeset/fix-940-delete-skill-cache-removal.md

@@ -0,0 +1,5 @@
+---
+"ornn-web": patch
+---
+
+Deleting a skill now removes its detail/versions cache (predicate covers name- and guid-keyed entries) so the page no longer refetches the deleted skill → 404 (#940)

.changeset/fix-941-delete-skill-sidebar-counts.md

@@ -0,0 +1,5 @@
+---
+"ornn-web": patch
+---
+
+Deleting a skill now also refreshes the My-Skills filter-sidebar facet + grants-summary counts so they stay consistent without a full-page refresh (#941)

.changeset/lifecycle-orbital-ring.md

@@ -0,0 +1,18 @@
+---
+"ornn-web": minor
+---
+
+feat(web): interactive skill-lifecycle orbital ring on the landing hero
+
+Overlay an orbital "skill lifecycle" ring on the looping hero video: eight
+agent-facing stages (search → preview → audit → install → execute → build →
+publish → share, and back to search) around a hairline circle, with an ember
+comet orbiting clockwise to signal the loop. Hovering / focusing / tapping a
+stage ignites the node, draws a connector to the centre, and morphs the hub
+into a detail card (agent-API hint + CTA) for that stage. Dark + light themes,
+reduced-motion fallback, mobile tap support, and EN + ZH copy.
+
+Also replace the two centred landing announcement modals with a single global
+announcement banner: a collapsed top-right headline pill that expands on hover
+into a dismissable stack, aggregating the hardcoded launch announcement and the
+dynamic announcements, shown on every page.

.changeset/loose-bugs-show.md

@@ -0,0 +1,2 @@
+---
+---

.changeset/release-notes-20260609.md

@@ -0,0 +1,4 @@
+---
+---
+
+chore: release notes for v0.11.0 (docs-only, no version impact)

.changeset/test-872-mirror-coverage.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Add unit test coverage for the GitHub mirror auth, REST client, and admin routes (#872)

.changeset/test-873-audit-coverage.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Add unit test coverage for the skill-audit module (prompts, repository, service, routes) (#873)

.changeset/test-874-crud-coverage.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Add unit + integration test coverage for the skills CRUD module (repositories, routes, service) (#874)

.changeset/test-875-generation-coverage.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Add unit test coverage for the skill-generation module (prompts, service, routes) (#875)

.changeset/test-876-search-coverage.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Add test coverage for the skills search module and remove a dead search type (#876)

.changeset/test-877-admin-routes-coverage.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Add route test coverage for admin skills, platform settings, and admin users (#877)

.changeset/test-878-me-users-coverage.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Add route test coverage for /me and /users surfaces (#878)

.changeset/test-879-notifications-coverage.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Add notifications test coverage (routes, service emitters, migration, bootstrap) and remove a dead test mock (#879)

.changeset/test-880-analytics-coverage.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Add analytics test coverage (service pass-through, route validation, PostHog tracker fail-open) (#880)

.changeset/test-881-settings-coverage.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Add settings + LLM-provider test coverage (migration, repositories, routes) (#881)

.changeset/test-882-content-quota-coverage.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Add broadcasts, announcements, and quota route/bootstrap/migration test coverage (#882)

.changeset/test-883-clients-infra-coverage.md

@@ -0,0 +1,5 @@
+---
+"ornn-api": patch
+---
+
+Add client and infra test coverage (LLM model-list client, sandbox client, Mongo connect) (#883)

.changeset/test-884-honest-web-coverage.md

@@ -0,0 +1,5 @@
+---
+"ornn-web": patch
+---
+
+Honest vitest coverage config: v8 provider, all-files reporting over src/** with justified excludes (#884)

.changeset/test-885-web-utils-coverage.md

@@ -0,0 +1,5 @@
+---
+"ornn-web": patch
+---
+
+Cover web utils and lib modules; drop two dead formatter exports (#885)

.changeset/test-886-web-components.md

@@ -0,0 +1,5 @@
+---
+"ornn-web": patch
+---
+
+Cover form and skeleton components; remove dead FileUpload and unused skeleton variants (#886)

.changeset/test-887-web-features.md

@@ -0,0 +1,5 @@
+---
+"ornn-web": patch
+---
+
+Raise page/feature component coverage; drop a dead type re-export (#887)

.changeset/test-931-backfill.md

@@ -0,0 +1,5 @@
+---
+"ornn-web": patch
+---
+
+Backfill per-component reset/derivation tests for the #888 effect refactors (#931)

.changeset/thin-tips-wave.md

@@ -0,0 +1,2 @@
+---
+---

.github/release-notes-20260609.md

@@ -0,0 +1,20 @@
+## Fixed
+
+- Runtime and mixed skills now run in Playground on more LLM providers.
+- Playground usage is charged correctly even when a chat is interrupted.
+- Deleting a non-latest skill version now works from name-based links.
+- The notification dropdown shows new broadcasts without a manual refresh.
+- Deleting a skill no longer leaves stale requests or incorrect counts.
+- Skill timestamps now follow the selected interface language.
+- Oversized (zip-bomb) skill uploads are now rejected server-side.
+- Few technical bugs fixed.
+
+## New Feature
+
+- Global announcement banner for launch and platform notices.
+- Skill-lifecycle ring on the landing hero.
+
+## Changed
+
+- Landing popups replaced by the unified announcement banner.
+- Technical enhancement.

.github/workflows/ci.yml

@@ -29,18 +29,52 @@ jobs:
       - uses: oven-sh/setup-bun@v2
       - run: bun install --frozen-lockfile
       - run: bun run test:coverage
-      # Codecov upload (#471). No token needed for public repos.
-      # `directory: ./coverage` picks up the per-package lcov files
-      # that bun test --coverage writes into each workspace.
-      - name: Upload coverage to Codecov
+      # ornn-api line-coverage floor (#889). Bun's bunfig `coverageThreshold`
+      # is silently ignored on the pinned Bun (1.3.x) — it prints the table
+      # but never fails the run — so the floor is enforced here by summing
+      # LH/LF across the lcov that `test:coverage` just wrote. ornn-web's
+      # floor is enforced inside the run by vitest `thresholds.lines` in
+      # ornn-web/vitest.config.ts. Floor = operator's ≥75% gate; ratchet up
+      # in a reviewed one-line diff as real coverage rises (measured 96.6%).
+      - name: Enforce ornn-api line-coverage floor (>=75%)
+        run: |
+          awk -F: '
+            /^LF:/ { lf += $2 }
+            /^LH:/ { lh += $2 }
+            END {
+              if (lf == 0) { print "no lines found in lcov"; exit 1 }
+              pct = 100 * lh / lf
+              printf "ornn-api line coverage: %.2f%% (%d/%d), floor 75%%\n", pct, lh, lf
+              if (pct < 75) { print "::error::ornn-api line coverage below 75% floor"; exit 1 }
+            }
+          ' ornn-api/coverage/lcov.info
+      # Codecov upload (#471, #889). No token needed for public repos.
+      # The pooled `bun` flag was split per-package so each workspace's
+      # lcov is attributed to its own flag (`api` / `web` / `sdk-ts`) and
+      # judged on its own target in codecov.yml. One upload step per flag
+      # because a single codecov-action invocation applies one flag set to
+      # every file it uploads.
+      - name: Upload ornn-api coverage to Codecov
+        uses: codecov/codecov-action@v4
+        if: ${{ !cancelled() }}
+        with:
+          flags: api
+          fail_ci_if_error: false
+          files: ./ornn-api/coverage/lcov.info
+      - name: Upload ornn-web coverage to Codecov
+        uses: codecov/codecov-action@v4
+        if: ${{ !cancelled() }}
+        with:
+          flags: web
+          fail_ci_if_error: false
+          files: ./ornn-web/coverage/lcov.info
+      - name: Upload TS SDK coverage to Codecov
         uses: codecov/codecov-action@v4
         if: ${{ !cancelled() }}
         with:
-          flags: bun
+          flags: sdk-ts
           fail_ci_if_error: false
-          # Per-package lcov files land under coverage/ in each
-          # workspace; codecov-action recurses by default.
-          files: ./ornn-api/coverage/lcov.info,./ornn-web/coverage/lcov.info,./sdk/typescript/coverage/lcov.info
+          files: ./sdk/typescript/coverage/lcov.info
 
   python-sdk-test:
     runs-on: ubuntu-latest
@@ -102,6 +136,8 @@ jobs:
           deps += cfg['project']['optional-dependencies']['dev']
           subprocess.check_call([sys.executable, '-m', 'pip', 'install', *deps])
           "
+          # pip itself is in the audited env — keep it current (PYSEC-2026-196, #935)
+          python -m pip install --upgrade pip
           pip install pip-audit
       - name: Audit dependencies
         working-directory: sdk/python

.gitignore

@@ -1,5 +1,6 @@
 node_modules/
 dist/
+coverage/
 .env
 .env.*
 !.env.sample.*

codecov.yml

@@ -1,35 +1,68 @@
 # Codecov configuration for chrono-ornn (#471).
 #
 # Reports are uploaded from two CI jobs:
-#   - `test`            → flag `bun` (per-workspace lcov)
+#   - `test`            → flags `api`, `web`, `sdk-ts` (one per workspace lcov)
 #   - `python-sdk-test` → flag `python` (sdk/python coverage.xml)
 #
-# Target is realistic-not-aspirational: 70% project coverage as the
-# initial commitment. Files that are deliberately not covered today
-# (god-files, generation prompts, infra shims) are excluded so the
-# headline number reflects code that's reasonably testable.
+# The pooled `bun` flag was split per-package (#889) so each surface is
+# judged on its own honest coverage rather than a blended headline that
+# lets a well-tested package mask a thin one. Per-flag status targets are
+# the real signal; the global project default is kept as a coarse backstop.
+#
+# Files that are deliberately not covered today (god-files, generation
+# prompts, infra shims) are excluded so each number reflects code that's
+# reasonably testable.
 
 coverage:
   status:
     project:
+      # Coarse backstop across all flags. Per-flag targets below are the
+      # signal that actually gates each package honestly.
       default:
         target: 70%
         threshold: 1%
         if_ci_failed: error
+      # ornn-api is well covered today — hold it near its real number.
+      api:
+        flags:
+          - api
+        target: 90%
+        threshold: 2%
+        if_ci_failed: error
+      # ornn-web is early — floor it at its measured level, ratchet later.
+      web:
+        flags:
+          - web
+        target: 14%
+        threshold: 2%
+        if_ci_failed: error
+      # TS SDK — informational until it has a real test suite.
+      sdk-ts:
+        flags:
+          - sdk-ts
+        target: auto
+        informational: true
     patch:
       default:
         target: 80%
         threshold: 5%
         if_ci_failed: error
 
-# Source-of-truth flags. Codecov sums across flags by default; both
-# are surfaced separately in the PR comment so a regression in one
-# language is obvious.
+# Source-of-truth flags. Each Bun workspace uploads under its own flag so
+# a regression in one package is isolated and visible (#889). Codecov sums
+# across flags for the headline; per-flag carryforward keeps a package's
+# last-known coverage when its upload is absent from a given PR.
 flags:
-  bun:
+  api:
     paths:
       - ornn-api/src/
+    carryforward: true
+  web:
+    paths:
       - ornn-web/src/
+    carryforward: true
+  sdk-ts:
+    paths:
       - sdk/typescript/src/
     carryforward: true
   python: