Initial commit -- did OMO do a good job?

2026-09-01 18:29:46 -04:00
commit 20a6c26f7c
179 changed files with 21704 additions and 0 deletions
@@ -0,0 +1,67 @@
+Todo 19 — news-web UI (CSR) with Leptos 0.8
+
+Completed: 2026-09-01
+
+Scope
+- Scaffolded `crates/news-web` as a Leptos 0.8 client-side-rendered (CSR)
+  web app, mirroring `runway-web`'s Trunk + Tailwind v4 toolchain.
+- Implemented three API client routes against the same-origin `/api/*` proxy:
+  - GET /api/stories
+  - POST /api/feedback
+  - GET /api/config
+- Built three UI views:
+  - Story list: table of clusters with title, source count, importance,
+    relevance, percentile badges, server decision, and per-row feedback
+    buttons (Interested / Not interested).
+  - Config view: read-only summary of topics, notification settings, and
+    sources.
+  - Error / loading states for both data fetches and feedback actions.
+- Kept the WASM bundle lean by avoiding `news-core` and any markdown/chart
+  libraries.
+
+Verification
+- `cargo check -p news-web --target wasm32-unknown-unknown` passes.
+- `trunk build --release` succeeds; generated WASM is
+  `news-web-263e7ab3a1a1c8aa_bg.wasm` at 525 KiB (well under the 1.8 MiB
+  budget enforced by todo 20).
+- `cargo fmt --all --check`, `cargo clippy --workspace --all-targets -- -D warnings`,
+  and `cargo test --workspace` all pass (87 tests total).
+
+Playwright QA — happy path
+- Seeded a temporary SQLite database with one source and one story cluster
+  ("Mars Rover Finds Water").
+- Started `target/debug/news-server` on port 3000 and `trunk serve` on
+  port 8080 with Trunk proxying `/api` to the backend.
+- Navigated to http://127.0.0.1:8080/ in Playwright.
+- Asserted the story title "Mars Rover Finds Water" rendered in the table.
+- Clicked the "Interested" feedback button.
+- Asserted the per-row "Saved" success indicator appeared.
+- Verified the network request log contained:
+  POST http://127.0.0.1:8080/api/feedback => 200 OK
+- Captured screenshot: `.omo/evidence/task-19-news-triage.png`
+
+Playwright QA — failure path
+- Stopped the backend server (port 3000).
+- Reloaded http://127.0.0.1:8080/ in Playwright.
+- Asserted the visible error state text "Could not load stories" appeared.
+- Verified the network request log contained:
+  GET http://127.0.0.1:8080/api/stories => 500 Internal Server Error
+- Captured screenshot: `.omo/evidence/task-19-news-triage-failure.png`
+
+Evidence files
+- `.omo/evidence/task-19-news-triage.png` — happy-path story list with
+  feedback success state.
+- `.omo/evidence/task-19-news-triage-failure.png` — failure-path error state.
+- `.omo/evidence/task-19-news-triage.txt` — this file.
+
+Notes / gotchas
+- Leptos 0.8's `Action::new` requires `Send` futures, which are not available
+  on `wasm32-unknown-unknown`. On WASM, use `Action::new_local` for actions
+  backed by `gloo_net` (or any non-Send JS future).
+- `news-server` stores `source.kind` and `raw_items.kind` as snake_case strings
+  (`rss_atom`, `news`). Seeding with `rss` or any other string causes a startup
+  parse error (`InvalidStoredValue`).
+- The Playwright MCP was configured to launch Google Chrome at
+  `/opt/google/chrome/chrome`; on this Arch box that required installing the
+  system `chromium` package and symlinking `/opt/google/chrome/chrome ->
+  /usr/bin/chromium` so the MCP could launch the browser.