[drakulavich]

Closes #129.

Planned together with #131. The two turned out unrelated (the progress line versus a note under db rows), so #131 has its own PR.

What

• The progress line names the collection: syncing sleep for the daily_sleep endpoint, syncing sleep-periods for sleep, and so on. collectionOf in run-command.ts maps an endpoint to its collection, and both the sync runner and fetch pass it to pageProgress.
• The page counter restarts with each request. The client numbers pages within one fetch() call (PageEvent.page), and the progress line shows that number. A range split into 30-day pieces counts each piece's own pages, and the row count still runs across them.

Why

Reproduced on main: the new fetch test shows fetching heartrate: page 1 … page 6 for three 30-day pieces of two pages each. The sync progress test shows syncing daily_sleep, which is the endpoint of the sleep collection.

Left as is: the issue also notes that the last frame repeats the previous row total when the final page is empty. That frame reports a page that really arrived with 0 rows, so I did not change it.

How checked

• New test in src/commands/fetch.test.ts: fetchCollection for hr over 61 days, through a real OuraClient with a stubbed fetch. It checks all six frames (fetching hr: page 1, 2 rows so far… through page 2, 9 rows so far…). The sync terminal-progress test now expects syncing sleep and syncing sleep-periods. Both fail on main's sources. The existing progress and client tests now carry the page field.
• Mutation, reverted after one run of src/commands/fetch.test.ts: pages = page.page; → pages++; in progress.ts: 1 of 17 fails (the new test).
• bun run check: typecheck, Biome, Knip clean; 712 tests pass.
• Codex review: Approve, no findings (report in the first comment).

Overlap: src/commands/run-command.ts is also changed by #173 (different lines).

🤖 Generated with Claude Code

https://claude.ai/code/session_018qh9xqviBHYmzVtmQWSKqA

[drakulavich]

Codex review (gpt-6-luna, final, ebcf1cc..1837f83):

Reviewed ebcf1cc..1837f83. I found no issues.

The four touched test files passed: 94 tests, 0 failures. The worktree also has a pre-existing modification to assets/demo.gif, which I left untouched.

Simplification: not needed

Verdict: Approve

[greptile-apps]

Confidence Score: 5/5

[Low impact] The PR appears safe to merge, with the previously reported terminal display issue remaining as a non-blocking concern.

Findings

1.  Old characters remain visible ▶

Summary

The PR makes terminal progress use collection names and restart page counts for each request.

• Progress lines name collections and restart page counts for each request.

The previous-review merge bases differ, so this review checked the full PR diff.

_{Reviews (2) · Last reviewed commit: "Merge remote-tracking branch 'origin/mai..." · Reviewed by Greptile}

[greptile-apps]

Old characters remain visible

Restarting the page counter can leave old text on the terminal. When one request ends at page 10 and the next starts at page 1, with the row total still the same width, the new line is one character shorter. show() writes only a carriage return before the text, so the previous trailing character remains visible. Clear the line or pad shorter frames before writing them.