* refactor(pagination): hoist cursor codec to utils/ for cross-domain reuse
The keyset cursor codec was asset-namespaced (app/assets/services/cursor.py)
but the wire format and encode/decode logic are domain-agnostic. Move it to
utils/cursor.py so the jobs endpoint can share one codec instead of importing
across domains or duplicating it.
* feat(jobs): cursor-based pagination on GET /api/jobs (BE-943)
Mirror the cloud jobs cursor (BE-885) on the OSS Python server so the
frontend sees one contract across runtimes.
- apply_sorting now appends the job id as a tiebreaker, making (create_time,
id) a stable keyset; without it, ties could reorder between pages.
- get_all_jobs accepts an opaque 'after' cursor (honored only for created_at
sort, like cloud), keyset-filters the sorted in-memory list, and returns
has_more + a next_cursor. Minted in offset mode too so a client can bootstrap
into keyset pagination.
- server.py /api/jobs parses 'after', returns next_cursor in the pagination
object, and maps a malformed cursor to 400 INVALID_CURSOR.
- Reuses the shared utils.cursor codec (base64url JSON {s,v,id,o}) so the wire
format matches cloud and assets exactly.
Tests: asc/desc multi-page round-trip, same-create_time tiebreaker, last-page
no-cursor, offset-mode bootstrap, execution_duration ignores cursor, malformed
cursor raises.
* refactor(jobs): return NamedTuple page, early-out on empty job set
Review feedback on the jobs cursor pagination:
- get_all_jobs now returns JobsPage, a NamedTuple, instead of a bare
4-tuple (callers unpack positionally either way).
- Early-out when the filtered job set is empty so paging code never has
to reason about indexing into an empty list. A malformed 'after'
cursor is still decoded first and rejected with INVALID_CURSOR.
- Document that job ids are server-assigned UUIDs, always present and
unique — the empty-string fallback in _job_id_key only shields
sorted() from a malformed dict, it is not part of the keyset
contract.
The /view endpoint returns text/plain for .svg files on some platforms
because Python's mimetypes module does not always include SVG by default.
Explicitly register image/svg+xml so <img> tags can render SVGs correctly.
Amp-Thread-ID: https://ampcode.com/threads/T-019d2da7-6a64-726a-af91-bd9c44e7f43c
* add 'is_default' to model paths config
including impl and doc in example file
* update weirdly overspecific test expectations
* oh there's two
* sigh
* Expand user path.
* Add test.
* Add unit test for expanding base path.
* Simplify unit test.
* Remove comment.
* Remove comment.
* Checkpoints.
* Refactor.
