feat: create_paper_note — quote-less paper-level commentary (#185 PR4/5)#191
Merged
feat: create_paper_note — quote-less paper-level commentary (#185 PR4/5)#191
Conversation
Foundation for the `create_paper_note` tool (PR4 part 1 of 3). Paper-level notes are commentary on a publication as a whole — no quote, no char_range, no fuzzy match needed. Adding a second synthetic-anchor namespace lets the resolver short-circuit them to AnchorStatus::Ok the same way #158's `.bib` import path does, but keeps them distinct in their own prefix so the TUI can render the two kinds in separate sections. - New `PAPER_NOTE_SENTENCE_ID_PREFIX = "paper-note:"` constant + `paper_note_sentence_id(paper_id)` helper. Disjoint from IMPORTED_SENTENCE_ID_PREFIX by construction (the prefix strings share no common prefix). - New `Anchor::is_paper_note()` predicate, mutually exclusive with `is_imported_synthetic` on a well-formed anchor. - `resolve_anchor` short-circuits paper-notes to Ok before any selector logic runs, mirroring the import-synthetic short-circuit. Tests: sentinel disjointness, stability per paper_id, predicate positives + rejections (quote / range), DB resolver short-circuit under non-matching body text. Refs #185. [tape-exempt: data-model + resolver only; no UI surface yet]
Quote-less paper-level commentary as an MCP tool and a TUI helper. Both surfaces produce annotations whose anchor carries only the `paper-note:<paper_id>` sentinel from C1; the resolver short-circuits them to AnchorStatus::Ok. - New tool `create_paper_note(paper_id, note, author)` on the MCP server. Verifies the paper exists at the boundary so a typo can't create a dangling note. Empty note rejected. Emits a `Created` event on the broadcast channel like other write tools. - `DataStore::create_paper_note` mirror for the TUI side (wired up in C3). Tests: - DataStore: persists with sentinel anchor; co-exists with anchored annotations on the same paper. - Server e2e: `create_paper_note` through the MCP tool surface persists with `is_paper_note()` and emits a Created event. - Server e2e: rejects missing paper + whitespace-only note. Test isolation: the three SCITADEL_DB-mutating tests now share an `ENV_LOCK` mutex; each holds it for the sync setup + write phase and drops before the .await on the broadcast channel. Sound because the broadcast Sender has already been populated for *this* server instance — a parallel test taking the lock next can't disturb the in-flight `rx.recv` resolution. Refs #185. [tape-exempt: MCP tool + DataStore wrapper; TUI rendering lands in C3]
Closes the user-facing half of #185 PR4. The reader-mode right pane now partitions annotations into two sections: - "── Paper-level notes ──" (above): unanchored commentary, drawn with a `¶` glyph. Replies to a paper-level note still render as children with the same threading semantics. - "── Threaded annotations ──" (below): the existing quote-anchored threads, unchanged. The `P` key in the paper-detail overlay opens a single-buffer paper-note prompt (no quote stage). Submission dispatches to `data.create_paper_note` which persists with the `paper-note:<paper_id>` sentinel from C1; the resolver short-circuits on next paper-open. New `AnnotationPrompt::PaperNote` variant is fully covered by state-machine tests (submit, empty cancel, backspace). VHS tape `tui-paper-note.tape` exercises the full flow: pre-existing paper-note renders in its own section, `P` opens the prompt, the new note appears alongside the first one. Added to the `tui-reader` shard with shard-coverage gate updated to match. Help text updated: "n: new | P: paper-note | J: focus". Refs #185.
…PR4 review) Two non-blocking nits from the review, both addressed: - Extract `paper_note_anchor(paper_id) -> Anchor` to scitadel-core so the MCP tool and the TUI DataStore wrapper share a single source of truth for paper-note anchor construction. Both paths now call the helper, which guarantees byte-identical anchors and prevents future drift between the two surfaces and the resolver. - Remove the now-stale `#[allow(dead_code)]` on `DataStore::create_paper_note` — the C3 commit already wired up the `P` keybind, the suppress is no longer needed. Refs #185. [tape-exempt: refactor + test only; no UI surface change]
…PR4) PR191 hit the 10m timeout in the tui-reader shard. The render step itself succeeded; the distinctness assertion then triggered the retry loop on tui-export-keybind.tape (historically the flakiest) and ran out the clock with 5 tapes in the shard. Two changes: - Per-shard `timeout-minutes` now expression-driven via the matrix entry (`matrix.shard.timeout-minutes || 10`), so a single shard can override the default without bumping every shard. - tui-reader → 14m. Empirically: 5 tapes × ~70s render + cargo build (3m) + retry loop on the flaky tape (3 × 70s in the worst case) ≈ 13m. 14m gives margin for the per-render overhead. The other 5 shards keep the 10m default. Parallel total wall-clock stays under 14m. Refs #185, #187. [tape-exempt: CI-only workflow change; no TUI/CLI surface touched]
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
P1 of #185. Adds quote-less paper-level commentary on a publication as a whole, surfaced through both MCP and TUI:
paper-note:<paper_id>sentinel anchor + resolver short-circuit (mirrors feat(annotations): synthetic sentence_id for unanchored imported notes #158's import-synthetic pattern in its own namespace, disjoint by construction).create_paper_note(paper_id, note, author)— verifies paper exists, rejects empty author/note. EmitsAnnotationEventKind::Createdon the broadcast channel like other writes.¶glyph) and "── Threaded annotations ──" (existing) sections.Pkeybind in the paper-detail overlay opens a single-buffer paper-note prompt.paper_note_anchor(paper_id)helper in scitadel-core so the MCP tool and the TUI DataStore wrapper produce byte-identical anchors.Commits
Test plan
Out of scope (follow-up)
Refs #185.