lesereisen-reader-spec.html — Issue #752
LR-0 type selector on /geschichten/new
LR-1 REISE badge on the list
LR-2 Journey reader (ordered cards, interlude asides, no position numbers)
lesereisen-editor-spec.html — Issue #753
LE-1 empty JourneyEditor layout
LE-2 editor with mixed items (documents + interludes, drag handles)
LE-3 inline note-editing state
LE-4 mobile layout
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Task 1: Create standalone FastAPI service scaffold with models, test framework,
and documentation. Includes ParseRequest, ParseResponse Pydantic models matching
OllamaExtraction contract, plus three passing tests validating model validation.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Drop unused MAX_CANDIDATES constant (not referenced in service)
- Keep detached-entity safety comment in resolveTags()
- Add 3 new partial-name match tests (23a/b/c) from #763
- Use resolveByName() API in test 28 (replaces findByDisplayNameContaining)
- Add NameMatches glossary entry from #763
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Assert that when the same person id is returned by two different token
fetches, the person appears exactly once in the result -- pinning
fetchPool's putIfAbsent dedup so a future refactor can't silently
double-classify a candidate.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
AC#4 (maiden alias -> direct) and AC#5 (alias first name -> fetchable +
classifiable) were each split across PersonRepositoryTest (the fetch) and
PersonServiceTest (the classifier with stubs) -- nothing walked
searchByName -> resolveByName end-to-end on real Postgres. Add two tests
in the existing @DataJpaTest slice that build a real PersonService over
the autowired repositories, persist a person with a MAIDEN_NAME alias and
one with an alias firstName, and assert both classify as direct.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The trigger hardcoded the multiple-people label for every count, so a
single did-you-mean picker announced "Mehrere Personen gefunden" to
screen readers while sighted users saw one name and a "Meintest du …?"
heading. Derive the trigger's accessible name from persons.length: a
single suggestion reuses the heading prop, two or more keep the
multiple-people label. Visible truncated name span unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
GLOSSARY entry for NameMatches (direct vs partial name-match strength and how
the search layer maps it); person/README adds resolveByName to the public
surface. No ADR — the matching rule is localized and justified inline.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A 1-item picker now reads "Meintest du …?" (a single direct match auto-selects
and never reaches the picker), while ≥2 keeps the "Person auswählen" framing.
The prompt lives in a visible, non-truncated panel heading (the trigger span
clips at 320px), and the "(auswählen…)" cue is dropped for the 1-item case.
DisambiguationPicker takes heading + showCue props; the page derives both from
ambiguousPersons.length. New search_disambiguation_did_you_mean key in de/en/es.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
resolveNames now delegates to PersonService.resolveByName and maps by match
strength: 1 direct → resolved (auto-select), ≥2 direct → ambiguous, 0 direct
with partials → ambiguous suggestions, 0 candidates → folded into full-text.
A single direct match no longer forces the picker when looser substring hits
coexist. The MAX_CANDIDATES cap moved into PersonService (after classification);
the MAX_NAME_LENGTH guard, resolved-cap overflow, and sender/receiver mapping
are preserved.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Token-set containment over all of a person's name components (firstName,
lastName, alias, each PersonNameAlias first+last, title) decides direct vs
partial. Orchestrates tokenize → cap(8) → fetch pool → classify → cap(10)
after classification, with an empty-token guard and a PII-free debug log of
the outcome bucket. MAX_TOKENS is a DoS control; the after-classify cap keeps a
direct match that sorts past position 10 among partials. Read-only transaction
keeps lazy nameAliases reachable during classification (ADR-022).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The direct-match classifier accepts alias firstName tokens, so the fetch must
surface candidates matchable only via an alias first name. Add a.firstName to
the searchByName LIKE clause (reuses the bound :query — injection-proof). The
person_name_aliases.first_name column already exists; no migration.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Lowercase, split on whitespace/hyphen/apostrophe, drop empties. Applied
symmetrically to query and candidate name components so "Anna-Maria" and
"Anna Maria" tokenize alike. Foundation for resolveByName direct matching.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
When tagsApplied is true, each resolvedTag renders as a 'Thema: Name'
chip with optional inline color style from the tag's resolved color.
Clicking × calls onRemoveChip('theme', tag.name).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Manual update since Docker compose backend runs old build; regenerate with
npm run generate:api once new backend is deployed.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Verifies the recursive CTE in findDescendantIdsByName expands a parent tag
to include all child IDs, and that findByNameContainingIgnoreCase matches
both parent and child names when the fragment appears in both.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Covers multi-tag match, no-match FTS fallback, mixed resolution, personRole
bypass, cap at 10, short-keyword skip, dedup, rawQuery suppression when all
keywords resolve, flag independence, colour propagation via resolveEffectiveColors,
and colour=null when depth constraint prevents resolution.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Keywords that substring-match the tag taxonomy become OR-union tag filters;
non-matching keywords stay as FTS text. Resolved tags surface in the
NlQueryInterpretation as TagHint objects with effective colours. The
rawQuery fallback is now guarded by hadStructuredMatch to prevent
double-apply when all keywords resolve.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Pre-implementation step for #743: ChipType union extracted from
InterpretationChipRow and +page.svelte into shared chip-types.ts;
resolvedTags/tagsApplied neutral defaults added to test fixtures.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Positional record fields added; all 3 construction sites updated with neutral
defaults; NlQueryParserService wired for TagService (4th constructor arg);
NlQueryParserServiceTest and NlSearchControllerTest synced.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Addresses @Sara review: browser tests in this spec fail silently when
the project path contains '+' (common in git worktrees). The comment
tells developers to copy the frontend directory to a clean path.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Addresses @Markus review: tags fetched by findByNameContaining live outside
any transaction; Hibernate's dirty-check never fires on them. The comment
removes the ambiguity for cold readers.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
When tagsApplied is true, each resolvedTag renders as a 'Thema: Name'
chip with optional inline color style from the tag's resolved color.
Clicking × calls onRemoveChip('theme', tag.name).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Manual update since Docker compose backend runs old build; regenerate with
npm run generate:api once new backend is deployed.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Verifies the recursive CTE in findDescendantIdsByName expands a parent tag
to include all child IDs, and that findByNameContainingIgnoreCase matches
both parent and child names when the fragment appears in both.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Covers multi-tag match, no-match FTS fallback, mixed resolution, personRole
bypass, cap at 10, short-keyword skip, dedup, rawQuery suppression when all
keywords resolve, flag independence, colour propagation via resolveEffectiveColors,
and colour=null when depth constraint prevents resolution.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Keywords that substring-match the tag taxonomy become OR-union tag filters;
non-matching keywords stay as FTS text. Resolved tags surface in the
NlQueryInterpretation as TagHint objects with effective colours. The
rawQuery fallback is now guarded by hadStructuredMatch to prevent
double-apply when all keywords resolve.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Pre-implementation step for #743: ChipType union extracted from
InterpretationChipRow and +page.svelte into shared chip-types.ts;
resolvedTags/tagsApplied neutral defaults added to test fixtures.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Positional record fields added; all 3 construction sites updated with neutral
defaults; NlQueryParserService wired for TagService (4th constructor arg);
NlQueryParserServiceTest and NlSearchControllerTest synced.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Capture the why behind deploying Ollama to prod/staging compose: the
corrected init recipe (supersedes ADR-028 §10's never-functional curl
loop), the OLLAMA_KEEP_ALIVE=-1 pin (so a future maintainer doesn't
optimize it away and reintroduce the post-idle cold-load 503), the
30->60s timeout NFR, and the memswap==mem hard-OOM trade-off.
Addresses #759 review (Markus #3, Nora #2).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The diagram declared Container(ollama, ...) twice — an alias collision that
renders a duplicate box. It also declared the backend->ollama relationship
twice. Keep the richer 'Ollama LLM Service' declaration and the more
specific 'NL query parsing (POST /api/generate)' relationship; drop the
duplicates.
Addresses #759 review (Markus #2).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Mirror the prod hardening in the dev stack: guard the model pull with
`ollama list | grep -q <model>` so an already-cached model exits clean
without a registry round-trip. Keeps dev and prod on one recipe.
Addresses #759 review (Tobias #1).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The init command unconditionally ran `ollama pull`, which contacts the
registry to verify the manifest digest even when the model is already on
the volume. A host reboot during a registry/upstream-network blip would
then fail init non-zero, the `service_completed_successfully` gate would
never be met, and the ollama service (hence NL search) would stay down
until the registry was reachable again.
Guard the pull with `ollama list | grep -q <model>` so a cached model
exits clean without any registry round-trip.
Addresses #759 review (Tobias #1).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
docker-compose.prod.yml declares the volume as `ollama-models` (hyphen),
so the compose-project-prefixed name is `archiv-production_ollama-models`,
not the underscored `archiv-production_ollama_models` the model-upgrade
guide documented. The documented `docker volume rm` would not have matched
the real volume.
Addresses #759 review (Tobias #2).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
application.yaml sets app.ollama.timeout-seconds: 60 (raised from 30 to
absorb the cold model load on the first query after an Ollama restart),
but DEPLOYMENT.md still documented 30. A doc that contradicts the shipped
value is a traceability defect.
Addresses #759 review (Markus, Felix, Elicit).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
NL search recovered after deploy but went 503 again after a few minutes:
Ollama unloads the model after its default ~5 min keep-alive, so the next
query cold-loads the 4.7 GB model and exceeds the backend's 30s read
timeout (ResourceAccessException -> SMART_SEARCH_UNAVAILABLE). Warm
inference is ~18s; the cold load after idle is what timed out.
- docker-compose.{prod,yml}: set OLLAMA_KEEP_ALIVE=-1 on the ollama
service so the model stays resident and never pays a cold-load penalty
during normal operation (verified on staging: `ollama ps` -> UNTIL
"Forever"; host has 47 GB free).
- application.yaml: raise app.ollama.timeout-seconds 30 -> 60 so the one
unavoidable cold load (first query after an Ollama restart, before the
model is pinned) completes instead of timing out.
Refs #758
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
NL search returned 503 (SMART_SEARCH_UNAVAILABLE / "Intelligente Suche
nicht verfügbar") on staging because Ollama was never reachable.
Two defects, both downstream of #737:
1. Ollama was added only to the dev docker-compose.yml. Staging/prod
deploy from the self-contained docker-compose.prod.yml, which had no
ollama service — so the backend (defaulting to http://ollama:11434)
hit a non-existent host (ResourceAccessException -> 503).
2. The merged model-init recipe never worked: the ollama/ollama image
ENTRYPOINT is `ollama` (so `command: sh -c ...` ran as `ollama sh ...`
-> "unknown command sh"), and the image ships no curl (so both the
readiness loop and the healthcheck could never pass).
- docker-compose.prod.yml: add ollama-model-init + ollama services and
the ollama-models volume, with the corrected recipe (entrypoint
override to /bin/sh -c, `ollama list` for readiness and healthcheck).
- docker-compose.yml: fix the same broken entrypoint/command and the
curl healthcheck so the dev stack actually starts Ollama.
Verified on staging end-to-end: model-init exits 0, ollama healthy,
backend reaches /api/tags, inference succeeds within the 8g limit.
Refs #758
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Markus (architect): document SearchFilterBar + the search/ components
(SmartModeToggle, InterpretationChipRow, SmartSearchStatus,
DisambiguationPicker) and the POST /api/search/nl relation.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
