Address claude-review round 4 on PR #868: tighten the assign_page_numbers
guard from `page_boundaries is not None` to a truthy check, so a PDF with an
empty boundary list no longer enters the trace span and fires the alarming
"NO page numbers assigned" warning for a harmless no-op.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Address claude-review round 2 on PR #868:
- Extract the use_page_aware branching into a pure `should_use_page_aware`
helper and cover the (doc_type, page_boundaries, page_aware_setting) matrix
in tests/unit/test_processor_routing.py (file+boundaries+enabled, empty
list, None, non-file doc types, disabled setting).
- Clarify the PageAwareChunker.chunk_text no-boundaries comment: the processor
pre-filters via should_use_page_aware, so that branch is a direct-call safety
net, not a production indexing path.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Address claude-review round 1 on PR #868:
- use_page_aware now gates on `bool(page_boundaries)` instead of
`is not None`, so a PDF that yields an empty boundary list takes the
char-based path explicitly (assign_page_numbers no-ops on []) rather than
the page-aware chunker's no-boundaries fallback. Same result, clearer intent.
- add test_oversized_page_with_leading_whitespace_offsets, exercising the
start+start_index offset path for an oversized page whose sub-chunks have
leading whitespace.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add PageAwareChunker, which splits paginated documents (PDFs) on page
boundaries first and only character-splits pages larger than chunk_size.
No chunk spans a page boundary, so page_number is always exact and stored
excerpts never lead with a neighbouring page's text. When chunk_size is at
least the largest page, this yields exactly one chunk per page: a
predictable vector count (== page count), a flat per-page embedding cost,
and zero cross-page overlap duplication.
Gated by DOCUMENT_CHUNK_PAGE_AWARE (default true). When false, the legacy
char-based DocumentChunker + post-hoc assign_page_numbers path runs
unchanged. Only doc_type="file" with page_boundaries (PDFs) takes the
page-aware path; notes/deck/news are unaffected.
Measured on a 15-page record (query "leadership award louis", target =
top-half of page 15): char-based degraded the target to dense-rank 10 at
cs=2048 (OCR) and mislabeled its page; page-aware restored rank 1 across
every fusion/modality and chunk size, with correct page labels and clean
snippets.
BREAKING CHANGE: PDFs are re-chunked page-aware by default. Existing
deployments will re-index PDF content on the next vector sync (different
chunk counts and page_number labels). Set DOCUMENT_CHUNK_PAGE_AWARE=false
to retain the previous char-based behaviour.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Tagging an existing file/folder emits only OCP\SystemTag\MapperEvent — never a
Node*Event — so tagged PDFs were previously only picked up by the hourly
scanner. Subscribe to the tag event and reconcile membership so adding/removing
the `vector-index` tag (re)indexes in near-real time.
- webhook_presets: add OCP\SystemTag\MapperEvent to the files_sync preset
(NC 32+, where MapperEvent gained getWebhookSerializable(); harmless on older
servers — it just never fires).
- webhook_parser: parse MapperEvent (objectType=files) into a path-less file
"reconcile" task. The payload carries only a fileid + tagIds (no name/path),
so assign and unassign both collapse to a reconcile.
- processor._reconcile_tag_event: resolve the fileid against the user's current
vector-index PDFs (find_files_by_tag). Present -> index with the resolved
path/etag; absent -> flip to delete. Naturally handles "an unrelated tag
changed" and a tagged folder's own fileid (no-op; the scanner still expands
folders to descendants).
- Unit tests for the parser branch and the reconcile.
The matching admin-UI preset change ships separately in the astrolabe app repo.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Address PR #858 review round 3:
- document_tier1_engine / document_ocr_provider now validate + normalize via
Settings.__post_init__ _enum_fields (the repo's canonical opt-in-enum pattern;
case-insensitive) instead of dynaconf Validators. A typo now raises ValueError
at load and "Gateway" normalizes to "gateway".
- classify_from_text gates no_text_layer/bad_text_layer on ocr_frac >=
OCR_PAGE_FRACTION, matching classify_pdf -- a "fast"-routed doc with a few junk
pages no longer emits a misleading flag (keeps the shadow vs hot-path
classification metrics consistent).
- build_ocr_backend warns when an EXPLICIT provider is misconfigured
(gateway without EMBEDDING_GATEWAY_URL, mistral without MISTRAL_API_KEY)
instead of silently returning None.
- Pypdfium2FastProcessor.health_check probes the import; documented why
OcrProcessor.health_check is unconditionally True (lazy per-tenant backends).
- Removed the leftover per-boundary / per-chunk debug logging loops.
Tests: enum normalization + rejection for the two new settings.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Address PR #858 review:
- 🔴 OcrProcessor now resolves its backend once and reuses it. Rebuilding per
call created a fresh GatewayTokenProvider each time -- discarding its M2M-token
cache, so every OCR'd document fetched a new token -- and a new Mistral client.
- 🔴 build_ocr_backend uses explicit ValueError (not assert, which is stripped
under `python -O`) for the gateway M2M triple.
- PIPELINE_TIER in the Qdrant payload now reflects the tier that actually
produced the doc: the registry stamps result.metadata["pipeline_tier"] and the
processor reads it (was hardcoded "fast", wrong for OCR/structured).
- Escalation now requires classification.page_count > 0, so a zero-page
(empty/corrupt) PDF isn't pointlessly sent to OCR; documented that a fast
FAILURE (encrypted/unopenable) is a hard failure and is not OCR-escalated.
- Documented the OCR page_boundaries separator-attribution choice.
- Downgraded the per-document page-boundary / page-assignment INFO logs to debug.
New tests: zero-page no-escalation, pipeline_tier stamping.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Replaces single-engine pymupdf4llm extraction with a tiered pipeline (Deck #205,
follows the tier-0 classifier #855). pypdfium2 becomes the default and only
hot-path PDF extractor; pymupdf4llm is deprecated to a rollback toggle.
Why: pymupdf4llm's O(n^2) find_tables drove the OOM (#852) and the form-PDF
parse timeouts (#856), carries AGPL/commercial licensing liability, and -- per
the benchmarks -- recovers near-zero usable tables on the real corpus. pypdfium2
(Apache/BSD) extracts the same text far faster (Student 1a.pdf: 120s timeout ->
0.2s) with no table-detection bomb.
- document_processors/pypdfium2_fast.py: tier-1 "fast" processor emitting text +
exact page_boundaries (the pdf_highlighter contract). pymupdf processor is now
tier "structured" (the rollback engine), registered but not default.
- registry: tiered routing in ProcessorRegistry. tier-1 fast extracts, then
classification is DERIVED from that text (classifier.classify_from_text -- no
PDF re-open), records the classification metrics, and escalates scanned /
no-text-layer docs to the "ocr" tier when document_ocr_enabled (default off;
no provider yet, so fast is terminal). Wires record_document_escalation + the
real "escalated" span attribute (was hardcoded False).
- Removes the separate _shadow_classify pass from vector/processor.py -- it
re-opened every PDF and re-extracted text (~0.5-1.3s/doc of pure duplicated
CPU that lowered throughput); classification now rides the tier-1 extraction.
- Settings: document_tier1_engine ("pypdfium2" default | "pymupdf" rollback,
enum-validated), document_ocr_enabled (default false).
Tests: pypdfium2 extractor, registry tiering (fast routing, rollback, classify
recording, OCR escalation on/off), classify_from_text. Full unit suite green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The vector-sync pipeline derived an indexed file's display title from the
document's embedded metadata (e.g. a PDF's /Title), falling back to the
filename only when absent. That embedded title frequently disagrees with how
the user named the file in Nextcloud and is confusing in the astrolabe
vector-viz UI (a passive consumer of the `title` payload field).
For files, always derive the title from the Nextcloud filename via a shared
`file_title_from_path` helper. Notes/deck/news keep their metadata titles.
A rename/move in Nextcloud keeps the fileid (doc_id) and content (etag/mtime)
but changes the path, so both the dedup claim and the scanner freshness gate
skip re-embedding and the stored file_path/title go stale. Add
`reconcile_document_path`: a metadata-only set_payload that refreshes
file_path + title on the existing real chunks without re-fetch/re-embed.
Wire it into both skip paths:
- dedup hit (etag unchanged on rename) via claim_existing_index(current_path=...)
- scanner incremental skip (etag changed, mtime stable)
Both reuse already-fetched payloads, so steady-state scans add no extra
round-trip (reconcile is a no-op when the path is unchanged).
Refs: Deck #204
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Address PR #855 round 2:
- 🔴 _shadow_classify swallowed all exceptions at DEBUG, so a systematic
failure (pymupdf bug, memory pressure) is invisible at LOG_LEVEL=INFO and
trips SonarQube S2221/S5754. Log at WARNING instead (still best-effort --
indexing is unaffected).
- classifier: use `with pymupdf.open(...) as doc` instead of manual try/finally.
- tests: release the Pixmap's native memory (del pix) in the image fixtures.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
First step of the tiered document-processor effort (Deck #203): a cheap, local
pre-pass that recommends which extraction tier a PDF should start in, emitting
metrics WITHOUT changing routing yet -- so we gather per-tenant doc-mix data
before turning escalation on.
document_processors/classifier.py: classify_pdf(content) -> DocClassification.
Page-sampled (bounded on large docs), <~1s. Cheap signals only -- text-layer
chars, a text-quality score (catches the "Student 147" failure where a text
layer exists but is mashed/space-less junk), and image coverage. A page that is
mostly a raster image routes to OCR: its content (handwriting, stamps) isn't in
any text layer. Deliberately no get_drawings/graphics-density signal -- it's
slow on the exact pages it'd flag, the hotfix's graphics_limit already makes the
parse safe, and the (future) tier-1 quality gate catches lost tables.
Validated on the sample corpus: born-digital 2-col arxiv and a digital student
record -> fast (tier 1); a scanned+handwritten form -> ocr (tier 3).
Wiring (vector/processor.py): _shadow_classify runs the classifier on PDFs in a
worker thread, best-effort (never blocks/fails indexing), gated by the new
DOCUMENT_CLASSIFY_ENABLED setting. Metrics: astrolabe_document_classified_total
{recommended_tier}, astrolabe_document_classifier_flag_total{flag},
astrolabe_document_text_quality histogram.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Address PR #852 round 3 (all 🟡, no blockers):
- config: DOCUMENT_PARSE_TIMEOUT_SECONDS is now float (default 120.0) so a
fractional value is honoured rather than silently stored in an int field;
matches anyio.move_on_after's float seconds.
- _isolation: comment that a clean rlimit MemoryError leaves the worker alive
in anyio's pool (vs the SIGKILL/BrokenWorkerProcess path that respawns) --
acceptable since RLIMIT_AS caps virtual address space, not RSS.
- processor: note the `if indexed is False` is a deliberate identity check --
a successful index (incl. dedup hit) returns None and must not be mistaken
for a parse failure.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Address PR #852 round 2:
- config: DOCUMENT_PDF_GRAPHICS_LIMIT validator is now gte=1 (pymupdf4llm treats
0 as "no cap", which would re-expose the OOM); documented the zero semantics
and that the per-worker mem rlimit needs a pod restart to change.
- processor: annotate `_index_document -> bool | None` and document the contract
so the `if indexed is False` check is explicit/type-checkable.
- tests: add the RLIM_INFINITY-hard branch assertion for _apply_mem_limit
(soft==target, hard stays unbounded).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Address PR #852 review:
- pymupdf.py: the metadata `doc` was only closed on the PdfParseFailed and
success paths, so a failure in `_extract_metadata`/`mkdir`/`get_settings`
leaked it. `doc` is only needed for metadata + page_count (the heavy parse
works from `content` bytes in the worker), so open it, read metadata, and
close it immediately under try/finally; drop the two later doc.close() calls.
- processor.py: a permanent parse failure early-returned from `_index_document`,
after which `process_document` still recorded record_qdrant_operation("upsert",
"success") + record_vector_sync_processing(success) -- counting an OOM/timeout
bomb as astrolabe_documents_indexed_total{status="success"}. `_index_document`
now returns False on that path and the caller skips the success metrics (the
failure is already recorded via document_parse_failed_total + the registry's
document_parse_total{error}).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The document processor crash-looped on one pathological PDF: pymupdf4llm's
table/graphics detection over a page with ~1M vector path items ballooned past
the 2 GiB pod limit. The parse ran in a thread, so nothing could interrupt or
memory-bound it -- a single bad file OOM-killed the whole pod.
Run the parse in an isolated worker subprocess (anyio.to_process, cancellable)
with an RLIMIT_AS memory cap and a wall-clock timeout, so a pathological file
fails THAT document instead of the pod (new document_processors/_isolation.py).
Also pass graphics_limit (default 5000) to to_markdown -- validated to cut the
known trigger page from 112 s to 23 s with bounded memory.
On a permanent parse failure the processor returns success=False (instead of
raising, which would retry 3x); vector/processor.py marks the placeholder
"failed" and skips indexing, and the scanner stops re-queuing failed placeholders
until the file changes -- so a doomed file no longer churns.
New per-tenant (per-pod env) settings: DOCUMENT_PDF_GRAPHICS_LIMIT,
DOCUMENT_PARSE_TIMEOUT_SECONDS, DOCUMENT_PARSE_MEM_LIMIT_MB. New metric
astrolabe_document_parse_failed_total{reason=timeout|oom|error} surfaces hard
failures that previously killed the process before any except ran.
First PR of the tiered document-processor effort (Deck #199); tier 0/1/3
pipeline tracked separately.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Starlette lifespan started `vector_sync_metrics_task` with undefined
names `task_producer` and `receive_stream`. Those locals only exist inside
the `_wire_vector_sync_state` helper; in the lifespan the transport is bound
as `ingest_transport`. The undefined reference raised `NameError`, which
aborted the background-sync task group and crashed startup in every
deployment mode ("Application startup failed. Exiting.").
Introduced by fbe70ecd ("feat: backend-agnostic vector-sync gauges").
Pass `ingest_transport.producer` / `ingest_transport.receive_stream` at both
call sites (single-user app.py:1791, OAuth/login-flow app.py:2012).
Also fix 10 pre-existing `ty` possibly-missing-attribute diagnostics: the
deck indexing code in scanner.py, processor.py and search/context.py reads
full-DeckCard-only fields (description, type, owner, etag, lastModified) off
`stack.cards`, typed `list[DeckCard | DeckCardSummary]`. Freshly-fetched
stacks from `get_stacks()` always hold full DeckCards (the summary
projection only happens in the tool layer), so narrow with
`cast(list[DeckCard], ...)` — matching the existing pattern in
server/deck.py.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
existing_principals() ran for every doc type when seeding acl_principals.
note/news_item/deck_card IDs are per-user (not globally unique) and chunk
point IDs are user-agnostic, so on an ID collision the merge would pull in
another user's principal and cross-surface their content via the
acl_principals search branch. It was also N wasted tenant-wide scrolls on
initial sync for those types. Gate the prior-principal merge on
doc_type == "file" (the only type with cross-user dedup + globally-unique
fileid); other types seed with the indexer only.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A file shared across many users — directly, or via a group folder shared
to a group — was parsed and embedded once per user. Chunk point IDs are
user-agnostic (uuid5(tenant_id, doc_id=fileid, chunk_index)), but the
per-user freshness gate filtered Qdrant by user_id, so two readers
ping-ponged: each overwrote the other's points and each kept seeing "not
indexed for me", reprocessing every scan. Production telemetry (note
386945, finding #5) measured identical docs re-processed every few hours
at 7-13s each, with PDF parse ~62% of per-doc cost.
Layer 1 — tenant-wide dedup:
- Thread the scanner's tag-REPORT etag into the file DocumentTask and the
chunk payload; index `etag` as a KEYWORD field.
- vector/sharing_state.find_indexed_content scrolls tenant-wide (no
user_id filter) for a non-placeholder point matching
(doc_id, doc_type, etag), gated on embedding_identity in Python so a
model switch correctly forces a re-embed.
- Scanner skips enqueue and the processor skips fetch/parse/embed when a
match exists (cross-worker race-guard before WebDAV read). Dedup is
fail-safe: a Qdrant error degrades to "process normally".
Layer 2 — observed-access ACL (no admin / GroupFolders API needed):
- Each point carries `acl_principals` = the set of user:<uid> whose
scanner has observed (hence can read) the file. The per-user tag REPORT
is the access oracle; group membership/GroupFolders enumeration is
admin-only and unavailable in multi-user modes.
- build_ownership_filter ORs MatchAny(acl_principals, ["user:<me>"]) so a
deduplicated shared/group-folder point surfaces to every reader;
verify-on-read (_verify_files) remains the precise ACL gate.
- Deletion/eviction become "release one user": drop the principal and
delete the points only when the set empties, so one user untagging a
shared file doesn't evict it for the others. Legacy points without the
field keep the original per-user delete.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Re-architect document ingest from the shared NATS-glued document-processor to a
per-tenant, in-process model owned by nextcloud-mcp-server (Deck #183). The MCP
server now owns both sides of ingest:
- Producer (api role): the scanner defers one job per changed document into the
app's Postgres via procrastinate (queueing_lock dedup; no execution lock, so a
crashed worker can't deadlock a doc — Qdrant upserts are idempotent).
- Consumer (worker role): `nextcloud-mcp-server worker` drains the queue and runs
the existing process_document pipeline; a periodic task reclaims jobs orphaned
in `doing` by a crash.
INGEST_QUEUE selects the transport (auto: postgres when DATABASE_URL is Postgres,
else the in-process anyio queue for SQLite/dev). procrastinate manages its own
tables (applied on a fresh DB at startup and by `db upgrade`). The vector-sync
status surface reads job counts from Postgres in postgres mode. procrastinate +
psycopg3 ship in the [postgres] extra; the app's own engine still uses asyncpg
(driver unification is a follow-up handled in the rendered Helm chart).
NATS JetStream, the Postgres-queue stub, the bus status subscriber, and nats-py
are removed.
BREAKING CHANGE: the external-NATS-ingest env vars are removed
(INGEST_MODE, STATUS_BACKEND, INGEST_BUS_URL, INGEST_BUS_NUM_REPLICAS,
FACT_EVENT_EMITTER). Use INGEST_QUEUE (memory|postgres) and the `worker`
command instead. TENANT_ID is retained (no longer NATS-subject-charset-validated).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Remaining items from the PR #831 Claude review:
- processor span symmetry: add "vector_sync.total_chars" to the sparse
embedding span (already on the dense span) and drop the redundant
"embedding.batch_size" attribute from both spans — it always equalled
vector_sync.chunk_count and would mislead once batching is split.
- metrics: document the deliberate "throughput counts only on full success"
contract in record_document_parse (partial extractions flagged
success=False are counted as a parse-error but never inflate
pages/chars/bytes throughput).
- config: extract _detect_base_provider() -> (family, model) as the single
source of truth for the provider-detection priority chain, shared by
get_embedding_model_name() and get_embedding_provider_family(). Preserves
the intentional gateway asymmetry (only the family method short-circuits).
- base.py: Optional[...] -> PEP 604 `... | None`; drop now-unused import.
Behavior unchanged (get_embedding_* outputs covered by test_config.py).
Refs Deck #175, PR #831.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Failed deletes no longer bump astrolabe_documents_indexed_total: the outer
except in process_document now gates doc_type on operation != "delete", so a
delete error is counted as processed-error but not as an indexing event.
Added test_failed_delete_is_processed_but_not_indexed.
- registry parse span: pass record_exception=True explicitly (matches
instrument_tool) and add a structured logger.warning on the parse-error path
(processor/tier/byte_size/duration_ms) for a Loki-aggregatable failed-parse
signal.
- test_error_does_not_increment_throughput: snapshot-before/delta pattern
instead of absolute 0.0 (counters are global singletons).
- config: document the deliberate gateway asymmetry between
get_embedding_model_name() (no gateway branch) and
get_embedding_provider_family() (short-circuits on gateway).
- Cleanup in touched scope: narrow `except (HTTPStatusError, Exception)` to
`except Exception` (drop now-unused import); convert registry signatures from
Optional[...] to `... | None`.
Refs Deck #175, PR #831.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Reviewer findings:
- Fix double-count of exhausted-retry failures: the inner final-retry branch
and the outer except both recorded a processing error. Consolidate to the
outer handler (single call site); inner branch keeps only the Qdrant-upsert
error metric. Regression test added.
- Deletes are no longer counted as indexing events: the delete success path
drops doc_type so astrolabe_documents_indexed_total is not inflated.
Regression test added.
- Reuse the already-resolved `settings` in _index_document instead of a second
get_settings() call.
- Use explicit `> 0` guards in record_document_parse / record_embedding instead
of truthiness checks.
SonarCloud:
- S1244 (BUG): replace float `==` equality in metric tests with pytest.approx.
- S5332 (hotspot): use https in the gateway-URL test fixture.
- S1192: extract the repeated "vector_sync.chunk_count" span-attribute literal
into a module constant.
Review nit: move the duplicated `_sample` test helper into a shared
`metric_sample` fixture in tests/unit/conftest.py.
Refs Deck #175, PR #831.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Make per-tier bottlenecks in the document-processing pipeline
(scan -> fetch -> parse -> chunk -> embed -> Qdrant upsert) visible via
metrics, traces, and structured logs. Today the document_processors layer
emits only a logger.info line: no metric, no span, and page counts live only
inside a log string. The single processing-duration histogram is unlabeled and
whole-document, so it cannot isolate parse vs embed vs upsert.
New astrolabe_* metric family (distinct from the mcp_* protocol metrics):
- astrolabe_document_parse_{duration_seconds,total} + pages/chars/bytes counters
recorded at the ProcessorRegistry.process() boundary (covers all current and
future processors uniformly)
- astrolabe_document_escalation_total (dormant; tiered-pipeline readiness)
- astrolabe_embedding_{duration_seconds,requests_total,chunks_total,chars_total}
- astrolabe_document_chunks_total, astrolabe_documents_indexed_total{source,status}
Tracing: new document_processor.parse child span + enriched embed/chunk span
attributes (provider/model/batch_size/chunk_count). Structured logs gain a
consistent field vocabulary (doc_id, doc_type, processor, tier, pages, chars,
byte_size, chunks, duration_ms, status) so Loki can aggregate without regex.
Tier-readiness: processor/tier are labels from day one and a tier property is
added to DocumentProcessor, so adding docling/OCR/LLM tiers later is additive
(new label values, never new metrics). Tenant comes from the kube namespace
label; mime_type/model are span attributes only (cardinality). Existing
mcp_vector_sync_*/mcp_qdrant_* are left untouched.
Refs Deck #175 (superset of #173 Phase 2). Dashboard/recording-rules follow-up
tracked on #175 for homelab-argocd.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds the seven §10.2 hook-point modules + five env vars so Astrolabe Cloud can
offload document processing to the external document-processor / embedding
gateway. Purely additive: with every setting unset the server behaves exactly
as today, so self-hosters are unaffected (Deck #92).
Hook points (all default to current monolith behavior):
- config: EMBEDDING_PROVIDER, INGEST_MODE, STATUS_BACKEND,
COLLECTION_METADATA_SOURCE, FACT_EVENT_EMITTER (+ supporting settings),
validated in Settings.__post_init__ (fail-fast STATUS_BACKEND=local with
INGEST_MODE=external); shared canonical.py.
- vector/payload_keys.py + acl_hash.py: cross-impl NAMESPACE/point_id (§2.2)
and BLAKE2b-128 ACL hash (§11), pinned by fixtures shared with the
document-processor repo.
- embedding/gateway_client.py: OpenAI-compatible GatewayProvider authenticating
via M2M OIDC client-credentials (separate realm); manual-only registry entry.
- vector/collection_metadata.py: sentinel-point / API metadata source with env
fallback.
- vector/queue/: hexagonal ingest producer ports + memory/NATS adapters
(Postgres seam); INGEST_MODE=external publishes mcp.ingest.requested.{tenant}
instead of the in-memory stream and skips the in-process processor pool. The
lifespan becomes a composition root across both deployment branches.
- vector/queue/status.py: STATUS_BACKEND=bus subscriber feeding a StatusStore
the vector-sync status endpoint reads.
- admin/payload_backfill.py: POST /api/v1/admin/payload-backfill (admin scope);
processor writes the new payload keys; query-side ACL pre-filter gated behind
ACL_PREFILTER_ENABLED (default off).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The vector index has always been strictly per-user: every Qdrant payload
carries a `user_id` and the search filter is `user_id == querying_user`.
A file Alice indexed cannot be discovered by Bob even if she has shared
it with him — Bob would have to re-index it under his own user_id to
make it searchable, which means duplicate index entries for every share
recipient.
Switch to ownership-with-ACL-expansion:
- New `nextcloud_mcp_server.search.access_filter` module:
- `list_accessible_owners(sharing_client, user_id)` calls the OCS
Sharing API (`shared_with_me=true`) and returns
`{user_id} ∪ {uid_owner of each share}`. Fails open to `[user_id]`
so a misbehaving Sharing API doesn't black-hole search.
- `build_ownership_filter(user_id, accessible_owners)` returns a
Qdrant `Filter` whose `should` branch matches either the new
`owner_id IN accessible_owners` field or the legacy `user_id` field.
The legacy branch keeps points indexed before this change reachable
without a migration backfill.
- Indexer payload (`vector/processor.py`) now writes `owner_id` alongside
`user_id`. `DocumentTask` gains an optional `owner_id` field; today the
scanner always runs as the owner so the processor falls back to
`user_id`, but the field is plumbed so a future shared-with-me crawler
can set the true owner without reshaping the payload contract.
- `SemanticSearchAlgorithm.search` and `BM25HybridSearchAlgorithm.search`
accept `accessible_owners` via kwargs and use the new ownership filter.
Default behaviour with no kwarg is unchanged (self-only).
- Both user-facing callers — the MCP tool path (`server/semantic.py`) and
the visualization Starlette route (`auth/viz_routes.py`) — compute
`accessible_owners` from the authenticated Nextcloud client before
invoking the search algorithm. Eviction, scanner deletion, placeholder,
and chunk-context paths intentionally keep the legacy `user_id`
semantics (those are "operations on a specific user's records", not
cross-user reads).
- 10 new unit tests in `tests/unit/search/test_access_filter.py` cover
self-only default, owner expansion, dedup, fallback fields, OCS
failure, and the legacy `should`-branch shape.
Pairs with cbcoutinho/astrolabe#89 — together they let an Astrolabe user
find content owners have shared with them without going through any
re-authorization flow or re-indexing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The ``BM25SparseEmbeddingProvider.__init__`` calls
``fastembed.SparseTextEmbedding(model_name="Qdrant/bm25")`` which
downloads ~50 MB of model weights from HuggingFace and loads them
into memory — observed >5 s wall-clock in production. The inference
methods (``encode_async``, ``encode_batch_async``) already wrap work
in ``anyio.to_thread.run_sync``, so the design intent is clearly to
keep FastEmbed off the event loop. That protection just didn't
cover the constructor.
Symptom in the Astrolabe Cloud per-tenant deploy (deck #102 smoke):
~30–90 s after a user enables semantic search, the pod tips into a
SIGKILL-restart cycle. Loki shows a single log line
Initializing BM25 sparse embedding provider: Qdrant/bm25
followed by nothing else from the event loop until exitCode 137.
Kubernetes ``/health/live`` httpGet probe timeout=5s fires 6 times
in a row, kubelet kills the container, restart, repeat.
Fix: switch ``get_bm25_service()`` to an async accessor that wraps
the first-time construction in ``anyio.to_thread.run_sync``. Two
existing call sites (``vector/processor.py:603``,
``search/bm25_hybrid.py:123``) update to ``await``. Both are
already inside async functions so the await is free.
New unit test pins the invariant by monkey-patching
``BM25SparseEmbeddingProvider.__init__`` with ``time.sleep(1)`` and
asserting a concurrent ``anyio.sleep(0.05)`` finishes promptly —
the test fails if the constructor ever runs back on the event loop.
Same pattern exists in ``OllamaEmbeddingProvider.__init__`` (sync
``httpx.get`` health-check). Ollama isn't enabled in any current
deploy; filed as a follow-up.
Refs:
- Astrolabe Cloud deck card #102 (smoke discovery)
- Sibling fix#799 (NullPool for cross-loop-asyncpg, same class
of "anyio bites you in production" bug)
Verified:
- ``uv run pytest tests/unit/`` — 1027 passed
- ``uv run ruff check`` clean on touched files
- ``uv run ty check`` clean on touched files
- New tests pass
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- pdf_highlighter.compute_chunk_bboxes_batch: drop unused chunk_text
destructure (SonarQube finding), and replace positional
page_boundaries[page_num - 1] with a key-based next() match so
reordered or non-1-indexed boundaries can't silently shift the bbox.
Convert touched f-string log to lazy %s formatting.
- vector/processor: rename the trace_operation span from
"vector_sync.generate_highlights" to "vector_sync.compute_chunk_bboxes"
to match what the function actually does.
- Add test_compute_chunk_bboxes_handles_unordered_page_boundaries —
reverses the boundaries list and asserts identical results to the
in-order case, guarding the boundary-lookup regression class.
- Pin pre-push-review skill to sonnet model.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- chunk_bboxes is now dict[int, list[tuple[...]]] holding the bbox list
directly, not {"bbox": ..., "page": ...}. The page from text-search was
stored but never read; page_number from offset-based assignment is
authoritative for the Qdrant payload.
- Add two unit tests for the documented omission contract: chunks whose
offsets fall outside every page boundary, and chunks whose text cannot
be located on the rendered page, are silently dropped from the result.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Drop chunk_bbox_page from Qdrant payload — viz endpoints never read it
(page_number is the canonical PDF page field).
- Bump upsert BATCH_SIZE 10 → 100 now that payloads no longer carry PNGs.
- compute_chunk_bboxes_batch: move doc.close() into finally, replace
unused stored_page_num with _.
- purge_page_images.py: switch to anyio.run() per project convention,
and wrap AsyncQdrantClient in try/finally so the aiohttp session is
always closed (the class doesn't implement async-context-manager).
- Decorate new bbox unit tests with @pytest.mark.unit so they run under
the fast-feedback selector.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Per-chunk PDF page renders (~150–700 KB base64 PNG each) were the dominant
disk consumer in production, repeatedly tripping `No space left on device:
WAL buffer size exceeds available disk space` on welcomed-malamute Qdrant.
Replace the inline highlighted_page_image / highlighted_page_number /
highlight_count fields with a small `chunk_bbox` field:
list[(x0, y0, x1, y1)] of normalized [0, 1] floats, ~32 bytes per chunk.
Astrolabe (the only known consumer) renders the highlight client-side as
a percentage-positioned overlay on top of the existing /api/v1/pdf-preview
render-on-demand path (cbcoutinho/astrolabe#76).
- pdf_highlighter: new compute_chunk_bboxes_batch() that reuses the
existing _find_chunk_bbox text-search path, skipping all pixmap/PIL/PNG
work.
- processor: store chunk_bbox + chunk_bbox_page in the Qdrant payload,
drop highlighted_page_image + friends, drop the base64 import.
- visualization /api/v1/chunk-context and auth/viz_routes: read
chunk_bbox instead of highlighted_page_image.
- vector/__init__: stop eagerly re-exporting `processor`/`scanner` —
fixes a pre-existing circular import (search.algorithms ->
vector.placeholder -> vector/__init__ -> processor -> scanner ->
server.semantic -> search.bm25_hybrid -> search.algorithms partial).
Test suite that was broken on master (test_bm25_hybrid.py et al.) now
collects and passes.
- scripts/purge_page_images.py: ad-hoc, idempotent migration that
delete_payload's the legacy keys from existing points. No reindex
required; legacy chunks render the page with no overlay.
Pairs with cbcoutinho/astrolabe#76. Frontend handles missing chunk_bbox
gracefully, so this can land in either order.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes out the remaining nits flagged in the round-6 review.
Critical:
- _verify_files contract comment now enumerates all None-return cases
(404 + malformed PROPFIND XML) and documents the false-eviction
trade-off; self-healing via re-indexing recovers
- int(r.id) cast at the SemanticSearchResult boundary now raises a
TypeError with explicit doc_type/value context instead of bubbling
up as an opaque "Search failed: ..." McpError
Design observations:
- nc_semantic_search_answer docstring documents the per-note
round-trip cost from the post-verification race guard
- News verification latency hint added to configuration.md
- SemanticSearchResponse exposes verified_count + dropped_count so
short result pages on high-ghost-density indexes are
distinguishable from genuine scarcity. verify_search_results now
returns (kept, dropped_count); production caller and tests updated
Minor:
- Comment clarifies the .get() fallback in verify_search_results is
defensive only (run_verifier always populates the entry)
- Eviction task-group guard narrowed from except Exception to
except RuntimeError (the only documented failure mode of
TaskGroup.start_soon on a closed group)
- Indexer logs a warning when a deck_card task is missing
board_id/stack_id, surfacing data-quality issues at index time
rather than at verification time
- New unit test covers the news verifier's non-numeric-id fail-open
path (one bad doc_id keeps the entire batch)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Enable ruff PLC0415 rule for all source files (tests excluded via
per-file-ignores). Move 136 inline imports to top-level across 33 files.
8 imports suppressed with noqa for legitimate reasons: circular
dependencies (client/__init__.py, context.py), optional dependency
guards (app.py document processors, auth/userinfo_routes.py), and
post-env-setup imports (smithery_main.py).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add type casts for Starlette app state access
- Add assertions for cipher, card, board, stack after initialization
- Add None checks for XML element text attributes
- Handle __package__ being None in tracing setup
- Fix TokenBrokerService initialization to use storage credentials
Resolves 42 type warnings from ty-check, enabling CI linting to pass.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Addresses reviewer feedback on PR #395 about O(n²) performance issue.
Changes:
- scanner.py: Add metadata field to DocumentTask with board_id/stack_id
- scanner.py: Populate metadata during deck card scanning (both initial and incremental sync)
- processor.py: Use metadata for O(1) card lookup via get_card() API when available
- processor.py: Fallback to iteration for legacy data without metadata
- context.py: Add _get_deck_metadata_from_qdrant() helper to retrieve metadata from Qdrant
- context.py: Use metadata for fast path lookup in chunk context expansion
- context.py: Add user_id parameter to _fetch_document_text() for metadata retrieval
Performance Impact:
- Before: O(boards × stacks × cards) iteration for each card lookup
- After: O(1) direct API call using stored board_id/stack_id
- Graceful degradation: Falls back to iteration for legacy data
Testing:
- All existing integration tests pass (test_deck_vector_search.py)
- Type checking passes with no new errors
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Adds comprehensive vector search support for Nextcloud Deck cards,
including semantic search indexing, chunk preview in the vector viz UI,
and proper deep linking to cards.
**Vector Search Indexing**
- Add deck_card scanning in scanner.py (scan_deck_cards function)
- Index cards from non-archived, non-deleted boards
- Store metadata: board_id, board_title, stack_id, stack_title, card_type, duedate, owner
- Content structure: title + "\n\n" + description (matches indexing format)
- Incremental sync based on lastModified timestamp
- Deletion tracking with grace period
**Vector Visualization Support**
- Add deck_card handler in context.py for chunk preview expansion
- Include board_id in search result metadata (bm25_hybrid.py, semantic.py)
- Expose metadata in viz_routes.py JSON responses
- Update vector-viz.js to construct proper Deck URLs: /apps/deck/board/{board_id}/card/{card_id}
- Update vector_viz.html filter label from "Deck" to "Deck Cards"
**Bug Fixes**
- Skip soft-deleted boards (deletedAt > 0) to prevent 403 Forbidden errors
- Applies to scanner, processor, and context expansion code paths
- Deck API returns deleted boards but rejects stack access with 403
**Testing**
- Add integration tests in test_deck_vector_search.py:
- test_deck_card_semantic_search: Filtered search with doc_type="deck_card"
- test_deck_card_appears_in_cross_app_search: Cross-app search includes deck cards
- test_deck_card_chunk_context: Chunk context fetching for viz preview
**Documentation**
- Update README.md: Add Deck cards to semantic search feature list
- Update semantic-search-architecture.md: Document deck_card support
- Update nc_semantic_search tool documentation
**Type Safety**
- Fix type narrowing for page_boundaries (could be None) using cast()
- Fix scanner.py payload None check for type safety
Resolves vector search for Deck cards across indexing, search, and visualization.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Add full integration for the Nextcloud News (RSS/Atom reader) app:
- Add NewsClient with complete CRUD operations for folders, feeds, and items
- Add 8 read-only MCP tools for listing/getting folders, feeds, items
- Add Pydantic models for News entities with camelCase alias support
- Add vector sync support for starred + unread items
- Add HTML to Markdown converter using markdownify for better embeddings
- Add Docker post-install hook to enable News app
- Add 25 unit tests for NewsClient API methods
Vector sync indexes starred and unread items, providing a balanced approach
that captures important (starred) and current (unread) content without
indexing the entire article history.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Previously, pymupdf4llm.to_markdown() was called twice - once in
PyMuPDFProcessor during indexing and again in PDFHighlighter during
visualization. Different image path lengths caused different character
offsets, leading to highlighted pages not matching their chunks.
Also fixed issue where all chunks on the same page showed all highlights
instead of just their own highlight. Now restores original page contents
between chunks using xref stream caching.
Changes:
- Add PDFHighlighter class requiring pre-computed page_boundaries and
full_text from document processor (no fallback extraction)
- Pass pre-computed data from processor to highlighter
- Extract page-relative portion of chunk text for cross-page chunks
- Add bounding box highlighting using text anchor search
- Run highlight generation in parallel with embedding/BM25
- Cache and restore page contents to isolate highlights per chunk
Results: Highlighting success rate improved from 51% to 95% (121/128).
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Implements optional context expansion for semantic search results that
fetches adjacent chunks (N-1 and N+1) from Qdrant to provide before/after
context. Removes configurable chunk overlap (default 200 chars) to avoid
duplicate text appearing in both context and excerpt.
Key changes:
- Add include_context and context_chars parameters to nc_semantic_search
and nc_semantic_search_answer tools
- Implement Qdrant cache fast path for chunk retrieval (avoids re-fetching
and re-parsing documents, especially important for PDFs)
- Add _get_chunk_by_index_from_qdrant() to fetch adjacent chunks
- Remove chunk overlap from before_context (last N chars) and after_context
(first N chars) to prevent duplicate text
- Fetch context in parallel with anyio.Semaphore (max 20 concurrent)
- Pass through page_number from SearchResult to SemanticSearchResult
- Remove document-level deduplication (keep chunk-level dedup from algorithm)
Context expansion is opt-in via include_context=true parameter. When enabled:
- Populates has_context_expansion, marked_text, before_context, after_context
- Adds truncation flags when context exceeds context_chars limit
- Falls back to document fetch for legacy data with truncated excerpts
Related: nextcloud_mcp_server/search/context.py:87-382,
nextcloud_mcp_server/server/semantic.py:161-255
The processor was not setting is_placeholder field when writing real
document chunks to Qdrant. This caused the placeholder filter to exclude
all documents (since None != False), resulting in 0 search results.
Now explicitly sets is_placeholder: False in payload when writing real
indexed chunks, allowing search filters to correctly distinguish between
placeholders and real documents.
Introduces a placeholder-based state tracking system to prevent duplicate
document processing during the gap between scanner queuing and processor
completion.
**Key Changes:**
1. **Placeholder Helper Functions** (`vector/placeholder.py`):
- `write_placeholder_point()` - Creates zero-vector placeholder when queuing
- `query_document_metadata()` - Queries for existing entry (placeholder or real)
- `delete_placeholder_point()` - Removes placeholder before writing real vectors
- `get_placeholder_filter()` - Filters placeholders from user-facing queries
2. **Scanner Updates** (`vector/scanner.py`):
- Replace `indexed_at` comparison with `modified_at` comparison
- Write placeholder before queuing each document
- Query per-document metadata instead of bulk-querying indexed_at
- Fixes bug where files were resubmitted every scan cycle
3. **Processor Updates** (`vector/processor.py`):
- Delete placeholder before upserting real vectors
- Ensures no duplicate points in Qdrant
4. **Query Filters** (all search files):
- Add `get_placeholder_filter()` to all user-facing queries
- Ensures placeholders never appear in search results or visualizations
- Applied to: bm25_hybrid.py, semantic.py, viz_routes.py, algorithms.py
**Architecture:**
- Placeholders use zero vectors with dimension from embedding service
- Payload includes `is_placeholder: True` flag for filtering
- Status field tracks: "pending", "processing", "completed", "failed"
- Deterministic UUIDs using uuid5 for consistent point IDs
**Impact:**
- Eliminates duplicate processing of same documents
- Fixes race condition where long-running documents get queued multiple times
- Prevents scanner from resubmitting files every scan cycle
- Maintains clean separation between in-flight and indexed documents
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
This commit fixes two critical issues with PDF processing:
1. **Text extraction mismatch (context expansion bug)**:
- Indexing used pymupdf4llm.to_markdown() producing markdown text
- Context expansion used page.get_text() producing plain text
- Different text formats caused character offset misalignment
- Search would find correct chunk, but expansion showed wrong section
- Fixed by making context.py use pymupdf4llm.to_markdown() consistently
2. **Diagnostic logging for page number assignment**:
- Added logging to verify page_boundaries exist in metadata
- Added logging to verify assign_page_numbers() assigns values
- Helps diagnose why page numbers show as null in search results
3. **mime_type storage bug**:
- Fixed incorrect field reference in processor.py:405
- Was using file_metadata.get("content_type", "")
- Should use content_type from WebDAV response
Changes:
- nextcloud_mcp_server/search/context.py: Use pymupdf4llm.to_markdown()
for PDF text extraction to match indexing method
- nextcloud_mcp_server/vector/processor.py: Add diagnostic logging for
page boundaries and assignment, fix mime_type storage
- tests/unit/client/test_webdav.py: Fix import sorting
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- scanner.py: Use file_info['id'] as doc_id instead of file_path
- scanner.py: Pass file_path in DocumentTask for content retrieval
- processor.py: Store file_path in Qdrant payload for later lookup
- context.py: Add _get_file_path_from_qdrant() to resolve file_id → file_path
- context.py: Update get_chunk_with_context() to handle file ID resolution
This makes the system resilient to file renames since file IDs are stable
identifiers in Nextcloud, while file paths can change.
This commit addresses multiple issues with async operations, PDF metadata
extraction, and type safety in document processing and search.
## Async/Await Fixes
- processor.py:259 - Added await for chunker.chunk_text(content)
- processor.py:270 - Added await for bm25_service.encode_batch(chunk_texts)
- tests/unit/test_document_chunker.py - Converted all 12 test methods to async
## PDF Metadata Enhancement
- pymupdf.py:143 - Added file_size metadata extraction
- pymupdf.py:145-206 - Refactored to extract text page-by-page
- Manually loop through pages instead of using page_chunks=True
- Generate page_boundaries metadata for precise page tracking
- Works around pymupdf.layout.activate() breaking page_chunks=True
- processor.py:32-66 - Added assign_page_numbers() helper function
- Assigns page numbers to chunks based on overlap with page boundaries
- Handles chunks spanning multiple pages
- processor.py:298-300 - Call assign_page_numbers() for PDF files
## Type Safety Fixes
- bm25_hybrid.py:184 - Removed int() conversion of doc_id
- semantic.py:131 - Removed int() conversion of doc_id
- viz_routes.py:275 - Removed int() conversion of doc_id
- Added comments documenting that doc_id can be int (notes) or str (file paths)
## Testing
- All 18 tests passing (12 unit + 6 integration)
- No type errors in modified files
- Container logs show successful processing
- Vector viz searches working correctly
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Track character offsets (start_offset, end_offset) for each chunk in vector
database metadata, enabling precise chunk highlighting in visualization pane.
Changes:
- processor.py: Store chunk_start_offset and chunk_end_offset in Qdrant metadata
- processor.py: Added metadata_version=2 to indicate position tracking support
- search/semantic.py: Return chunk positions from search results
- server/semantic.py: Expose chunk positions in API responses (SemanticSearchResult)
Enables viz pane to:
1. Display exact matched chunk with surrounding context
2. Highlight the precise portion of text that matched the query
3. Build user trust by showing what the RAG system actually retrieved
Position tracking uses ChunkWithPosition dataclass from document_chunker.py
which provides character-accurate offsets in the original document.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
