Cascade soft-delete tenant_members when a workspace is removed so the
tenant switcher no longer shows ghost entries as "#<id>". Also filter
orphaned memberships in /auth/me and switch to another workspace after
delete when one remains.
Add new styles for anchored form popups in the wiki browser, improving the user interface for folder actions and confirmations. Update internationalization strings for move confirmation titles in English, Korean, Russian, and Chinese locales to enhance user experience across multiple languages. Refactor existing components to integrate the new popup styles and ensure consistent behavior.
Add new API functions for managing wiki folders, including creating, updating, deleting, and listing folders. Introduce a new WikiFolder interface to represent folder structure and enhance the WikiBrowser component with folder actions. Update internationalization strings for folder-related actions in English, Korean, Russian, and Chinese locales. Refactor existing code to accommodate folder hierarchy and improve user experience in the wiki browser.
Enhance the internationalization support by adding optional descriptions for the embedding model in English, Korean, Russian, and Chinese. Update the UI to reflect these changes, ensuring users are informed about the optional nature of the embedding model when configuring knowledge base settings.
Revamp the wiki taxonomy planning process to assign coherent directory paths to entity/concept pages in a single pass, ensuring consistent reuse of existing folders. This change eliminates the need for per-page category assignments during the reduce phase, preserving user-curated categories.
Consolidate category path parsing and deduplication into a shared utility, improving consistency across the service and repository layers. Update related functions and tests to reflect these changes, ensuring efficient handling of category paths in the wiki ingestion process.
Centralize the wiki category-path normalization (separator/quote cleanup,
type-label stripping, dedup, depth cap) into a single types helper shared by
the service, repository, and taxonomy layers so stored and queried paths are
always cleaned identically and cannot drift.
Flatten category_path into fixed category_l1/l2/l3 columns (kept in sync on
write) and rewrite ListDistinctCategoryPathsByType to group, count, and
paginate directly in cross-dialect SQL, returning the page of child folders
plus the total distinct count. This removes the prior full-column scan and
in-memory pagination (whose maxPaths cap was effectively dead) and makes Total
and later pages correct.
Add the level columns + composite index to the wiki migrations and the repo
test schema, and cover the SQL grouping/counting/pagination with a new test.
Persist category and hierarchy metadata on wiki pages (new migration), expose
category-path retrieval and related repository/service/handler/router APIs, and
update wiki ingestion (batch, cite, taxonomy) plus agent wiki prompts to build
and consume the taxonomy. Regenerate swagger docs for the new endpoints.
Wiki summary/entity/concept/page-modify prompts instructed the LLM to
reproduce  verbatim, so high-entropy image URLs passed
through generation and got corrupted (dropped UUID segments, duplicated
or mangled path prefixes, even CJK chars inserted into the bucket name),
producing object keys that don't exist -> 404.
Mask all image URLs to opaque placeholders at the generateWithTemplate
chokepoint and deterministically rehydrate after generation; drop
corrupted/hallucinated placeholders so broken links never reach storage.
Introduce embed_channels storage, publish-token and session-token auth,
origin allowlist enforcement, shared Redis rate limiter, public embed API
routes, webhooks with SSRF validation, and session listing by embed source.
Three related improvements to the IM-session experience:
- Auto-generated titles: a new IM session starts untitled and gets a
content-based title from its first message via the same GenerateTitleAsync
flow web sessions use, instead of an identity title like "user · dm". Falls
back to the identity title when the first message has no text. All platforms.
- In-flight reply recovery (web chat): opening a session whose reply is still
being generated on an IM channel made the web client attach via
continue-stream and fail (the IM answer never streams through this server),
flashing a "stream failed" toast and forcing a manual refresh. Treat a failed
attach to an IM reply (channel === 'im') as recoverable — poll until it
completes, then reload the thread; on timeout surface the standard error.
Web/api replies keep the original behaviour, untouched.
- Source-filter classification: the sidebar filter treats "web" as sessions
with no IM mapping, but the list join excluded soft-deleted mappings, so
/clear'd IM sessions leaked into the user's own web chats. Include soft-deleted
mappings so a session ever bound to an IM channel stays under its platform.
Adds a repository regression test.
- Enhanced the `reparseKnowledge` function to accept optional process configuration overrides, allowing users to adjust parsing settings during reparse operations.
- Updated the frontend components to integrate reparse capabilities, including the `UploadConfirmDialog` and `KnowledgeBase` views, providing a user-friendly interface for confirming reparse actions.
- Added new localization strings for reparse confirmation and process configuration details in multiple languages.
- Modified the backend API to handle reparse requests with optional configuration, ensuring proper validation and persistence of overrides.
- Introduced new types and interfaces to manage reparse-specific data structures in the store, improving type safety and clarity in the codebase.
- Removed the `enable_multimodal` field from the `ChunkingConfigOverride` interface and related components to streamline configuration.
- Updated the logic in `UploadConfirmDialog.vue` to default `enableParentChild` to false when initializing from knowledge base info.
- Modified the `mergeChunkingConfig` function to ensure that `EnableParentChild` is authoritative, allowing explicit control over its state.
- Added a new test case in `knowledge_process_config_test.go` to verify the behavior of parent-child toggles in process configuration resolution.
- Updated the logic for `graph_enabled` in `UploadConfirmDialog.vue` to require both `nodeExtractConfig.enabled` and `graphEnabled` to be true.
- Added a new test case in `knowledge_process_config_test.go` to ensure that `GraphEnabled` is false when `ExtractConfig.Enabled` is false.
- Adjusted the `ResolveProcessConfig` function to reflect the updated dependency between graph and extract configurations, ensuring proper behavior in the process configuration resolution.
- Updated the `CreateKnowledgeFromFile` function to accept an optional `processConfig` parameter for per-upload parse configuration overrides.
- Introduced the `KnowledgeProcessOverrides` type to define various parsing settings, including parser engine rules and chunking configurations.
- Modified the API client and related components to handle the new process configuration, improving flexibility in knowledge file uploads.
- Updated documentation and API specifications to reflect the new `process_config` parameter in the upload process.
- Added a new `UploadConfirmHost` component to facilitate user confirmation for uploads, integrating with the updated upload logic.
- Enhanced user experience by providing detailed configuration options during the upload process.
This commit removes Jaeger tracing support from the project, including the associated Docker configurations, Go modules, and middleware. The changes simplify the codebase by eliminating unused tracing functionality, which was previously integrated for observability. Documentation and scripts have also been updated to reflect the removal of Jaeger references.
Moving knowledge with mode=reuse_vectors copies vector index entries
directly between knowledge bases, which only works when both KBs are
bound to the same vector store (CopyIndices is routed through the source
store). Nothing rejected a cross-store reuse_vectors move, so it would
copy target-KB rows into the source store and then delete the source
indices — corrupting vector data. The move endpoint validated only the
embedding model, and ListMoveTargets surfaces same-model KBs regardless
of their store binding, so a cross-store target is reachable.
Reject reuse_vectors moves whose source and target KBs are bound to
different vector stores:
- MoveKnowledge handler: synchronous 400 before enqueueing, with a
message pointing the caller at reparse mode.
- moveOneKnowledge: re-check before marking the knowledge as processing,
so a rejected move leaves it untouched rather than stranded.
- moveKnowledgeReuseVectors: a top-of-function guard as defense in depth
for any directly enqueued task, satisfying the pre-existing contract
noted in the comment ("callers must reject cross-store moves").
reparse mode is unaffected: it re-indexes into the target store from
scratch and is the supported way to move across stores.
- Updated logic to manage <think> and </think> tags in chat message content, ensuring accurate representation of thinking states.
- Introduced a new function to handle the closing of thinking events, enhancing clarity and reducing redundancy in event emissions.
- Adjusted event handling for reasoning content to support historical message reconstruction, improving user experience during chat interactions.
- Increased the default worker pool size for the asynq server from 16 to 32 to improve performance during concurrent uploads.
- Updated the .env.example and docker-compose.yml files to reflect the new default concurrency setting.
- Enhanced the handling of FAQ entries in the tool results, ensuring distinct identification through standard questions, improving user experience and clarity in the output.
- Refactored related components to utilize the new FAQ attributes, ensuring consistent display and interaction across the application.
- Eliminated all instances of the final_answer tool from the agent's configuration and logic, transitioning to a model where the agent concludes responses with plain text instead.
- Updated event handling and response analysis to reflect this change, ensuring that the agent's output is now directly delivered as text without relying on a dedicated final answer tool.
- Adjusted localization strings across multiple languages to remove references to final_answer, enhancing clarity and consistency in user-facing messages.
- Improved the overall structure of the agent's response handling to streamline the process and reduce potential confusion regarding tool usage.
refactor(agent): streamline answer handling and preamble logic
- Updated the agent's response handling to treat any text streamed before a tool call as a preamble, ensuring it is not included in the final answer.
- Removed the superseded marker logic from the event emission process, simplifying the flow of answer content and enhancing clarity in the UI.
- Adjusted the event handling to ensure that preamble segments are correctly retracted and relocated into the steps tree, improving the overall user experience.
- Enhanced the agent's internal logic to better manage answer segments and their states during streaming, leading to a more coherent response structure.
- Updated the `resolveStoreGroups` and `buildRetrievalParams` functions to handle multiple knowledge bases more effectively.
- Changed the structure of `buckets` to store pointers to `KnowledgeBase` objects instead of just IDs, allowing for richer data handling.
- Enhanced the retrieval parameter construction to differentiate between FAQ and document-type knowledge bases, ensuring correct routing for vector and keyword retrieval.
- Improved comments for clarity on the logic behind retrieval parameter construction based on knowledge base types.
- Updated the QA execution logic to differentiate between user-requested cancellations and actual errors, preventing unnecessary error events from being emitted when a stop is requested.
- Enhanced logging to provide clear feedback when a QA session is cancelled by the user, improving the user experience and reducing confusion.
- Added context cancellation checks to ensure proper handling of stopped generation scenarios, maintaining the integrity of the session flow.
Vector store create and raw connection-test accepted user-supplied
connection addresses without SSRF validation, while every other
admin-entered endpoint (e.g. model BaseURL) already runs
ValidateURLForSSRF. An admin could point an engine address at an
internal host and use the connection probe as an SSRF oracle.
Add validateConnectionAddrSSRF, applied at the two user-input
boundaries only:
- CreateStore: validated before any network dial.
- TestStoreRaw: a new TestRawConnection wrapper runs an engine-type
allowlist, required-field validation, and SSRF validation before
delegating to TestConnection. TestConnection stays validation-free
for trusted callers (env stores, stored configs already validated
at create time).
Validation covers every address-bearing field a driver dials
(Addr, qdrant Host+Port, weaviate Host and GrpcAddress) and
fails closed for unmapped engine types. The Elasticsearch probe no
longer follows redirects, closing a 302-to-internal bypass.
docker-compose whitelists the bundled vector-store service hostnames
(qdrant, milvus, weaviate, doris-fe) via SSRF_WHITELIST_EXTRA so the
out-of-box experience is preserved; postgres is intentionally not
whitelisted. .env.example and the OpenSearch dev guide document how
to whitelist external or local-dev addresses.
Cloning a knowledge base previously copied only the storage path strings
(knowledge.FilePath and chunk.ImageInfo.URL), so the source and the clone
shared the same physical objects in the storage backend. Once the original
file and extracted images are deleted on source removal, the clone is left
with dangling references and its document and images become unreadable —
data loss that occurs even for same-store clones.
Add a CopyFile primitive to the FileService interface and implement it in
every backend: server-side CopyObject on the object stores
(s3/obs/cos/oss/tos/ks3/minio), io.Copy on local, and a no-op on dummy.
Destinations use the knowledge-owned layout and reuse the existing
path/object-key guards; a sentinel ErrCrossBackendCopy is returned when the
source scheme does not match the backend.
Use CopyFile to deep-copy the document file in cloneKnowledge and the
extracted images in CloneChunk and cloneFAQKnowledgeBase via a shared
cloneChunkImageInfo helper that deduplicates identical image URLs per clone
and rewrites them to the new objects. Copied objects are cleaned up
best-effort if a clone fails partway through. A clone-time preflight rejects
cloning into a target bound to a different storage backend when the tenant
pins providers via StorageEngineConfig.
Adds unit tests for local CopyFile (independent copy survives source
deletion, traversal rejection, cross-backend rejection), cloneChunkImageInfo
(empty/multi/dedup/parse-failure/OriginalURL handling), and the storage
provider preflight.
Introduce opendataloader and PaddleOCR-VL parser engines with tenant-level
settings UI, replace liteparse, and harden Excel/PPT/Markdown parsing.
Optional odl-hybrid sidecar stays local-build only and is excluded from
default dev-start and full profiles.
Large scanned PDFs (hundreds of pages) were slow and fragile end-to-end.
This change addresses the parse, transport, and task-scheduling layers:
docreader (parse + transport):
- Parallelize per-page scanned rendering across processes (forkserver/fork),
with serial fallback. ~4-7x faster on large scanned PDFs; pdfium is not
thread-safe so we fan out across processes. Configurable via
DOCREADER_PDF_RENDER_PARALLELISM.
- Add server-streaming ReadStream RPC: emit one meta frame then one frame per
image, so documents with many page images are no longer capped by the unary
gRPC message-size limit (a 874-page PDF produced ~193MiB of images, far over
the 50MB cap) and memory is bounded on both ends. Unary Read is kept for
backward compatibility; the Go production reader switches to ReadStream.
VLM:
- Make the VLM HTTP timeout configurable (VLM_HTTP_TIMEOUT_SECONDS) and raise
the default 90s -> 180s so dense scanned-page OCR does not time out with
"context deadline exceeded".
Async task queues:
- Isolate high-volume, model-heavy fan-out tasks into dedicated asynq queues so
a single large document cannot saturate the shared worker pool and block
user-facing document parsing:
image:multimodal -> "multimodal"
chunk:extract -> "graph"
question:generation -> "question"
- Register the new queues in the server weight map and the cancel inspector's
scanned-queue set (so cancelling a knowledge still purges its pending tasks).
Reconstructing full document content from chunks relied on position math
(offset = content length - (EndAt - lastEndAt)) that assumes
len([]rune(Content)) == EndAt - StartAt. Two upstream behaviors break that
invariant and caused the merged "全文" view to drop or duplicate text:
- The parent-child chunker prepends a synthetic table header to split
tables; that header is zero-width in position space (start == end), so
Content is longer than EndAt - StartAt.
- Chunk Content can retain HTML entities (", > ...), inflating the
rune count relative to the source span.
Replace position-based trimming with text-overlap matching: find where the
accumulated text's suffix reappears at the head of the next chunk and join
there. Positions are only used to size the search window. This naturally
skips the prepended header and is immune to entity length drift.
Extract the logic into searchutil.AppendWithOverlap / MergeTextChunks and
reuse it across the three backend stitching sites (wiki_ingest
reconstructContent, graph mergeChunkContents, chat_pipeline
mergeOverlappingChunks). Mirror the same algorithm in the frontend
doc-content merge. Also fix the markdown hr/heading styles (thin solid
divider, primary heading color).
The wiki ingest pipeline was passing raw locale codes (e.g. "zh-CN",
"en-US") directly into LLM prompts where a human-readable language name
is expected. The model would see "zh-CN" instead of "Chinese (Simplified)"
in its instructions, which could degrade output quality.
Two call sites were affected:
- SlugUpdate.Language (ProcessWikiIngest, retract path): consumed by the
reduce phase to tell the page editor model what language to write in.
- lang variable (mapOneDocument): consumed by entity/concept extraction,
chunk citation classification, and index page rebuild — all LLM prompts.
Wrap both with types.LanguageLocaleName() which maps locale codes to
descriptive names (e.g. "zh-CN" → "Chinese (Simplified)"). Unknown
locales fall through unchanged so no information is lost.
Introduced a new test file for the knowledge repository, implementing various scenarios to ensure the reliability of the finalizing subtask counter. Key tests include:
- Concurrent promotion of subtasks to verify atomic behavior.
- Handling of partial decrements to ensure the row remains in "finalizing" state.
- Safety checks to prevent underflow of the pending subtask count.
- Regression tests to confirm that updates do not overwrite the pending counter.
These tests address previous issues with the finalizing process and enhance overall stability.
The finalizing subtask counter (pending_subtasks_count) could get stuck at a
non-zero value, leaving knowledge permanently in "finalizing" and never
promoted to "completed". Root cause and related hardening:
- UpdateKnowledge does a full-row Save; pending_subtasks_count was NOT in the
omit list, so any concurrent enrichment subtask that loaded the row, did slow
work (LLM call), then saved an unrelated field wrote back the STALE counter,
clobbering decrements other subtasks performed in between. Add
PendingSubtasksCount to omitFieldsOnUpdate so only the atomic helpers
(SetFinalizing / FinalizeSubtask / explicit column writes) ever touch it.
- FinalizeSubtask no longer gates the finalizing->completed promote on a
separate re-read of the counter. Every caller unconditionally runs the
guarded promote UPDATE (WHERE pending_subtasks_count = 0), the single
authoritative atomic check, so a racy/stale read can't strand the row.
- Decrements run on a context detached from the caller's cancellation
(context.WithoutCancel + timeout) so graceful shutdown / preemption can't
silently skip a decrement.
- Reconcile seeded vs actually-enqueued subtask slots (summary/question/graph)
and release any shortfall so un-enqueued planned slots (e.g. graph with
NEO4J off) don't strand the row.
- Reparse paths reset pending_subtasks_count via an explicit column write since
full-row Save now omits it.
Question generation is also moved off the synchronous single-task path onto a
batched async fan-out (windows of text chunks), each batch independently
queued / retried / traced, grouped under a postprocess.question span.
The finalizing subtask counter (introduced when wiki ingest was counted)
could leak slots, leaving a fully-parsed doc stuck in "finalizing" until
the housekeeping sweep wrongly marked it "failed".
- wiki ingest: a doc skipped in map (knowledge deleted / no chunks /
insufficient text) produced no docResult and was not a failedOp, so
neither the success nor the dead-letter drain fired. Drain the slot on
that terminal skip path.
- summary & question: the drain was keyed on the span-error variable,
which assumes "err != nil => asynq will retry". Several branches set
that variable yet `return nil` (insufficient text content, KB/knowledge
fetch failures) - terminal, no retry - so the drain was skipped. Key
the drain on the value actually returned to asynq (named retErr)
instead, so terminal nil-returns drain and only retried errors wait
for the final attempt.
Also fix the trace panel header flashing "已完成" mid-wiki: the latest
attempt's root span closes while async post-pipeline subspans keep
running, so trace.status read terminal while the row was still
"finalizing". Prefer parse_status on the latest attempt while it is
non-terminal so the panel header, LIVE badge and doc card agree, and add
the "finalizing" status label to all locales.
Wiki ingest runs asynchronously after the parse pipeline and was not
counted in pending_subtasks_count, so a document flipped to "completed"
while wiki generation was still minutes away (30s debounce + batch +
retries). That hid the in-progress state and dropped the stop-parse
affordance before wiki actually finished.
Count wiki as a single enrichment subtask (when WikiEnabled and the doc
has text chunks, matching the enqueue condition) so the row stays in
"finalizing" until wiki is done. The batch worker drains the slot at the
op's terminal state: once on successful map, and once when an op is
dead-lettered after exhausting in-batch retries. Retract ops (deleted
knowledge) are skipped. FinalizeSubtask guards both the decrement and the
promote, so a wiki op enqueued before this accounting shipped is a
harmless no-op on an already-completed row. A wiki op that never drains is
bounded by the existing housekeeping finalizing sweep.
Manually uploaded text documents could get stuck showing "generating
summary" forever, and editing/re-saving a document mixed its trace into
the previous attempt.
Three related fixes:
1. finalizeIndexedKnowledgeState no longer marks a text document
completed the moment chunks are indexed. Doing so made
KnowledgePostProcess hit its non-processing guard and skip the
summary/question/graph fan-out, stranding summary_status on
"pending". The row now stays "processing" (still retrievable via
enable_status=enabled) and post-process remains the sole authority
that drives processing -> finalizing -> completed.
2. ProcessManualUpdate now allocates a fresh span-tracking attempt via
OpenAttempt. Previously it ran with attempt 0, so processChunks
dropped every stage span and KnowledgePostProcess fell back to
LatestAttempt, piling the new run's summary/wiki subspans onto the
prior attempt's trace.
3. Enrichment workers (summary, question, graph extract) skip when a
newer attempt has superseded theirs. A stale subtask from a previous
upload/edit/reparse must not read deleted chunks or decrement the new
attempt's pending_subtasks_count, which would race-promote the row to
completed before the new attempt finishes.
Updated TestFinalizeIndexedKnowledgeState to expect the deferred
completion for text documents.
- Added a check to skip regex processing for non-image content, improving performance when no inline base64 payload is present.
- Updated regex pattern to ensure it correctly matches base64 image data URIs.
wizardchen