* feat: support multi-tag for document knowledge
- Add knowledge_tag_relations join table for many-to-many tag-knowledge mapping
- Migrate existing single tag_id data to the new table
- Update Knowledge struct, filter, payload types (TagID -> Tags/TagIDs)
- Rewrite repository layer: SetKnowledgeTags, GetKnowledgeTags, ListIDsByTagIDs
- Update tag count references to query knowledge_tag_relations
- Update service/handler layers for multi-tag create/list/batch-update
- Frontend: multi-select tag sidebar, multi-tag chip display, tag toggle in popup
- FAQ tags remain single-tag model (unchanged)
* feat(knowledge): tag edit dialog with dynamic overflow and inline creation
- Add TagEditDialog component with search, multi-select, and confirm
- Replace inline t-popup tag selector with dialog-based editing
- Tags overflow with +N badge, visible count adapts to container width via ResizeObserver
- Date/card layout: fixed date width, tags flex to fill remaining space
- Add persistent new-tag input bar in dialog (Enter to create, auto-select)
- Tag creation emits callback to refresh parent tag list immediately
- Apply same overflow + dialog trigger to both card view and list view
- Add i18n keys: tagEditDialogTitle, tagEditSearch, tagSelectedCount, tagOverflowTip, tagNewPlaceholder
* fix: exclude soft-deleted knowledge from tag reference counts
CountReferences, BatchCountReferences, and DeleteUnusedTags now
JOIN with knowledges table to filter out soft-deleted records
when counting knowledge_tag_relations. Previously they queried
knowledge_tag_relations directly without the deleted_at check,
causing stale tag counts after knowledge deletion.
* fix: refresh tag counts after async knowledge deletion completes
Add a second delayed tag list refresh (2.5s) after tag deletion to capture
updated reference counts for remaining tags. The original immediate refresh
handles removing the deleted tag from the list, while this delayed refresh
ensures other tags' counts are correct after the async deletion finishes.
---------
Co-authored-by: lyingbug <xiayula2014@gmail.com>
Move batch reparse work to an asynq task to avoid HTTP timeouts, add i18n
for all locales, guard in-flight documents, and prevent concurrent delete/reparse.
This commit introduces the Batch Reparse feature for the Knowledge Base.
Users can now select multiple documents and trigger a re-parsing pipeline in batches, streamlining the recovery from parse failures.
Tighten shared-agent readiness checks, surface not-ready toasts in the
selector, validate chat-model overrides in resolveChatModelID, and
improve configure deep-links for shared agents.
Block chat when agents lack an explicit chat or rerank model, guide users
through the selector and editor with field highlight, and align session QA
checks with the same readiness rules on backend and frontend.
Let admins send real requests against configured chat, embedding, rerank,
VLM, and ASR models from settings, with thinking controls shown only when
the model supports them.
Add recent=true to knowledge search so the mention picker can list recent
files without a keyword, and consolidate KB/FAQ/file chip styling across
chat input and message history.
Add compact agent filter beside channel list titles with selected name
display, fix embed localStorage agentId read on rebind, add IM agent
rebind tests, and deep-link agent editor counts to the matching tab.
Return provider-specific diagnostics when connectivity tests return zero
results, including SearXNG unresponsive_engines details instead of a
misleading API key hint.
Add agent complete wait timeout and answer fallbacks for cancel paths,
restore Telegram Markdown on finalize with plain-text retry, and show
answer-only final IM frames instead of the collapsed think summary.
Wait for agent complete without a short timeout, unify Telegram stream
parse mode, remove obsolete FinalAnswer handling, isolate per-section
newline tracking, add lifecycle tests, and update IM streaming docs.
Mirror Web AgentStreamDisplay/RagPipelineProgress in IM text streams:
retract live answers into 思考过程, show Web-aligned tool titles with
query/summary from structured data, and collapse thinking on finalize
instead of dropping it entirely.
Default unset agent thinking to explicit false so thinking_type models
receive disabled/enabled wire params. Render reasoning inside the quick-answer
pipeline timeline only when thinking events arrive, avoiding a flash of the
thinking step when thinking mode is off.
The runnable_by_viewer field was never populated for built-in agents,
so TenantRoleViewer users always hit 403 on agent-chat. Drop the field
and RBAC check so agent-chat matches knowledge-chat access for viewers.
Scope auth headers to feed fetches, use feed-scoped external IDs and
content-based incremental fingerprints, tolerate partial feed failures,
store feed URLs in settings, and fix credential configured detection.
Support connecting to MCP servers that require OAuth2, with a Cursor-like
zero-config flow: the user provides only a URL and the system performs
metadata discovery, dynamic client registration (DCR), and the
authorization code flow with PKCE on first connect.
- Persist per-user access/refresh tokens and per-service DCR client
credentials, encrypted at rest (AES-256-GCM)
- Store the PKCE verifier/state during the flow via Redis with an
in-memory fallback for Lite mode
- Cache MCP clients per (service, user) so tokens stay user-isolated
- Add endpoints for authorize-url, callback, status, and token revoke;
the callback is allowlisted for unauthenticated access
- Surface a structured "authorization required" hint when a tool call
hits an unauthorized MCP service
- Frontend: add an auth-type selector with OAuth config, inline
authorization status/actions, and render the connection test result
inline in the editor drawer instead of a stacked centered dialog
When editing a data source whose tree loads one level at a time (Feishu
wiki), a saved deep-node selection was invisible: only top-level spaces
were listed, so the selected node could not be seen and its ancestor
space rendered as unchecked.
- Add Connector.ResolveResourceAncestors: given selected resource IDs,
return the ancestor IDs that must be expanded to reveal them. Feishu
walks up via GetWikiNode (O(depth) per selection, no full-tree walk);
Notion/Yuque return empty.
- Fix wikiNode.ParentNodeID JSON tag: the Feishu API field is
parent_node_token, not parent_node_id, so GetWikiNode can return the
parent and the ancestor walk works in production.
- Expose POST /datasource/{id}/resource-ancestors (service + handler +
route) and a matching frontend API client.
- Edit-mode picker reveals hidden selections by expanding and lazily
loading each ancestor level, so selections are visible and checked.
- Gate child fetching on treeFullyLoaded instead of a partial-children
heuristic, avoiding a redundant request for full-tree connectors.
Listing resources for a Feishu data source eagerly walked the entire wiki
tree (one API call per node across every space), which exceeded the request
timeout for large knowledge bases. ListResources now takes a parentID and
returns a single level at a time: empty lists the wiki spaces, a space/node
ID lists only its direct children. The frontend fetches children on demand
when a node is expanded. Notion (full tree in one call) and Yuque (flat)
keep their existing behavior by ignoring non-empty parentID.
Refs: https://github.com/Tencent/WeKnora/issues/1672
PaddleOCR-VL renders tables as HTML with per-cell text-align styles.
These waste tokens on layout markup and bypass the chunker's
table-protection logic (which only recognizes Markdown pipe tables),
causing large tables to be split mid-row.
Convert inline HTML <table> blocks to GFM Markdown tables when possible,
and strip presentational attributes (style/class/align/...) for tables
using rowspan/colspan that Markdown cannot express.
Refs: https://github.com/Tencent/WeKnora/issues/1725
GET with a JSON body was non-standard and confused integrators. Register
POST as the preferred method and keep GET for existing clients; update SDK,
MCP server, and API docs accordingly.
Refs: https://github.com/Tencent/WeKnora/issues/1727
- Add IM service dependency to CustomAgentHandler for managing agent-related IM channels.
- Implement DeleteChannelsByAgent method to stop and soft-delete IM channels associated with a custom agent upon its removal.
- Update ListChannelsByTenant to exclude soft-deleted custom agents, ensuring consistent overview lists after agent removal.
The angle-bracket image target parser declared a local variable named
`close`, which shadows Go's built-in `close` function and is flagged by
revive's redefines-builtin-id rule, breaking `make lint`. Rename it to
`closeIdx`.
DingTalk websocket (Stream) channels stopped delivering @-mentions after
a day or two and only recovered when the IM channel was toggled off/on.
The factory fired the SDK's Start once and relied entirely on the SDK's
internal auto-reconnect, with no liveness supervision on WeKnora's side.
On long-running connections the SDK can enter a zombie state where the
connection object stays alive but no longer receives messages, leaving no
way to recover automatically. The cancel func was also ineffective because
the SDK ignores the context after the initial dial, so stopping a channel
leaked the old connection.
Add a reusable RunSupervised helper that keeps a long connection alive and
proactively recycles it every 6h on top of the SDK's own reconnect, which
bounds the worst-case outage and clears stuck connections. Wire DingTalk to
it, add LongConnClient.Close (disables AutoReconnect before closing so the
SDK does not resurrect the connection), and make the cancel func actually
tear the connection down. Drop the now-unused stream client reference from
the adapter, which only sends replies via sessionWebhook or OpenAPI.
checkSQLInjectionRisks and InjectAndConditions compiled their regexps
with regexp.MustCompile on every invocation. Both run on the datasource
SQL path — InjectAndConditions is called once per security filter
(tenant, soft-delete, hidden-KB, search-scope), so a single secured query
recompiled ~20 patterns. InjectAndConditions also compiled the identical
tail-clause pattern twice.
Hoist them to package-level vars (compiled once at init). Behavior is
unchanged — same patterns, covered by the existing TestValidateSQL_*
and TestInjectAndConditions tests.
Benchmarks (apple M, -benchmem):
InjectAndConditions 4347 ns, 72 allocs -> 1023 ns, 7 allocs
checkSQLInjectionRisks 36074 ns, 560 allocs -> 12984 ns, 12 allocs
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
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.
ParseLLMJsonResponse recompiled its ```json fence-extraction regexp on
every call that fell through the direct-unmarshal fast path — i.e. every
fenced LLM response on the graph-extraction path. Hoist it to a
package-level var.
Adds a test (none existed) covering direct, fenced, no-lang and
prose-wrapped inputs, plus a benchmark.
Benchmark, fenced payload (-benchmem):
before: 3205 ns/op, 5712 B/op, 54 allocs/op
after: 1315 ns/op, 693 B/op, 15 allocs/op
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
hasCommandSubstitution recompiled its three regexps with regexp.MustCompile
on every call, and ValidateArgs calls it once per argument — so validating
an N-arg command compiled 3N patterns. Hoist them to a package-level slice
compiled once at init.
Behavior is unchanged (same patterns); covered by the existing
TestScriptValidator_ValidateArgs / ValidateAll tests.
Benchmark, ValidateArgs with 8 args (-benchmem):
before: 16374 ns/op, 40142 B/op, 505 allocs/op
after: 2092 ns/op, 32 B/op, 1 allocs/op
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
WikiChunkCitationPrompt placed the per-batch <chunks> block before the
static rules, output schema and the per-document-stable <candidate_slugs>.
Within one document only ChunksXML changes between batches, so the stable
content trailing the chunks could not form a long common prefix and
providers re-billed the static rules on every citation batch (issue #1687
measured ~6.4% prefix-cache hit rate on wiki synthesis).
Reorder to [rules | output schema | candidate_slugs | chunks] so every
batch after the first shares the long [rules | candidate_slugs] prefix.
No instruction text was dropped; only block order and the inline
above/below directional references changed. Verified the cacheable prefix
grows from 281 to 2366 bytes and is byte-identical across batches.
Addresses #1687
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
When the dominant heading level repeats and becomes the split backbone,
a long section is sub-split via SplitText and every sub-chunk inherited
the section-level breadcrumb. Deep `###`/`####` headings that define a
sub-chunk's meaning were dropped from Chunk.ContextHeader, so queries
about content nested far below a shallow heading failed to retrieve it.
Track the heading hierarchy across the section and assign each sub-chunk
the breadcrumb in effect at its start offset, so the deepest active
heading reaches the breadcrumb. Adds a regression test.
Fixes#1674
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>