github.com/Tencent/WeKnora changelog and release notes, page 8

v0.6.021 May 2026

fix(compose): stop publishing docreader gRPC port to the host by @lyingbug in #1308

What's Changed

fix(compose): stop publishing docreader gRPC port to the host by @lyingbug in #1308
feat(frontend): show UI build version on system info page by @lyingbug in #1313
fix(org): searchable join bypasses invite code expiry by @tsukiga-kirei in #1324
fix(frontend): make rerank model optional in agent editor save by @guangyang1206 in #1317
feat(cli): v0.3 — extended management surface and session subtree by @voidkey in #1315
refactor(retriever): introduce factory for KB-scoped retrieve engine resolution by @ochanism in #1310
fix(session): scope session access by user by @wolfkill in #1309
feat(system-info): surface DB migration errors with troubleshooting links by @lyingbug in #1335
fix(session): map ErrSessionNotFound to 404 across all handlers (#1309 follow-up) by @lyingbug in #1336
fix(migration): ensure pg_trgm is created before trigram index in 000041 by @lyingbug in #1334
fix(kb): map ErrKnowledgeBaseNotFound to 404 across handler helpers by @lyingbug in #1339
feat(cli): v0.4 - output contract hardening and mainstream alignment by @voidkey in #1342
feat: modify the skills registration directory by @1141520944 in #1344
fix: extend document processing timeout for large files by @HelloWeit in #1346
fix(model): hide Ollama tip for remote models by @wolfkill in #1330
fix(knowledge): prevent infinite scroll race condition causing missing documents by @caiyuze-cpu in #1356
feat(cli): v0.5 — agent CRUD + chunk subtree + audit-driven cleanup by @voidkey in #1354
feat(auth): 连接docreader支持auth by @CodeMarker0 in #1359
fix(docreader/auth): harden gRPC TLS/Token rollout from #1359 by @lyingbug in #1361
fix: redact sensitive fields and prevent credential loss on edit by @ochanism in #990
fix(docreader): bump grpcio floor to 1.78.0 to match generated proto by @toy0116 in #1363
refactor(credentials): /credentials subresource, at-rest encryption, and test-connection UX by @lyingbug in #1364
feat(embedding): add Zhipu AI embedder implementation by @Wellbek in #1365
fix(embedding): repair broken comment in Zhipu embedder by @lyingbug in #1366
test(repo): add wiki_config + indexing_strategy to sqlite test DDL by @toy0116 in #1362
fix: guard against empty choices and message=None in LLM response by @qizwiz in #1369
feat(cli): v0.6 — output contract, doc wait, status, multi-id delete by @voidkey in #1368
feat(knowledge-base): validate vector store bindings on create, copy, and delete by @ochanism in #1372
fix(agent): exclude processing docs from prompt by @wolfkill in #1370
feat(rbac): multi-tenant RBAC — schema, enforcement, member management, ownership, audit log (#1303) by @lyingbug in #1367
fix(session): update AgentQA to conditionally disable title based on request by @lyingbug in #1375
feat(frontend): redesign knowledge document tag selector by @lyingbug in #1376
feat(obs): 支持华为云obs存储 by @CodeMarker0 in #1377
feat(url): 支持docreader不上传替换白名单url的图片 by @CodeMarker0 in #1378
feat(tenant): richer workspace-switch toast that survives the post-switch reload by @lyingbug in #1379
feat(tenant): remember user's last-active workspace across logins by @lyingbug in #1381
fix(auth): unify DISABLE_REGISTRATION and auth.registration_mode by @lyingbug in #1384
chore(env): expose tenant RBAC + per-user tenant cap as env knobs by @lyingbug in #1385
fix(docparser): unescape MinerU markdown image syntax by @nixidexiangjiao in #1382
fix(moonshot): pin temperature=1 for models that reject other values by @nixidexiangjiao in #1383
style(menu): tighten session list density in sidebar by @lyingbug in #1387
fix(menu): avoid refreshing session list when opening an existing chat by @lyingbug in #1388
fix(menu): stop truncating session titles when extra space is available by @lyingbug in #1389
feat(session): persist last request state for UI restoration by @lyingbug in #1390
revert: drop colored role chip styling on tenant members page by @lyingbug in #1391
fix(i18n): escape '@' in invite email placeholder by @lyingbug in #1392
fix(chunker): keep top-level heading chunks separate by @wolfkill in #1396
doris: add configurable compatibility modes and guard mode switches by @SilentMil in #1395
feat: add vLLM server URL configuration for MinerU by @xdgu0275 in #1405
chore: update Go version to 1.26.0 in go.mod by @lyingbug in #1407
fix(frontend): mirror cross-tenant superuser Admin role in UI gates by @lyingbug in #1406
refactor(agent): simplify grep_chunks tool to a single regex query by @lyingbug in #1410
feat(frontend): make Vite dev proxy target configurable by @lyingbug in #1412
fix(chat): stabilize history pagination and keep message order by @lyingbug in #1409
feat(search): fan-out KB retrieval across bound vector stores by @ochanism in #1386
fix(knowledge): complete indexed documents immediately by @wolfkill in #1408
fix(client): align UpdateAgent request types with internal API by @HQF2017 in #1398
fix(session): scope wiki fixer to shared KB tenant by @wolfkill in #1413
fix(chat): improve history rendering stability by @lyingbug in #1419
fix(docparser): preserve MinerU markdown and persist relative images by @M1dnightSUN in #1404
fix(docparser): address review feedback on #1404 (MinerU markdown & image persistence) by @lyingbug in #1420
docs: add Chinese RBAC guide and link with shared space docs by @lyingbug in #1421
fix(frontend): guard knowledge list against stale updates by @iceloon in #1424
fix(chat): adjust user message container to support pre-wrapped text by @ThinhTT84 in #1425
fix(repository): correct query syntax to get correctly built-in models #1373 by @ThinhTT84 in #1428
chore(release): v0.6.0 by @lyingbug in #1422
docs(readme): flatten Latest Updates into single-line per-version list by @lyingbug in #1430

New Contributors

@HelloWeit made their first contribution in #1346
@caiyuze-cpu made their first contribution in #1356
@Wellbek made their first contribution in #1365
@qizwiz made their first contribution in #1369
@nixidexiangjiao made their first contribution in #1382
@SilentMil made their first contribution in #1395
@xdgu0275 made their first contribution in #1405
@M1dnightSUN made their first contribution in #1404
@iceloon made their first contribution in #1424
@ThinhTT84 made their first contribution in #1425

Full Changelog: v0.5.2...v0.6.0

Open source →

Additional notes

New Features

NEW: Tenant RBAC (Role-Based Access Control) — the headline of this release (#1303). WeKnora now enforces a per-tenant role matrix on every mutating route, with per-KB resource ownership. Highlights:
- 4-tier role matrix: Owner (one per tenant; can additionally delete the tenant) ⊃ Admin ⊃ Contributor (full owner of own resources, read-only on others) ⊃ Viewer (read-only). Two exceptions: cross-tenant superuser (User.CanAccessAllTenants=true) is implicit Admin in any tenant they switch into; API-Key-synthesized virtual users are pinned Admin in their owning tenant.
- Per-KB resource ownership: chunk → knowledge → kb → creator_id; same chain applies to FAQ entries, generated questions, KB tags and wiki pages. custom_agents.creator_id + custom_agents.runnable_by_viewer (default true) control agent ownership and viewer-callability.
- Two guard families: role guards (Viewer() / Contributor() / Admin() / Owner()) for tenant-level infra (models, vector stores, IM channels, …) and ownership guards (OwnedKBOrAdmin(), OwnedAgentOrAdmin(), OwnedChunkKBOrAdmin(), …) for resource writes. KB-access guard wired at the route layer for chunk / knowledge / knowledgebase routes (no per-handler helpers).
- Tenant members: invite / remove / role-change endpoints; new /leave endpoint; per-tenant audit log with daily retention sweep (default 90 days, audit_logs.created_at indexed); tenant_members table now drives membership (lifted from per-user to per-tenant in Plan 3); cross-tenant share managed by source-tenant Admin+.
- Configurable: tenant.enable_rbac (default true); false enters an "audit-only" grace window. New env knobs WEKNORA_TENANT_ENABLE_RBAC, WEKNORA_TENANT_MAX_PER_USER. RBAC state logged at startup. See Authentication and authorization.
NEW: Tenant Member Management & Multi-Workspace UX — invite-only gate, member listing UI with role chips, tenant identity surfaces reworked; tenant switcher in the user menu; tenant switch always redirects to KB list and clears tenant-scoped client state; last-active workspace persisted across logins; pending invitations dialog with polling + global invitation bell; rich workspace-aware notifications on login / tenant switch (raw-message handling, styled chips, survives page reload); QuickNav entry for members; "leave workspace" surfaced in i18n.
NEW: Self-Service Workspaces — any user can create their own tenant (capped per user via env knob); creation dialog with i18n; tenant name + description editable inline; cross-tenant superuser mirrored as Admin role chip in the UI.
NEW: weknora CLI v0.3 / v0.4 (GA) — graduates from preview to GA with comprehensive verb-noun subtree coverage:
- agent subtree: list / view / invoke / check / status / edit / delete / create (full agent CRUD with config rendering).
- chunk subtree: list / view / delete (with curation rationale).
- session subtree: list / view / delete.
- search subtree: chunks / kb / docs / sessions (replaces flat search).
- kb: new edit, pin, empty, check, status verbs; delete and other commands harmonized.
- doc: new download, view, wait (multi-target wait-all), unlink, upload --recursive; upload flag expansion; delete accepts multiple IDs.
- auth: new refresh and token verbs; transparent 401 retry transport.
- context CRUD: add / list / remove / use.
- link / unlink for project-level KB binding.
- mcp serve — curated stdio MCP server so AI clients (Claude Code, Cursor, …) can drive WeKnora directly; includes MCP chunk_list tool.
- Globals: --format, --json field-select, --jq, --paginate, --all-pages (canonical catch-up), --input, --log-level, --from-url, NDJSON output, bare-JSON output path, signal-aware contexts.
- Removed: envelope infrastructure (errors → stderr); --dry-run; internal/agent aiclient package; v0.0 scaffolding.
NEW: KB Retrieval Fan-out Across Vector Stores — a single KB can now bind to multiple vector stores; retrieval engine fans out queries across all bound stores and merges results. KB editor validates bindings on create / copy / delete. Retriever resolution introduces a factory pattern for KB-scoped engine selection.
NEW: AES-256-GCM At-Rest Encryption for MCP and Data Source credentials with graceful key-rotation handling. Sensitive fields redacted in API responses; new /credentials subresource pattern prevents credential loss on edit.
NEW: Docreader gRPC TLS + Token Auth (#1359) — app → docreader connection can be hardened with TLS + bearer-token authentication; docreader gRPC port is no longer published to the host by default; grpcio floor bumped to 1.78.0 to match generated proto.
NEW: Zhipu AI Embedder — first-class Zhipu embedding provider.
NEW: Huawei Cloud OBS object storage joins Local / MinIO / AWS S3 / Volcengine TOS / Alibaba Cloud OSS / Kingsoft Cloud KS3 / Huawei OBS.
NEW: vLLM URL configuration for MinerU doc parser.
NEW: Apache Doris compatibility modes — configurable Doris compat modes with mode-switch guards.
NEW: Docreader image URL whitelist — trusted URLs can be served as-is without re-uploading into WeKnora storage.
NEW: Server-Side User Preferences — per-user font / theme / memory-feature toggle persisted on the server; per-user KB pinning replaces tenant-wide pin model; "Shared by me" label across surfaces.
NEW: User favorites & recents under the user menu.
NEW: creator_name on agents and knowledge bases for visibility across surfaces.
NEW: Per-session last-request state persistence for UI restoration after reload.
NEW: Knowledge document tag selector redesign.
NEW: vue-i18n notification templates support raw message handling with styled chips.
NEW: Custom agent service supports KB sharing.

Improvements

IMPROVED: Frontend offline + legacy browser support hardened.
IMPROVED: Chat history rendering stability — pagination preserves message order; menu no longer refreshes the session list when opening an existing chat; session titles no longer truncate when extra horizontal space is available; session list density tightened in sidebar.
IMPROVED: Session — wiki fixer now scoped to shared KB tenant; session access scoped by user (security hardening); agent-chat rejects requests early when agent_id is missing.
IMPROVED: KB — indexed documents complete immediately instead of waiting for an extra sweep; vector store bindings validated on create / copy / delete; ErrKnowledgeBaseNotFound mapped to HTTP 404 across all handlers; ErrSessionNotFound mapped to HTTP 404 across all handlers.
IMPROVED: audit_log.Stop() no longer deadlocks when Start() is never called.
IMPROVED: Organization searchable join no longer bypasses invite code expiry.
IMPROVED: Chunker no longer merges top-level heading chunks.
IMPROVED: Moonshot models — moonshot-v1-* / kimi-k2.5 / k2.6 now pin temperature=1 automatically (they return HTTP 400 for any other value); kimi-k2 / k2-turbo / k2-thinking left untouched.
IMPROVED: MinerU markdown image syntax unescape — \!\[\]\(\) is restored to ![]() so downstream image extraction works.
IMPROVED: Test-connection — surfaces upstream and SSRF errors verbatim; falls back to stored apiKey when test-connecting an existing model.
IMPROVED: Test infrastructure — vector store tests now use a fake Elasticsearch server; knowledge base repository gains user pinning methods.
IMPROVED: Embedding pipeline — Zhipu AI embedder lands; broken comment in Zhipu embedder repaired.
IMPROVED: Sqlite test DDL augmented with wiki_config + indexing_strategy.
IMPROVED: agent exclude processing docs from prompt.
IMPROVED: LLM response — guard against empty choices and message=None.
IMPROVED: Configurable API proxy target for frontend dev environment.
IMPROVED: DISABLE_REGISTRATION now drives registration_mode too; removed redundant WEKNORA_AUTH_REGISTRATION_MODE env override.
IMPROVED: Tenant RBAC + per-user tenant cap exposed as env knobs.
IMPROVED: Auth — JWT tenant_id claim honored in middleware; tenant-scoped client state cleared on tenant change.
IMPROVED: gin per-route logs silenced; env config banner emitted at startup.
IMPROVED: Frontend — hide UI mutation surfaces for Viewer / non-creator; tenant switcher mirrors cross-tenant superuser Admin role in UI gates; role-aware UI gates no longer leak write affordances after tenant switch; agent editor rerank model now optional; Ollama tip hidden for remote models.
IMPROVED: System Info page surfaces UI build version, DB migration errors with troubleshooting links.
IMPROVED: Logger — logger.CloneContext propagates TenantRole.
IMPROVED: SSE / fetch paths — dropped insecure X-Tenant-ID short-circuit.
IMPROVED: Settings sidebar nav items grouped into labeled sections.

Bug Fixes

FIXED: API — agent-chat early reject when agent_id missing; deprecated tenant ConversationConfig field and KV write path removed.
FIXED: RBAC — chunk-id ownership chain for generated-question delete; sharing routes gated, tenant-disable shared agent → Admin+; ungated mutating routes plugged; FAQ + tag mutating routes aligned with KB ownership matrix; org-tenant gate gaps from Plan 3 closed; cross-tenant superuser organization owner pinned in DB instead of derived at runtime; remaining organization mutating routes gated with Admin+; dedup pending join/upgrade requests per (org, tenant, type); allow source-tenant Admin+ to manage cross-tenant shares; rbac-ui org owner row identified by tenant_id (not user_id).
FIXED: Client — UpdateAgent request types aligned with internal API.
FIXED: Frontend — input field agent selection logic improved for shared agents; permissions enhanced across KB and agent views; security — command-palette recent searches namespaced per (user, tenant); tenant switch away from tenant-scoped routes; tenant-members inline editing input attributes; chat/enableMemoryOverride simplified.
FIXED: i18n — @ escaped in invite email placeholder; "Shared by me" label added; chat titles and "leave workspace" updates across multiple languages; RBAC messages for tenant admin requirements.
FIXED: Docparser — MinerU markdown image syntax unescaped.
FIXED: Migrations — pg_trgm created before trigram index in 000041.
FIXED: Compose — docreader gRPC port no longer published to the host.
FIXED: Credentials — redact sensitive fields and prevent credential loss on edit.
FIXED: Auth — connection to docreader supports auth; gRPC TLS/Token rollout from #1359 hardened.

Refactoring

REFACTOR: knowledgebase — removed TogglePinKnowledgeBase from KnowledgeBaseRepository interface (replaced by per-user pinning).
REFACTOR: Tenant switch navigation unified to always redirect to KB list.
REFACTOR: Tenant member — tenant ID resolution simplified in handlers; tenant-access guards centralized in middleware.
REFACTOR: Custom-agent — KB sharing support split out.
REFACTOR: Organization — tenant-based access control; tenant-level membership transitions.
REFACTOR: Retriever — factory pattern for KB-scoped engine resolution.
REFACTOR: Agent — grep_chunks tool simplified to a single regex query.
REFACTOR: Frontend — GlobalCommandPalette, InputField, sidebar, menu, UserMenu templates streamlined for readability.
REFACTOR: CLI — comprehensive v0.3 / v0.4 cleanup: dropped --dry-run, dropped envelope infrastructure (errors to stderr), introduced bare-JSON output path, dropped internal/agent aiclient package (Go 1.26), --limit / --all-pages canonical pagination, auth security audit (gh CLI parity hardening), pre-PR audit fixes.
REFACTOR: Credentials — /credentials subresource pattern introduced.

Infrastructure & Build

BUILD: Go bumped to 1.26.0 in go.mod.
BUILD: grpcio floor bumped to 1.78.0 to match generated proto.
BUILD: Migrations — audit_logs.created_at index added; daily retention sweep job.
BUILD: Frontend — skill registration directory updated.

Documentation

DOC: New docs/RBAC说明.md (Chinese RBAC guide) and docs/wiki/安全认证/RBAC说明.md, linked with shared space docs.
DOC: docs/RBAC documents Contributor vs OwnedXxxOrAdmin selection rule.
DOC: Issue templates require concrete app/UI versions (not "latest").
DOC: CLI — cli/README.md, cli/AGENTS.md + cli/CHANGELOG.md brought in sync with v0.3 / v0.4 surface; stale e2e refs cleared; CI parity test added.

v0.5.213 May 2026

chore: release v0.5.1 by @lyingbug in #1104

What's Changed

chore: release v0.5.1 by @lyingbug in #1104
docs(readme): add Chrome Extension and ClawHub Skill badges by @lyingbug in #1105
fix(embedding): OpenAIEmbedder.doRequestWithRetry shadows err · retur… by @toy0116 in #1116
fix(chat): include history in model fallback by @Aphroq in #1112
fix(memory): dedupe episode write on duplicate streaming Done by @tsukiga-kirei in #1125
feat(wiki): enforce strict citation tracing and ontology reuse in Wik… by @lFeynman in #1120
feat(openmaic-classroom): 增加基于wiki知识图谱概念的微课堂生成支持，更新需求构建模板 by @peterchan2020 in #1119
fix: oidc auth login by @Windfarer in #1127
feat(ks3): add KS3 object storage support by @ChenRussell in #1109
fix(im): clamp reconnect backoff to avoid int64 overflow busy loop by @lyingbug in #1128
feat(frontend): unify fonts across the app and add per-user font/size/theme preferences by @Puma7 in #1121
fix(i18n): remove duplicate liveIndicator key in ko-KR locale by @lyingbug in #1129
feat(frontend): respect BASE_URL so the SPA works behind a path-prefi… by @toy0116 in #1115
fix: include shared knowledge in knowledge search by @Aphroq in #1111
fix(agent): correct graph config reporting by @Aphroq in #1114
fix(data-analysis): capture LOCAL_STORAGE_BASE_DIR at init time (#1040) by @guangyang1206 in #1130
fix(ingest): prevent filename-driven LLM hallucinations on scanned PDFs by @Puma7 in #1122
feat(kb-list): group pinned knowledge bases into a labeled section by @lyingbug in #1132
feat(frontend): replace standalone search page with global ⌘K palette by @lyingbug in #1135
feat(kb-list): source column + description subtitle, fix <style> XSS by @lyingbug in #1136
fix(files): fall back to global file service when tenant storage config is unavailable by @sqkstwj in #1096
refactor(knowledge): split 9.8k-line service file by responsibility by @lyingbug in #1134
Revert "refactor(knowledge): split 9.8k-line service file by responsibility" by @lyingbug in #1138
fix(storage): merge tos/s3/oss/ks3 tenant configs in buildStorageConfig (#1117) by @voidborne-d in #1137
feat(chunking): adaptive 3-tier strategy with editor UI and live debug panel by @Puma7 in #1124
refactor(knowledge): split 9.8k-line service file by responsibility by @lyingbug in #1139
feat(chunker): coalesce tiny adjacent heading chunks; drop redundant TierRecursive by @lyingbug in #1140
fix(frontend): rework chunking debug UX in the KB editor by @lyingbug in #1141
refactor(chunker): phase 1 cleanup — overlap bug, profile threading, protected spans, embedding trim, parent-child strategy by @lyingbug in #1142
fix(chunker): recognise CN chapter titles and multi-level numeric headings by @lyingbug in #1143
fix(knowledge): polish chunking-settings debug trigger and strategy hint by @lyingbug in #1148
fix: 为前端静态资源设置 Cache-Control，修复升级后需重新登录才能看到新页面 by @lyingbug in #1149
fix(frontend): make general-settings font picker actually work by @lyingbug in #1150
fix(graph): recover JSON from malformed/truncated fences in extract_entity (#1113) by @draix in #1131
fix: 修复流式回答think/answer混淆及对话结束标记问题 by @peterchan2020 in #1126
fix(agent): 统一最终答案渲染并去除重复/空思考块 by @lyingbug in #1153
fix(docreader): throttle heavy parser concurrency by @wolfkill in #1151
chore(agent): add fire-once stream diagnostic logs (agent + OpenAI layer) by @lyingbug in #1157
fix(frontend): render markdown tables in chat wiki drawer by @lyingbug in #1158
fix(frontend): polish model settings UI and fix disabled brand color by @lyingbug in #1162
feat: cli foundation by @voidkey in #1144
perf(wiki): subgraph API and interactive exploration for large graphs by @lyingbug in #1167
perf(wiki): lazy-load sidebar page list with virtual-scroll tabs by @lyingbug in #1169
fix(frontend): allow build scripts for pnpm to unblock Docker build by @lyingbug in #1170
fix(frontend): pass allow-build flags to pnpm install in Docker build by @lyingbug in #1171
fix(frontend): configure pnpm v11 build approvals in workspace file by @lyingbug in #1172
Fix authentication API documentation links by @aqilaziz in #1163
fix: propagate user attachments to agent query in AgentQA by @lyingbug in #1175
fix(frontend): use npm to match committed lockfile by @YunfanDu in #1174
Pr/wiki requeue fix by @toy0116 in #1155
perf(wiki): move ingest log to event table and index to on-demand API by @lyingbug in #1176
fix(wiki-ingest): harden lock-conflict handling and reduce retry churn by @lyingbug in #1177
feat(retriever): 接入 Apache Doris 4.1 作为向量数据库 by @tianlinzx in #1154
fix(retriever/doris): code review cleanup by @lyingbug in #1180
feat(wiki/graph): 优化操作帮助入口与图例视觉 by @lyingbug in #1181
fix(wiki-ingest): reconcile summary links and log feed when reduce LLM fails by @lyingbug in #1182
feat:cli v0.1 by @voidkey in #1179
chore(deps): bump github.com/mattn/go-isatty from 0.0.20 to 0.0.22 in /cli by @dependabot[bot] in #1193
chore(deps): bump github.com/spf13/cobra from 1.8.1 to 1.10.2 in /cli by @dependabot[bot] in #1192
chore(deps): bump actions/setup-node from 4 to 6 by @dependabot[bot] in #1191
chore(deps): bump docker/metadata-action from 5 to 6 by @dependabot[bot] in #1187
chore(deps): bump actions/download-artifact from 4 to 8 by @dependabot[bot] in #1186
chore(deps): bump docker/build-push-action from 3 to 7 by @dependabot[bot] in #1188
chore(deps): bump actions/setup-go from 5 to 6 by @dependabot[bot] in #1189
chore(deps): bump github.com/mattn/go-runewidth from 0.0.16 to 0.0.23 in /cli by @dependabot[bot] in #1190
chore(dependabot): expand coverage beyond /cli by @lyingbug in #1194
chore(deps): bump github.com/google/jsonschema-go from 0.4.2 to 0.4.3 by @dependabot[bot] in #1212
chore(deps): bump github.com/volcengine/ve-tos-golang-sdk/v2 from 2.7.23 to 2.9.4 by @dependabot[bot] in #1211
chore(deps): bump github.com/aws/aws-sdk-go-v2/service/s3 from 1.83.0 to 1.101.0 by @dependabot[bot] in #1213
chore(deps): bump github.com/minio/minio-go/v7 from 7.0.91 to 7.1.0 by @dependabot[bot] in #1216
chore(deps): bump github.com/sashabaranov/go-openai from 1.40.5 to 1.41.2 by @dependabot[bot] in #1214
chore(deps): bump github.com/aliyun/alibabacloud-oss-go-sdk-v2 from 1.4.1 to 1.5.1 by @dependabot[bot] in #1210
chore(deps-dev): bump less from 4.3.0 to 4.6.4 in /frontend by @dependabot[bot] in #1208
chore(deps): bump github.com/duckdb/duckdb-go/v2 from 2.5.4 to 2.10502.0 by @dependabot[bot] in #1215
chore(deps): bump github.com/go-openapi/strfmt from 0.25.0 to 0.26.2 by @dependabot[bot] in #1218
chore(deps): bump github.com/aws/aws-sdk-go-v2/credentials from 1.19.11 to 1.19.16 by @dependabot[bot] in #1217
chore(deps): bump gorm.io/gorm from 1.30.0 to 1.31.1 by @dependabot[bot] in #1209
chore(deps): update minio requirement from >=7.2.18 to >=7.2.20 in /docreader by @dependabot[bot] in #1206
chore(deps-dev): bump less-loader from 12.3.0 to 12.3.2 in /frontend by @dependabot[bot] in #1207
chore(deps): update lxml requirement from >=6.0.2 to >=6.1.0 in /docreader by @dependabot[bot] in #1205
chore(deps): update oss2 requirement from >=2.18.0 to >=2.19.1 in /docreader by @dependabot[bot] in #1203
chore(deps): update cos-python-sdk-v5 requirement from >=1.9.38 to >=1.9.42 in /docreader by @dependabot[bot] in #1200
chore(deps): bump docker/setup-buildx-action from 3 to 4 by @dependabot[bot] in #1199
chore(deps): bump docker/login-action from 3 to 4 by @dependabot[bot] in #1195
chore(deps): bump actions/upload-artifact from 4 to 7 by @dependabot[bot] in #1196
chore(deps): bump actions/checkout from 3 to 6 by @dependabot[bot] in #1197
chore(deps): bump @types/papaparse from 5.5.0 to 5.5.2 in /frontend by @dependabot[bot] in #1198
chore(deps): update paddlepaddle requirement from <4.0.0,>=3.0.0 to >=3.3.1,<4.0.0 in /docreader by @dependabot[bot] in #1201
chore(deps): bump vue from 3.5.17 to 3.5.34 in /frontend by @dependabot[bot] in #1204
chore(deps): bump github.com/weaviate/weaviate from 1.33.0-rc.1 to 1.37.2 by @dependabot[bot] in #1219
chore(deps): bump dompurify and @types/dompurify in /frontend by @dependabot[bot] in #1202
chore(deps): bump github.com/aws/aws-sdk-go-v2/config from 1.32.14 to 1.32.17 by @dependabot[bot] in #1221
chore(deps): bump github.com/chromedp/chromedp from 0.14.2 to 0.15.1 by @dependabot[bot] in #1223
chore(deps): bump nltk from 3.9.2 to 3.9.4 in /docreader by @dependabot[bot] in #1229
feat: add Tencent VectorDB retriever backend by @langcaiye in #1184
chore: bump Go to 1.26 and slim docreader dependencies by @lyingbug in #1231
chore(deps): bump google.golang.org/grpc from 1.80.0 to 1.81.0 by @dependabot[bot] in #1228
chore(deps): bump gorm.io/driver/postgres from 1.5.11 to 1.6.0 by @dependabot[bot] in #1227
chore(deps): bump github.com/ollama/ollama from 0.11.4 to 0.23.2 by @dependabot[bot] in #1226
chore(deps): bump github.com/milvus-io/milvus/client/v2 from 2.6.2 to 2.6.4 by @dependabot[bot] in #1224
chore(deps): bump github.com/jackc/pgx/v5 from 5.7.2 to 5.9.2 by @dependabot[bot] in #1239
chore(deps): update github.com/go-openapi/swag to v0.26.0 and remove unused dependencies by @lyingbug in #1240
chore(deps): bump golang.org/x/crypto from 0.50.0 to 0.51.0 by @dependabot[bot] in #1225
perf(wiki-ingest): scale to 4w-doc KBs + generic task queue / dead-letter by @lyingbug in #1241
feat: support Tencent VectorDB keyword retrieval by @langcaiye in #1234
fix(chat): 为 LLM 非流式/流式调用添加超时保护，防止 worker 被 hung 请求永久阻塞 by @ymkzs in #1238
fix(chat): LLM 兜底超时仅在上层无 deadline 时生效，并修复 ChatStream raw HTTP 路径的 cancel 泄漏 by @lyingbug in #1242
fix(frontend): 修正 LLM 调用超时文案与实际行为一致，上限放宽至 3600s by @lyingbug in #1243
feat(agent): human-in-the-loop approval for MCP tool calls (#1173) by @lyingbug in #1245
docs(cloud-image): add cloud-agnostic image packaging scripts by @lyingbug in #1249
fix(client): correct GetChunkByIDOnly endpoint by @HQF2017 in #1254
fix(cloud-image): enhance firstboot and cleanup scripts for improved security and functionality by @lyingbug in #1255
feat(web-search): add SearXNG provider (#1166) by @lyingbug in #1246
feat(cloud-image): support apt-based docker install for restricted-egress hosts by @lyingbug in #1260
chore(env): remove WEKNORA_SANDBOX_DOCKER_IMAGE from .env.example by @lyingbug in #1261
refactor(agent): rebuild multi-turn history from DB, drop llmcontext layer by @lyingbug in #1263
docs(api): full audit and update of API documentation by @voidkey in

Note truncated.

Open source →

Additional notes

🚀 New Features

NEW: weknora CLI v0.2 — the official command-line client lives under cli/. Mirrors the gh CLI <noun> <verb> convention with 10 top-level commands (api, auth, chat, context, doc, doctor, kb, link, search, version). Highlights:
- Hybrid search and streaming RAG chat against any knowledge base.
- Project-level binding via weknora link writing .weknora/project.yaml (vercel/netlify pattern); subcommands auto-resolve --kb from the link.
- Stable JSON envelope ({ok, data, error, _meta, dry_run, risk}) on every --json invocation; closed error-code registry enforced by an AST scanner test.
- Agent affordance: --dry-run for write commands, exit-code 10 + input.confirmation_required for non-interactive destructive writes, per-command "AI agents:" guidance auto-shown when CLAUDECODE / CURSOR_AGENT is set. Operational contract in cli/AGENTS.md.
- Multi-context auth (login / logout / list / status), OS keyring + 0600 file fallback for credentials, both API-key and password (JWT) modes.
- Health check via weknora doctor (4 statuses: ok / warn / fail / skip).
- See cli/README.md for install + 5-minute quickstart.
NEW: Adaptive 3-tier chunking — documents are now profiled before splitting and routed to one of three strategies: heading-aware (Markdown structure), heuristic (form-feeds, multilingual chapter markers DE/EN/ZH, all-caps titles, visual separators), or recursive (the modernized legacy splitter as a fallback). Auto-strategy is the new default for fresh KBs; existing KBs keep their previous behavior until the user opts in. See Chunking.
NEW: Human-in-the-loop approval for MCP tool calls (#1173) — when an MCP tool is marked sensitive, the agent now pauses and surfaces a ToolApprovalCard in the chat UI. Approval state is persisted (so refreshing the page does not lose context), enforced per user, and hardened for concurrent multi-instance deployments. See MCP integration.
NEW: Anthropic chat provider — first-class support for Claude models, including streaming through the Anthropic gateway and reasoning_content round-tripping for thinking-mode providers.
NEW: Apache Doris 4.1 retriever backend — Doris joins pgvector / Elasticsearch / Milvus / Weaviate / Qdrant / Tencent VectorDB as a supported vector store, with native stream-load ingest and hybrid query.
NEW: Tencent VectorDB retriever — full-text / keyword retrieval against Tencent Cloud VectorDB.
NEW: KS3 (Kingsoft Cloud) object storage — joins Local / MinIO / AWS S3 / Volcengine TOS / Alibaba Cloud OSS as a supported storage backend.
NEW: SearXNG web search provider (#1166) — self-hosted, federated metasearch as a first-class web search option, with zero-config defaults and hardened secret handling.
NEW: Global Command Palette — replaces the standalone search page with a global ⌘K palette that fuzzy-searches knowledge bases / chats / commands and can directly start a new chat from a result.
NEW: Cloud-image packaging scripts — scripts/cloud-image/ ships prepare.sh, firstboot.sh, cleanup.sh, and systemd units for producing reproducible self-hosted images (validated on Tencent Lighthouse; cloud-agnostic). Includes apt-based Docker install for restricted-egress hosts and idempotent firstboot with pinned image versions. See Cloud-image scripts.
NEW: KB editor — chunking settings panel surfaces the new strategy selector (Automatic / Markdown-optimized / Smart structure detection / Classic) plus advanced options for token limit per chunk and language hints. Sharper inline help text on every setting explains when defaults apply and when to tune.
NEW: Chunking debug panel — embedded "Test with sample text" panel under the chunking settings. Paste a snippet, hit Run preview, see selected tier, rejected tiers + reasons, document profile, size distribution stats over the full chunk set, and per-chunk cards with breadcrumb + content preview. Read-only, no DB or embedding side effects, 5-second server-side timeout.
NEW: POST /api/v1/chunker/preview endpoint backing the debug panel. Returns selected_tier, tier_chain, rejected[], profile, chunks[], and stats. Capped at 64k input runes / 500 chunks per response.
NEW: Per-tenant RRF (Reciprocal Rank Fusion) tuning — RRFK, RRFVectorWeight, RRFKeywordWeight are now configurable on the tenant RetrievalConfig. Defaults preserve the previous hardcoded behavior (k=60, weights 0.7/0.3).
NEW: Dedicated query-understanding model — agents can now route the query-rewrite / understanding step to a cheaper, faster model than the main reasoning model.
NEW: Document-level KB list filters with explicit batch-management UX (multi-select, batch delete, pinned-group section).
NEW: Frontend font picker + per-user UI preferences (font family, font size, theme) with a migration latch so legacy settings carry over safely.
NEW: OpenMaiC Classroom skill — generate micro-classroom content from knowledge-graph concepts, with an updated requirement-builder template.

⚡ Improvements

IMPROVED: Agent multi-turn history is now rebuilt from the database on every turn — the dedicated llmcontext storage layer (in-memory / Redis) has been removed entirely. Eliminates cache invalidation bugs, avoids attachments being dropped between turns (fix: propagate user attachments to agent query in AgentQA), and simplifies deployment (no extra Redis namespace required).
IMPROVED: Wiki ingest scaled to 40k-document KBs — operations move through a generic task queue with dead-letter handling, conflict retries are bounded, requeue counts are capped, and the wiki ingest log is moved off the request path to a dedicated wiki_log_entries table with an on-demand API.
IMPROVED: Wiki page-link graph performance — new subgraph API + interactive exploration UI so large graphs no longer hang the browser; documentation clarifies the distinction between the wiki page-link graph and the entity-relation knowledge graph.
IMPROVED: Wiki sidebar lazy-loads page list with virtual-scroll tabs; image / graph overview now shows operation help entry and refined legend visuals.
IMPROVED: Langfuse observability — spans expanded across the chat pipeline (retrieval, rerank, agent step), end-to-end TTFB is logged on both ends of the chat stream, and natural-stop candidates are recorded when the model returns no tool call.
IMPROVED: LLM call timeout hardening — non-stream / stream LLM calls now have a defensive fallback timeout (300s chat / 600s stream by default, configurable up to 3600s on the agent editor), only applied when no upstream deadline is present. Prevents worker pools from being permanently blocked by hung provider requests, and stops cancel leaks on the raw-HTTP streaming path.
IMPROVED: GPT-5 / o-series compatibility — MaxTokens is now mapped to MaxCompletionTokens for models that require the new field.
IMPROVED: Chunker recursive priority — splitBySeparators now genuinely walks separators by priority and recursively re-splits oversize sub-pieces with the next-priority separator. Mirrors the Python reference. Without this fix, a "one paragraph break followed by a long run of newline-separated lines" pattern could emit ~1900-rune chunks at chunkSize=300.
IMPROVED: ChunkOverlap default consolidated to 80 (~15% of ChunkSize). Previously the Go DefaultConfig used 64, the knowledge service used 50, the Python docreader used 100, and the frontend form initialised to 100. All paths now align.
IMPROVED: ContextHeader (Markdown breadcrumb) lives on Chunk.ContextHeader, separate from Chunk.Content. Restores the End-Start == len(Content) invariant that the document-reconstruction path in knowledge.go relies on for summary generation and UI highlighting. Eliminates a duplicate-heading regression where the section heading appeared twice in a chunk's body.
IMPROVED: Embedding pipeline — exponential backoff (200/400/800/1600/3200 ms) replaces the previous fixed 100ms × 5 retry loop, with context-cancellation between attempts. sanitizeForEmbedding caps single embedding inputs at 20k runes with a warning log on overflow.
IMPROVED: SplitParentChild forces children onto the recursive tier, skipping per-parent profile passes (previously paid N extra O(N) document scans).
IMPROVED: Heuristic splitter snaps overlap start to the nearest semantic boundary or newline instead of slicing mid-line / mid-word.
IMPROVED: Validator flow — when every tier is rejected, the chain returns the legacy tier's output directly instead of running SplitText a second time.
IMPROVED: Token limit per chunk — when set, ChunkSize is auto-clamped to a per-language character budget (with a 10% safety factor). Prevents overshooting embedding model token caps on CJK content where 1 char ≈ 0.6 tokens.
IMPROVED: KB-config API — strategy, tokenLimit, languages use pointer DTOs server-side so a payload omitting a field means "no change" while an explicit empty / 0 / [] resets to default. Previously these were write-once fields.
IMPROVED: Wiki prompts enforce strict citation tracing and ontology reuse, with dedicated handling for contradictions and per-rule conflict policy.
IMPROVED: Chunker recognises CN chapter titles and multi-level numeric headings; surrounding whitespace is trimmed before embedding; protected spans are honoured during heuristic splitting; tiny adjacent chunks are coalesced in the heading splitter.
IMPROVED: Frontend nginx serves static resources with gzip + correct Cache-Control headers.
IMPROVED: Feishu connector tolerates partial wiki-node listing failures instead of aborting the whole sync.
IMPROVED: KB list — pinned KBs are now grouped under a dedicated section; the type column is replaced with a richer source + description subtitle.
IMPROVED: Frontend — SPA respects BASE_URL and now works correctly behind a path-prefix reverse proxy.
IMPROVED: GitHub issue / PR templates translated to English and rewritten; Dependabot grouping + monthly cadence applied across all ecosystems.

🐛 Bug Fixes

FIXED: Mimo / DeepSeek-class providers — reasoning_content is now passed back to providers that require it for multi-turn thinking, and historical agent steps in the frontend correctly re-render reasoning_content instead of dropping it.
FIXED: Embedding pipeline — OpenAIEmbedder.doRequestWithRetry no longer shadows err and returns (nil, nil) on connection failure, which previously caused callers to SIGSEGV.
FIXED: Agent — quick-answer (RAG) mode excludes wiki-only KBs; rerank model requirement relaxed for custom agents; data_analysis toggle moved to the retrieval section and the stage is now opt-in per agent (#1244); attachments replayed correctly across multi-turn history; trailing thinking events that duplicate the final answer are suppressed; whitespace-only thinking events dropped; unified rendering when the model skips the final_answer tool; stream answer no longer mixes think and answer content; conversation end marker now reliably shown.
FIXED: Wiki ingest — concurrent lock conflict no longer exhausts retry budget; summary links and feed log reconciled when reduce LLM fails; JSON recovered from malformed / truncated fences in extract_entity; cap on requeueFailedOps retry count prevents queue pile-up.
FIXED: Storage — tos / s3 / oss / ks3 tenant configs are merged in buildStorageConfig (#1117); fall back to the global file service when tenant storage config is unavailable; sanitized document HTML now forbids <style> blocks.
FIXED: Data-analysis tool captures LOCAL_STORAGE_BASE_DIR at init time (#1040); shared knowledge included in knowledge search; graph config reporting corrected; multimodal pipeline unblocks when a provider:// image read fails.
FIXED: FAQ — PostgreSQL cast precedence in FAQ search; ingest no longer hallucinates summaries from filenames on empty PDFs; OCR text preserved on ingest QA follow-up; insufficient-content gate covered by tests.
FIXED: Middleware — camelCase secret fields are masked in request logs.
FIXED: Chat — model fallback now includes history; pipeline data-analysis stage no longer runs for agents that did not opt in.
FIXED: Memory — episode write deduped on duplicate streaming Done.
FIXED: IM — wecom reconnect backoff clamped to avoid int64 overflow → busy loop; OIDC auth login fixed.
FIXED: Frontend — markdown tables render in chat wiki drawer; radio button theme + disabled selection unified; external doc-link style and outline add-button styling unified; model-settings UI polished (disabled brand color); pnpm v11 build approvals configured for Docker build; npm used in production build to match the committed lockfile; LaTeX formula loading restored in the rare-case markdown path; KB-editor i18n: chunkOverlap initial value aligned with backend default (80, not 100); description texts updated.
FIXED: Chunker — Chunk.Start / End rune-offset invariant restored; heuristic applyOverlap aligned to boundaries; preview endpoint stats computed over the FULL chunk set; empty / whitespace-only sample returns friendly 400; recursive split fixed by QA audit (pointer DTOs, profile reuse, stats correctness).
FIXED: Docreader — heavy parser concurrency now throttled.
FIXED: Frontend i18n — duplicate liveIndicator key removed from ko-KR; preference fallback no longer infectious.
FIXED: Client SDK — GetChunkByIDOnly endpoint corrected; SDK stdout logging silenced via opt-in slog.

🔧 Refactoring

REFACTOR: internal/application/service/knowledge.go (9.8k lines) split by responsibility into focused files (knowledge_create.go, knowledge_delete.go, knowledge_clone_move.go, knowledge_faq.go, knowledge_faq_import.go, knowledge_post_process.go, knowledge_process.go, knowledge_util.go).
REFACTOR: Agent — dedicated internal/agent/approval/ package for the MCP human-in-the-loop gate; tools gain a shared exec_context.go.
REFACTOR: Web-search — dropped dead api_key path in the SearXNG provider; SSRF whitelist tests reformatted.
REFACTOR: CLI — command surface aligned with mainstream gh-style conventions across releases; unused v0.0 scaffolding removed.

🧱 Infrastructure & Build

BUILD: Go bumped to 1.26; docreader Python dependencies slimmed (removed legacy ocr/ package, retired storage.py, retired download_deps.py).
BUILD: Major dependency bumps across server-deps and frontend-deps groups (Dependabot grouped, monthly cadence): gorm, pgx/v5, aws-sdk-go-v2, weaviate, volcengine-tos, alibabacloud-oss, milvus, ollama, chromedp, duckdb-go/v2, minio-go/v7, sashabaranov/go-openai, volcengine-tos, aliyun-oss, vue 3.5.34, dompurify, papaparse, less/less-loader, plus GitHub Actions (checkout@6, upload-artifact@7, setup-go@6, setup-node@6, docker/*).
BUILD: Migrations — 000040_wiki_log_entries, 000041_task_queue_and_wiki_indexes, 000042_mcp_tool_approval; SQLite init migration updated to match.

📚 Documentation

DOC: New docs/CHUNKING.md — strategy explanations, settings reference with use-case presets, token-limit guide per embedding model, debugging workflow, and known trade-offs.
DOC: New docs/zh/mcp-approval.md describing the MCP human-in-the-loop approval flow.
DOC: New docs/cloud-image/README.md and docs/cloud-image/tencent-lighthouse.md covering cloud-image packaging.
DOC: API documentation — Swagger annotations restored across handlers, Swagger regenerated via make docs, and the hand-written docs/api/*.md rewritten to match current routes. Includes a new docs/api/auth.md.
DOC: New cli/README.md, cli/AGENTS.md, top-level CLI mention added to main README, plus a CHANGELOG and ADR section under cli/.

github.com/Tencent/WeKnora

Release timeline

Releases

What's Changed

New Contributors

New Features

Improvements

Bug Fixes

Refactoring

Infrastructure & Build

Documentation

What's Changed

🚀 New Features

⚡ Improvements

🐛 Bug Fixes

🔧 Refactoring

🧱 Infrastructure & Build

📚 Documentation