DeepSeek Harness Plugin Hub

Publish and manage complete Harness Profiles. Discover Plugins for your next setup.

Explore

PluginsPresetsDocsNews

Community

Publish a pluginContactReport an issue

Resources

Plugin Hub on GitHubDeepSeek HarnessSystem statusPrivacy notice
© 2026 DeepSeek Harness Plugin HubPowered byPaxTech

Independent and unofficial. Not affiliated with, authorized by, or endorsed by DeepSeek.

Bailian Kb Dsh — DSH Plugin for DeepSeek Harness
← Plugins

bailian-kb-dsh

Bailian Kb Dsh

Bailian knowledge-base tools for DeepSeek Harness: kb_search and kb_chat over the DashScope RAG API, plus the bl CLI management skill.

The plugin will be installed here. Keep web if you are unsure.

npx -y @deepseek-ai/dsh plugin --profile web add bailian-kb-dsh@0.1.19
READMECompatibilityVersions

Compatibility and provenance

Bailian Kb Dsh is published as bailian-kb-dsh and currently resolves to version 0.1.19. The Hub verifies its manifest and preserves the exact installation source for reproducible installs.

DSH compatibility
*
Runtime surfaces
web
Release source
npm
Registry updated
9/20/2026

Versions

0.1.19stable
8/25/2026
0.1.18
stable
8/24/2026
Show 1 more versionCollapse versions
0.0.0-beta-e023af5-202608241135beta
8/24/2026

Related plugins

Loading related plugins…

Latest
0.1.19
DSH
*
HMR
Process restart
Tree shaking
Safe tree shaking not declared
Unpacked size
282.5 kB
Files
36
Surface
web
License
Apache-2.0
Source
npm
GitHub
★ 320
Weekly downloads
60
Security scan
✓ v0.1.19 scan passed
Last push
9/1/2026
View source ↗Project homepage ↗
README badge

Click the badge to copy Markdown for your README.

Do you maintain this Plugin?Claim benefit · Priority security scan

Verify the GitHub repository declared in package.json to manage this listing. After you claim it, Hub will prioritize a security scan of the current version and publish the result when it passes.

Claim this Plugin →
Report an issue
DeepSeek Harness Plugin Hub
ProfilesPluginsCategoriesNewsDocsSign inManage Profiles
ProfilesPluginsCategoriesNewsDocsSign in

Related plugins

More verified plugins in search-research.

Weknora@wxg-prc-cpg/dsh-weknoraWeKnora knowledge retrieval tools for DeepSeek Harness (dsh): semantic search, document reading and RAG/agent answers over your own knowledge bases.Free Searchdsh-free-searchFree web search for DeepSeek Harness: 13 engines (Bing/DuckDuckGo/AnySearch/SearXNG/Exa/Tavily/Keenable/Firecrawl keyless; Parallel/Perplexity/SerpBase/DeepSeek with key) + time filtering + platform search + web_fetch, with web settings UI.Find Plugindsh-find-pluginFind DeepSeek Harness plugins inside the agent — live GitHub dsh-plugin topic search, ranked by stars.Mimirdsh-mimirResearch-assistant plugin suite (literature search, research wiki, LaTeX compile, independent subagent review) for the DeepSeek Harness

README

Bailian Knowledge Base for DeepSeek Harness

Knowledge-base retrieval tools for DeepSeek Harness, backed by Aliyun Model Studio (Bailian).

Bailian Console · 中文文档 · DeepSeek Harness · API Documentation

What is this?

bailian-kb-dsh is a plugin (and dsh bundle) that gives a DeepSeek Harness agent access to knowledge bases hosted on Aliyun Model Studio (Bailian). It registers two model-facing tools — kb_search for raw evidence and kb_chat for grounded answers — and ships a settings page plus a management skill for the bl CLI.

Retrieval happens through retrieval services you deploy on Bailian: a service binds one or more knowledge bases to an embedding/rerank configuration and is addressed by its agent_id. The plugin keeps the list of deployed services in front of the model, so it can decide whether the question is answerable from your knowledge at all.

Features

  • Two model-facing tools — kb_search returns scored chunks with source references; kb_chat returns a complete grounded answer
  • Service awareness — the workspace's deployed retrieval services are injected into the conversation, so the model knows what it can look up instead of guessing service ids
  • Low-friction setup — sign in to the Bailian console from the settings page to fill in the API key and workspace id; an existing bl CLI login is adopted automatically
  • Settings page — a "Bailian KB" section in the web UI for credentials, default services, and service-cache inspection
  • Management skill — bundled bailian-kb skill teaching the agent the bl CLI workflow for creating knowledge bases, ingesting documents, and deploying services

Requirements

  • DeepSeek Harness with its plugin runtime (@deepseek-ai/dsh-*), Node.js >= 22.12
  • An Aliyun Model Studio account: a workspace id and a DashScope API key (get one)
  • At least one deployed retrieval or Q&A service in that workspace — create one in the console or with bl knowledge service create / bl knowledge service deploy
  • The bl CLI (npm install -g bailian-cli) for anything on the management side — creating knowledge bases, ingesting documents, deploying services. Retrieval itself calls the API directly and never shells out, so kb_search / kb_chat work without it

Installation

dsh plugin --profile web add bailian-kb-dsh

The CLI adds the bundle to the profile's layer stack; no manual YAML editing required. To remove it:

dsh plugin --profile web remove bailian-kb-dsh

Verify the plugin is composed — dsh --profile web --dump-config should list a tool-bailian-kb row.

Configuration

Option 1 — Settings page (recommended)

After installation, Settings → Bailian KB appears in the web UI:

  • Fetch from console login — opens the Bailian console in a browser on the host machine; when you finish signing in, the API key and workspace id of that account are stored on the host (the key never travels to the browser). Each login requests a freshly issued key, so switching accounts is one click.
  • API key — write-only: the stored value is never echoed back, only reported as configured or not.
  • Workspace id / default retrieval service / default Q&A service — editable with echo; the service ids can be picked from the cached service list. Clearing a value falls back to the layers below.
  • Retrieval service cache — shows when the injected service list was last fetched, how many services it holds, and offers a manual refresh for a service you just created.

If you have already run bl auth login, the API key and workspace id are adopted once from ~/.bailian/config.json at startup. A value you deliberately clear is never re-filled.

Option 2 — Environment and credential files

# ~/.dsh/.env, or the credential store at ~/.dsh/.credentials.yaml
DASHSCOPE_API_KEY=sk-xxx                    # required
BAILIAN_WORKSPACE_ID=ws-xxx                 # required
BAILIAN_DEFAULT_RETRIEVE_AGENT_ID=aid-xxx   # optional
BAILIAN_DEFAULT_CHAT_AGENT_ID=aid-xxx       # optional

Option 3 — Profile patch

The bundle inserts its own entry into the profile; you can override it by id in ~/.dsh/cordis.patch.yml or the profile's patch file. An override replaces the whole config object (no deep merge):

- id: tool-bailian-kb
  config:
    defaultRetrieveAgentId: aid-search-service
    defaultChatAgentId: aid-chat-service
    chatTimeoutMs: 600000

Disable the plugin with - id: tool-bailian-kb plus disabled: true.

Config fields

The config doubles as the bailian-kb settings section, so edits in the settings page or settings document apply to the next call without a restart.

FieldTypeDefaultMeaning
workspaceIdstring?—Bailian workspace id; the API host is the workspace subdomain https://<workspaceId>.<endpointHost>
endpointHoststringcn-beijing.maas.aliyuncs.comHost suffix; replace for another region or a private deployment
defaultRetrieveAgentIdstring?—Service kb_search falls back to when the caller omits agent_id
defaultChatAgentIdstring?—Service kb_chat falls back to when the caller omits agent_id
agentVersionstring?—beta (draft) or a published version number; defaults to the latest published version. Not exposed to the model
chatTimeoutMsnumber300000kb_chat timeout — the server side is a minutes-scale agentic loop

Resolution order

ValueSettings layer (settings page)Entry config (profile patch)Credential store / env
DASHSCOPE_API_KEYwrite-only control—DASHSCOPE_API_KEY
workspace id✅ workspaceId✅ workspaceIdBAILIAN_WORKSPACE_ID
default retrieval service✅ defaultRetrieveAgentId✅ defaultRetrieveAgentIdBAILIAN_DEFAULT_RETRIEVE_AGENT_ID
default Q&A service✅ defaultChatAgentId✅ defaultChatAgentIdBAILIAN_DEFAULT_CHAT_AGENT_ID

Every value is re-read per call, so a rotated key or a switched workspace takes effect immediately. The API key and workspace id are mandatory: without them the tools fail with a message pointing at these configuration paths. Default services are optional — when the workspace has exactly one deployed service for a scene, that one is used.

Tools

ToolParametersReturns
kb_searchquery, agent_id (required), top_k? (default 5, applied client-side), images? (image URLs for multimodal)Scored chunks with source references, plus a total
kb_chatmessage, agent_id (required)The complete answer plus a request_id

agent_id is required in both schemas: the schema cannot tell the model whether this deployment pins a default, and discovering a missing default at call time wastes a round-trip. The configured default is still honoured for programmatic calls that omit it.

The list of deployed services — id, name, and scene — is injected into the conversation as a context message, refreshed periodically and whenever a bl knowledge service command changes the inventory. Long lists are truncated with the total stated, so the model never mistakes a partial list for the full inventory.

Errors

  • HTTP 4xx — most often an agent_id that no longer exists, so the service list is refreshed and appended to the error message for immediate recovery
  • HTTP 5xx — passed through unchanged
  • Missing credentials — the message names the configuration paths (~/.dsh/.env, ~/.dsh/.credentials.yaml, the settings page) and links to the console key page
  • kb_chat timeout — the message explains the server-side multi-turn retrieval and suggests retrying or switching to kb_search

Known limitations

  • kb_chat buffers the server stream, so there is no progress output while it runs.
  • top_k is a client-side cut: the request body carries no such parameter, and how many chunks the server returns is decided by the service configuration.
  • Service names carry the routing signal. The service list API does not return a description field yet, so the model judges what a service covers from its name alone. Name your services after their content (Product docs retrieval, not Service 1).
  • At most two pages per scene are fetched; beyond that the injected list is marked as truncated.

Development

pnpm --filter bailian-kb-dsh run build       # tsc → dist/ (node half) + tsdown → dist/web/client.js (browser half)
pnpm --filter bailian-kb-dsh run typecheck   # node and web tsconfigs
pnpm --filter bailian-kb-dsh run test

For local integration, add the working copy to a dev profile (the patch file is watched by HMR):

dsh plugin --profile dev add <this-repo>/packages/bailian-kb-dsh

Internal design notes — context injection strategy, service cache layout, refresh triggers — live in docs/kb-dsh/runtime-behavior.md; the maintenance checklist is docs/agents/dsh-plugin.md.

Contributing

Bug reports, feature requests, and PRs are welcome. See CONTRIBUTING.md for developer setup and the contribution workflow.

License

Apache 2.0