Zleap-AI/dsh-sag--packages-dsh-sag0

@zleap-ai/dsh-sag

Local SAG personal knowledge base plugin for DeepSeek Harness

包名
@zleap-ai/dsh-sag
版本
0.1.0
最近更新
2026年8月29日

安装

$npx -p @deepseek-ai/dsh dsh plugin --profile web add github:Zleap-AI/dsh-sag#1a50da21b32dc5e4e361fed945d071bc7feea3b4&path:packages/dsh-sag

@zleap-ai/dsh-sag

dsh-sag lets DeepSeek Harness use the local SAG personal knowledge base for search, reading, upload, note ingestion, and document management.

Compatible with DeepSeek Harness 0.1.1-rc.2, the current npm latest (there is no final 0.1.1 release yet). SAG must include the Connect dsh setting and local connector API. The unmodified SAG 1.8.3 release does not include this capability; use the build from SAG PR #154, or a later SAG release that includes it.

Quick start

Install and start SAG first, then install the plugin:

dsh plugin --profile web add @zleap-ai/dsh-sag
dsh plugin --profile web exec dsh-sag doctor

You can start using the plugin when doctor reports that SAG is connected. The command runs inside the web profile, so dsh-sag does not need to be on the system PATH. The plugin discovers common local addresses automatically; the normal setup does not require Python or manual credentials.

Connect SAG

If doctor cannot find SAG, save a connection with any one of these routes:

# Discover a running local SAG automatically
dsh plugin --profile web exec dsh-sag setup

# Use a connection file exported by SAG
dsh plugin --profile web exec dsh-sag setup ./sag-dsh.json

# Use the local SAG address directly
dsh plugin --profile web exec dsh-sag setup --url http://127.0.0.1:8000

Check the saved connection:

dsh plugin --profile web exec dsh-sag doctor

Ask dsh naturally

  • “Search my SAG knowledge base for the DW-2412P30 upload limit and cite the source text.”
  • “Upload /Users/me/Documents/product-manual.pdf to SAG, then summarize it after processing finishes.”
  • “Save the following meeting decisions as a note in SAG: …”

The plugin provides status checks, knowledge-source creation and listing, search and source-text reading, file upload, text ingestion, and document listing, reprocessing, and deletion. SAG advertises the operations available in the current installation; the plugin does not call capabilities that SAG has not enabled.

Update or remove

dsh plugin --profile web update @zleap-ai/dsh-sag
dsh plugin --profile web remove @zleap-ai/dsh-sag

Connection recovery

Make sure SAG is running, then run dsh plugin --profile web exec dsh-sag setup and dsh plugin --profile web exec dsh-sag doctor. If automatic discovery fails, export sag-dsh.json from SAG or connect with the local-address command above.

To let dsh manage the Python engine directly, see advanced embedded mode.