dsh-dingtalk-connector

DingTalk AI Tables (multi-dimensional aitable) hold project data, customer records, and content schedules in fragmented Bases, while DeepSeek Harness sessions and tool chains run in a separate environment. Copy-pasting between the two, scribbling ad-hoc REST scripts, and switching back to check results is slow and error-prone. dsh-dingtalk-connector closes that gap: after installation it exposes 10 dingtalk_aitable_* tools inside Harness so you can discover Bases, read tables and fields, query records by ID or filter, batch-write changes, and export BOM-prefixed Chinese CSV files that Excel opens correctly—without ever leaving the conversation.

The capability set centers on three principles: read completely, write safely, export reliably. A diagnose tool checks CLI version, token status, and registered commands in one call. A scan tool probes candidate Bases for actual readability and tags each as open or blocked. Record queries support both targeted lookups and full-table pulls. Batch writes are gated behind a read-first, write-locked design: write and delete are both disabled by default, and deletion additionally requires a per-action confirm flag (a double-lock), deliberately leaving no room for model hallucination to meet an over-broad grant. Scheduled export jobs run inside the DSH host process on a daily or weekly cadence, dropping the selected table to a local CSV file so the nightly manual export simply stops being needed.

It suits knowledge workers and small teams who keep structured data in DingTalk AI Tables and simultaneously use DeepSeek Harness for analysis, drafting, and automation. The plugin does not replace the DingTalk client; it gives a Harness session a second, read-only (or controlled write) entry point into table data. Ask in plain language—"export all records with status pending review"—and seconds later the CSV is sitting in your local directory, no window-switching and no hand-written API calls involved.