Introduction

The philosophy of DSH is “everything is a plugin”. In the DeepSeek Harness Web profile, usage information is scattered across provider dispatch, retries, cached tokens, and session history; if you wish to view this after restarting and filter by provider, model, and time range, the following introduces a community repository: xie-tj/dsh-token-usage-ledger.

The community directory is an independent site with no official affiliation to DeepSeek / High-Flyer. This article does not describe it as an official app store.

What is it

xie-tj/dsh-token-usage-ledger provides a persistent usage ledger and a visualized Usage dashboard for the DeepSeek Harness Web profile. The repository owner is xie-tj; as there is no separate “maintainer” field documented, this article only refers to it by repository name and installation commands.

The license is MIT.

It solves the following issues:

  • Records every provider dispatch, including failures and retries, and saves input, output, and cached tokens.
  • Backfills data using session history.
  • View usage by provider, model, and time range in Settings → Usage.
  • Provides an independent persistent ledger: retains history after restart and performs best-effort backfill on existing sessions.

Core Features

The capabilities verified below are listed.

Recording and Backfilling

  • Records every provider dispatch, including failures and retries.
  • Saves input, output, and cached tokens.
  • Backfills data using session history.
  • Independent persistent ledger: retains history after restart and performs best-effort backfill on existing sessions.

Display and Filtering

  • View usage in Settings → Usage.
  • Supports filtering by provider, model, and recent 7 days / 30 days.
  • Curves, bar charts, and API request counts retain precise values.
  • Model details use K/M/B/T compact units.
  • Hover tooltips on token traffic charts show daily input, output, cached, and total.
  • Displays cached hit tokens separately, and retains failed, unmeasured, and retry requests.

Data and Interfaces

  • Opens the usage_ledger storage domain version 2.
  • Provides the Remote method usageLedgerPlugin/snapshot, returning events, models, and daily.

Installation and Enabling

First install it to the Web profile, then start the Web profile. The commands verified in the documentation are as follows:

dsh plugin --profile web add github:xie-tj/dsh-token-usage-ledger
dsh --profile web

After installation, open Settings → Usage to view usage.

In production environments, the Git spec should be pinned to a commit SHA. For example:

dsh plugin --profile web add github:xie-tj/dsh-token-usage-ledger#<commit-sha>

The current version is built against the @deepseek-ai/dsh 0.1.0-rc.8 dependency line, requiring Node.js ^22.19.0 || >=24.0.0.

Typical Usage

View Usage

  1. Install to Web profile:
dsh plugin --profile web add github:xie-tj/dsh-token-usage-ledger
  1. Start Web profile:
dsh --profile web

  1. Open Settings → Usage.

  2. Use provider, model, and recent 7 days / 30 days filters.

Uninstall

dsh plugin --profile web remove dsh-plugin-usage-ledger

Uninstalling does not delete existing data in the usage_ledger storage domain.

Build from Source

pnpm install
pnpm run build

Verify before submitting:

pnpm run typecheck
pnpm run build
pnpm run test
pnpm run pack:check

Runtime Dependencies and Limitations

Host Dependencies

Host plugins have hard dependencies on the following services:

  • storageDomain: Opens the usage_ledger storage domain version 2.
  • sessions.
  • sessionPersistence.

Optional dependencies:

  • settings.

cordis.patch.yml will not install the storage-domain backend, session persistence provider, nor add service routes for custom profiles. Persistence capabilities depend on the backend configured for the profile in storageDomain.

Data Boundaries

  • The ledger is best-effort derived data, not part of request transactions.
  • The plugin does not estimate tokens, calculate fees, discounts, cache prices, or quota impact.
  • The page merges cache read/write display and does not plot them separately; it also does not show prices or amounts.
  • Persistence retention, cleanup, export, and migration tools are not implemented.
  • Regression tests do not replace actual Web profile startup tests.

Current Limitations

  • Web currently has no workspace selector.
  • Legacy sessions may display an unknown route.
  • The plugin settings card is currently read-only with no runtime configuration options.

Use Cases and Notes

Suitable for those who wish to view persistent usage, cached tokens, and failed/retry requests in a DSH Web profile, and filter by provider and model.

Please note the following:

  • The plugin loads into the Web profile as DSH plugin code and runs with permissions equivalent to the current dsh process; you should check the source code and license before installing.
  • The license is MIT.
  • In production environments, the Git spec should be pinned to a commit SHA.
  • Uninstalling does not delete existing data in the usage_ledger storage domain.
  • The plugin does not calculate fees, discounts, cache prices, or quota impact.

Conclusion

The value of this repository lies in persisting usage data to an independent usage_ledger storage domain and providing a visualized interface in Settings → Usage to view by provider, model, and time range.

The directory page link did not appear in the verified materials, so this article does not list a specific address.

GitHub: https://github.com/xie-tj/dsh-token-usage-ledger