dsh-code-lens
Run the following command in DeepSeek Harness:
dsh plugin install lisycotana/dsh-code-lens
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install lisycotana/dsh-code-lens in your DeepSeek Harness terminal; the plugin source lives at https://github.com/lisycotana/dsh-code-lens and it registers the codeLens session projection and the code_lens tool upon install.
About this plugin
In DeepSeek Harness code mode, the model writes a TypeScript program that orchestrates tool calls. The program receives every sub-call argument and result in full; the model sees only whatever the program finally returns. The session log records all of it, yet nothing surfaces the three-way split. dsh-code-lens fills that gap.
Two capabilities ship: a codeLens session projection that exposes whole-session hidden-call totals directly in the UI (hiddenBytes being the payload the program consumed, the log preserved, and the model never saw), and a code_lens tool that lets the model audit its own program after the run. Zero external dependencies, read-only access to the session event stream, no append re-entry, and it stays loadable in headless assemblies.
Designed for developers and model tuners who build complex tool-orchestration programs in code mode and need visibility into how many files were read and how many APIs were called behind the scenes, rather than trusting only the final sentence the model reports.
Use Cases
- After a code-mode program reads multiple files and calls APIs, inspect the hidden payload total directly in the UI
- Let the model audit its own sub-call history after the run via the code_lens tool
- Load the observability and audit panel in a headless assembly with zero external dependencies
Best For
- Developers writing TypeScript tool-orchestration programs in code mode
- Tuners who need to quantify hidden tool-call context consumed by the model
- Automation engineers running the DeepSeek Harness headless profile
Related Plugins
A method pack that makes AI coding agents plan against your real baseline, prove completion with fresh evidence, and reduce reworks and unsafe changes.
Turns the DeepSeek Harness session into a captain that builds a durable sub-agent team, splits goals into dependency-aware tasks, and coordinates work via direct messages and a live Web UI.
Gives coding agents design judgment, letting Claude Code, Cursor, and 70+ agents generate and iterate high-quality UI, presentations, and graphics right from the terminal.
Run the Pi ecosystem's plugins on DeepSeek Harness, unmodified, via a compatibility layer that implements Pi's public extension ABI on DSH's native services.