Proposal: experimental host for DeepSeek Harness (dsh)

Open
#264 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
45/100
Issue type
Feature
Clarity
Mostly clear
Activity status
Active
Tech stack
node.js, python, typescript

Research direction

Start with plugin/opencode/tokeneconomizer.ts and the recent adapter PRs mentioned in the issue, then read docs/PLUGIN.md for host conventions. Check whether dsh 0.2.x supports tool-free single turns and persistent per-tool settings before settling the side-call and pruning behavior. Done means an experimental host adapter with offline tests, documented verified behavior and tested versions, plus live-session evidence that learned context reaches the model.

Written by the indexing model from the issue text.

Description

Would you take a PR adding DeepSeek Harness (dsh, npm @deepseek-ai/dsh) as an experimental host?

There are no dsh entries in this repo yet, but several community ports exist as separate repos, and each one reimplements the engine in TypeScript. I'd rather add dsh to this repo and keep this repo's Python engine.

Plan:

  • A thin TypeScript plugin, modelled on plugin/opencode/tokeneconomizer.ts, that forwards dsh's session, prompt and tool events to te-hook and returns injected context. No second engine.
  • First cut: micro-skills and the loop guard. Side calls would run through dsh headless "<prompt>" with a --patch overlay that strips tools. If dsh cannot run a tool-free single turn, side calls stay off by default, as on Droid and agy.
  • Tool pruning only once I've confirmed dsh has a persistent per-tool setting; until then "nothing yet", like opencode.
  • Offline tests, a verified-behaviour section in docs/PLUGIN.md, a supported-versions row (dsh 0.2.x is still pre-release, so I'd list the exact versions tested), and evidence from a live session that a learned fact reached the model.

Before I start, is this in scope? And are there conventions for new hosts that I should follow beyond what the recent adapter PRs show?

Dominant language
Python
Stars
7.4k
Forks
500
Avg merge
1d 9h
Merged PRs (30d)
124

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from microsoft/TokenEconomizer

All issues in microsoft/TokenEconomizer

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.