Back to marketplace

better-model-provider

Models & Data

sanshanya/better-model-provider

Per-model capability declaration for DeepSeek Harness: reasoning-effort levels and request modalities for OpenAI-compatible providers, with settings UI.

  • ai-providers
  • deepseek
  • deepseek-harness
  • dsh-plugin
  • reasoning-effort
GitHub Stars
6GitHub
Views
0DSH Plugin Hub
Forks
1GitHub
Open issues
1GitHub Issues
Manifest version
0.0.1better-model-provider
Latest push
Aug 20, 2026GitHub
License
MITTypeScript
Plugin type
Host + ClientRuns in both Host and Web Client

README

View source

better-model-provider

Per-model capability editing for DeepSeek Harness: reasoning-effort levels (with wire spellings), input modalities, and token capacities — declarations on your own routes, sparse overrides on official-catalog routes.

Custom models: edit declarations. Official models: edit overrides. Provider configuration remains official.

中文

Editing one model row

Why

Two per-model fields stayed YAML-only: reasoningEfforts and input. Until declared, the picker shows no effort control and image sessions refuse the model (... does not accept image input). This page edits them plus contextWindow / maxTokens — one row fully configures one model.

Install

dsh plugin --profile web add github:sanshanya/better-model-provider

GitHub installs print one pnpm allowBuilds key to add, then rerun add. A local link: install must npm install && npm run build first. Restart dsh web, and the Settings sidebar gains Model capabilities.

dsh plugin --profile web rm better-model-provider

Use

  1. Configure the provider + API key on the official Models page first; keys and lifecycle never move here.
  2. Expand a row: pick Custom under reasoning effort, check the levels (also check off, blank wire → null, if Off should exist); check image under input modalities for vision models; fill capacities in K/M spelling (380K, 1M). Apply — the picker and image admission follow immediately.
  3. Official-catalog routes: Manage official models — each field writes one sparse modelOverrides leaf; official capacities stay unless overridden; reasoning wire spellings are never guessed; Reset to official defaults lifts exactly what you overrode.
  4. Manage official providers (N) lists installed catalog routes; pick one and the first override you apply materializes its profile.

Dedicated-adapter apps (built-in DeepSeek / OpenAI Codex) declare capabilities in their own adapter code and never appear here.

Compatibility

Contract line @deepseek-ai/dsh-api-remotes ^0.1.0-rc.7 (verified against the published harness); surfaces outside the contract degrade silently. Development gates, live lanes, and invariants: see CONTRIBUTING.md.

License

MIT

Comments

0
Newest first