feat: add OrcaRouter provider preset for DeepSeek V4

[Marc-oss-hub]

Summary

Adds a deepseek-copilot.provider setting (deepseek | orcarouter) so DeepSeek V4 Pro & Flash can be served through the OrcaRouter model gateway instead of (or alongside) the official DeepSeek API. It also runs gateway-level, zero-trust security for AI agents on the same endpoint — screening every prompt/response and governing every tool call on a default-deny basis, with no application code changes.

What this changes

• New deepseek-copilot.provider setting. deepseek (default) keeps today's behavior; orcarouter switches the extension to the OrcaRouter gateway.
• When orcarouter is selected:
  • requests go to https://api.orcarouter.ai/v1
  • model IDs are namespaced automatically (deepseek/deepseek-v4-flash, deepseek/deepseek-v4-pro) — OrcaRouter rejects bare model names, so this is what makes the gateway route correctly
  • 401 / 402 error links point at the OrcaRouter console
• The model picker refreshes when the setting changes; baseUrl / modelIdOverrides are ignored while orcarouter is active (they describe the DeepSeek / third-party proxy path).
• Docs updated: README (EN + 中文), package nls localization, marketplace keywords.

Why a provider preset

The extension already ships baseUrl + modelIdOverrides as a generic escape hatch, but that expects users to know OrcaRouter's endpoint and its namespaced model-ID convention. A named preset makes the gateway a first-class option in the Settings UI — one toggle, correct model IDs, correct error links — the same treatment the official API gets. OrcaRouter is OpenAI-compatible on /v1/chat/completions, so the existing DeepSeek request path (streaming, thinking/reasoning_effort, tools) works unchanged; only the endpoint and model IDs differ.

Files changed

• src/config.ts — getProvider(), provider-aware getBaseUrl() / getApiModelId()
• src/endpoint.ts — OrcaRouter base URL + host matcher
• src/client/error/index.ts, src/client/consts.ts, src/client/types.ts — OrcaRouter error-link provider mapping
• src/provider/index.ts — react to the provider setting change
• src/consts.ts — OrcaRouter external URLs
• package.json, package.nls.json, package.nls.zh-cn.json — the new setting + localization
• README.md, README.zh-cn.md — usage docs

Getting started

1. Set "deepseek-copilot.provider": "orcarouter" in settings.
2. Run DeepSeek: Set API Key and paste an OrcaRouter API key.
3. Pick DeepSeek V4 Pro or DeepSeek V4 Flash in Copilot Chat as usual — thinking mode, vision, and tools all route through the gateway.

Verification

• npm run compile (tsc) clean; npm run lint (oxlint) 0 errors; npm run format:check clean.
• Live test against the gateway using the extension's exact request shape: POST https://api.orcarouter.ai/v1/chat/completions for deepseek/deepseek-v4-flash (stream + thinking/reasoning_effort) and deepseek/deepseek-v4-pro (stream + tools) both returned valid completions. Both model IDs are present in the gateway's /v1/models catalog.

I'm an engineer on the OrcaRouter team.