feat: add DeepSeek V4.1 Flash and model retirement notices

[Vizards]

Related issues

Related to #265.

Problem

The model picker does not expose deepseek-flash or explain the official routing changes for the old V4 model names. Add DeepSeek V4.1 Flash and migration notices so users can choose the new entry and plan existing conversations around the transition.

Changes and scope

• Add DeepSeek V4.1 Flash (deepseek-flash) with native vision, thinking mode, tool calling, CNY/USD prices, and a model ID mapping setting.
• Keep all three old entries selectable, with warnings alongside their original descriptions. V4 Flash and Vision Exp show routing notices instead of old prices. Pro keeps its prices until September 14, 2026 at 12:00 Beijing time (UTC+8), then shows the temporary V4.1 Flash routing and billing notice. The existing pricing refresh timer updates this metadata.
• Custom endpoints and mapped IDs receive generic migration guidance without official routing or price claims. Model mapping changes refresh the picker.
• Preserve old request IDs, conversation handling, and Vision Proxy choices. New Flash uses the existing native Vision Exp request path and is excluded from Vision Proxy candidates.
• Align English/Chinese guidance with the official API documentation and clarify that maxTokens=0 uses the API default. Versioning remains managed by release-please.

Validation

• CI passed for cf3d938: lint, formatting, compilation, and VSIX packaging. git diff --check also passes.
• 27 local regression cases passed with mocked VS Code/network responses, covering retirement timing and refresh, preserved descriptions, pricing, mappings, thinking modes, images, replay, output limits, streaming, and integration with feat: add configurable request headers #266 request headers.
• VS Code Stable 1.136.2 / Copilot 0.64.1, official API: verified picker warnings/prices, four thinking settings, user images, old Vision Proxy paths, mappings, output limits, cancellation/retry, and observed cache hits.
• Actual Copilot UI via computer-use: new Flash tool images in High and None modes, multiple tools, follow-up turns, and continuing the same conversation after Reload Window. UI/API runs preceded the feat: add configurable request headers #266 rebase; request-header integration was checked after rebase, and subsequent commits only changed wording.

Known limits

• The existing mixed user/tool message ordering defect can return HTTP 400 on both new Flash and Vision Exp. It also exists on main and is left for a separate fix.
• Copilot removes historical tool-image pixels before passing history to the provider. Reload checks confirmed text and reasoning recovery, not restoration of historical pixels. The additional Vision Exp tool-image UI comparison was not completed.
• The future Pro cutoff was simulated locally. Observed cache hits do not guarantee reuse across different prefixes, model switches, or upstream routing changes.

Final checks

• I reviewed and can explain every change
• I reviewed any user-facing text changes in both English and Chinese
• This pull request contains no API keys, authorization headers, private prompts, private code, or unreviewed sensitive logs

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-10T12:28:05.481054Z	cf3d938	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[Copilot]

🟢 Approval recommended

The implementation consistently integrates the new model and retirement metadata, with successful CI and no unresolved defects found.

Pull request overview

Adds DeepSeek V4.1 Flash while preserving legacy model behavior and communicating official migration timelines.

Changes:

• Adds V4.1 Flash with native vision, thinking, tools, pricing, and ID mapping.
• Adds endpoint-aware retirement notices and cutoff-based Pro pricing.
• Updates English and Chinese documentation and settings text.

File summaries

File	Description
src/provider/vision/sources/vscode/index.ts	Preserves legacy proxy selection.
src/provider/retirement.ts	Defines retirement notice logic.
src/provider/models.ts	Adds warnings and conditional pricing.
src/provider/index.ts	Refreshes metadata for mapping changes.
src/i18n.ts	Adds bilingual model and migration text.
src/consts.ts	Registers V4.1 Flash capabilities and pricing.
src/config.ts	Clarifies output-limit behavior.
resources/walkthrough/show-models.nls.zh-cn.md	Updates Chinese picker guidance.
resources/walkthrough/show-models.md	Updates English picker guidance.
README.zh-cn.md	Documents models and migration in Chinese.
README.md	Documents models and migration in English.
package.nls.zh-cn.json	Updates Chinese setting descriptions.
package.nls.json	Updates English setting descriptions.
package.json	Adds the model mapping configuration.

Review details

• Files reviewed: 14/14 changed files
• Comments generated: 0
• Review effort level: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.