feat: add DeepSeek as LLM provider

[rushichavda]

Summary

• Adds DeepSeek AI as a new LLM provider with DeepSeek V3 (deepseek-chat) and DeepSeek R1 (deepseek-reasoner) models
• DeepSeek's API is OpenAI-compatible, so the integration uses ChatOpenAI with a custom base URL (https://api.deepseek.com) — same pattern as xAI/Grok and Moonshot/Kimi
• Users just need to set DEEPSEEK_API_KEY and select DeepSeek from the provider list

Changes

• src/model/llm.ts — Added deepseek- prefix model factory and deepseek fast model variant
• src/utils/env.ts — Registered DeepSeek provider config with DEEPSEEK_API_KEY
• src/components/ModelSelector.tsx — Added DeepSeek provider with V3 and R1 models to the UI
• env.example — Added DEEPSEEK_API_KEY placeholder

Test plan

• TypeScript type-check passes (tsc --noEmit)
• All existing tests pass (bun test — 9/9)
• Manual testing with a DeepSeek API key (requires account at https://platform.deepseek.com/)

Closes #111

[virattt]

Thanks!

I think we'll cut off any new LLM integrations after this one for now.

Let's all focus on new / improved functionality for the Dexter agent instead.

There are some cool features that we can clone from OpenClaw, like:

• dynamic prompts
• self-correcting memory
• imessage / whatsapp integration
• etc

[rushichavda]

Sure thing, can you create issues for this.