Add DeepSeek V4 model support (direct API + OpenRouter)

[junaid-hasan]

• Add deepseek-v4-pro and deepseek-v4-flash to both direct DeepSeek provider catalog and OpenRouter model list
• Retain deepseek-chat and deepseek-reasoner (deprecated 2026/07/24)
• Fix reasoning/thinking for direct DeepSeek API:
  • _supports_reasoning_extra_body() now recognizes deepseek.com URLs
  • Use DeepSeek-native format: extra_body={'thinking': {'type': 'enabled/disabled'}}
  • Add top-level reasoning_effort param (high/max) per DeepSeek API spec
• OpenRouter path unchanged — continues using existing reasoning format

Tested end-to-end via Gauss agent with deepseek-v4-pro, deepseek-v4-flash, deepseek-chat (direct + OpenRouter), all passing.

What does this PR do?

Adds DeepSeek V4 model support and fixes thinking mode for the direct DeepSeek API endpoint. DeepSeek's new V4 models (deepseek-v4-pro, deepseek-v4-flash) use a different thinking mode format
than OpenRouter — extra_body={"thinking": {"type": "enabled"}} + top-level reasoning_effort instead of extra_body={"reasoning": {...}}. The direct API endpoint was previously unreachable for
thinking mode because _supports_reasoning_extra_body() only returned true for OpenRouter URLs. The existing deepseek-chat / deepseek-reasoner models are being deprecated July 2026, so the
catalog needs the V4 replacements.

Related Issue

Fixes # (no existing issue — this is a new provider model update)

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

• gauss_cli/models.py — Added deepseek-v4-pro and deepseek-v4-flash to direct DeepSeek provider catalog and OpenRouter model list
• run_agent.py — _supports_reasoning_extra_body() now recognizes deepseek.com base URLs
• run_agent.py — _build_api_kwargs() uses DeepSeek-native thinking format ({"thinking": {"type": "enabled/disabled"}} in extra_body, reasoning_effort as top-level param)

How to Test

1. Set DEEPSEEK_API_KEY env var with a valid DeepSeek API key
2. gauss --provider deepseek --model deepseek-v4-pro -q "What is 2+2?" — should respond with thinking mode enabled
3. gauss --provider openrouter --model deepseek/deepseek-v4-pro -q "What is 2+2?" — should work via OpenRouter as before
4. Verify gauss_cli/models.py lists deepseek-v4-pro and deepseek-v4-flash in provider menus

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: Debian 13 (Equivalent to Ubuntu 26.04)

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A
• N/A — no new config keys, no architecture changes, no doc changes needed
• N/A — no tool behavior changes
• N/A — no cross-platform impact (pure provider config + API param mapping)

For New Skills

• This skill is broadly useful to most users (if bundled) — see Contributing Guide
• SKILL.md follows the standard format (frontmatter, trigger conditions, steps, pitfalls)
• No external dependencies that aren't already available (prefer stdlib, curl, existing Gauss tools)
• I've tested the skill end-to-end: gauss --toolsets skills -q "Use the X skill to do Y"

Screenshots / Logs

   === Direct DeepSeek API ===
   deepseek-v4-flash (non-thinking): 4 (1.4s)
   deepseek-v4-pro  (thinking):      Paris (2.2s)
   deepseek-v4-pro  (tools+thinking): listed directory contents (69s)

   === OpenRouter ===
   deepseek/deepseek-chat:            4 (1.9s)
   deepseek/deepseek-v4-pro (think):  Berlin (19.5s)

   All 5/5 passing.