Summary

Two related enhancements to the reasoning/footer system:

1. Effort control for Claude Code

Claude Code now supports an --effort <level> CLI flag (low/medium/high) that controls reasoning depth. This adds support via Untether's existing reasoning config system:

• Add "claude" to REASONING_SUPPORTED_ENGINES with engine-specific levels (low/medium/high)
• Claude runner passes --effort <level> to the CLI subprocess
• Effort level injected into StartedEvent meta and shown in footer
• /config reasoning page now appears for Claude with 3 level buttons (vs 5 for Codex)
• Codex runner also injects reasoning into meta for consistency

2. Model version numbers in footer

_short_model_name() now extracts version numbers from Claude model IDs:

• claude-opus-4-6 → opus 4.6
• claude-sonnet-4-5-20250929 → sonnet 4.5
• claude-haiku-4-5-20251001 → haiku 4.5
• Bare names (opus) and non-Claude models (gemini-2.5-pro) unchanged

Footer example: opus 4.6 · medium · plan

Files changed

• src/untether/telegram/engine_overrides.py — per-engine reasoning levels
• src/untether/runners/claude.py — --effort flag, effort in meta
• src/untether/runners/codex.py — effort in meta (consistency)
• src/untether/markdown.py — version extraction, effort in meta line
• src/untether/telegram/commands/config.py — reasoning button for Claude, dynamic level buttons
• tests/test_meta_line.py — updated assertions, effort tests
• tests/test_config_command.py — Claude reasoning tests
• docs/reference/runners/claude/runner.md — --effort flag
• docs/how-to/model-reasoning.md — updated supported engines