Document local LLM (Ollama) as community-maintained alternative

[neuromechanist]

Summary

The project has moved to OpenRouter API as the production default due to superior performance. The local LLM setup (Ollama/GPU) should be documented as a community-maintained alternative rather than the primary deployment path.

Current State

• Many docs emphasize local GPU/Ollama deployment as the main path
• Code for local LLM support still exists and works
• Production uses OpenRouter API exclusively

Proposed Changes

• Update README.md to lead with OpenRouter as default
• Move local LLM docs to a docs/community/local-llm.md section
• Add "Community Maintained" badge/notice to local LLM docs
• Keep code intact for community contributors who want local deployment
• Update DEPLOYMENT.md to prioritize OpenRouter path

Rationale

• OpenRouter provides faster, more reliable results
• No GPU hardware requirements for users
• Local LLM code remains available for those who need offline/air-gapped deployment
• Reduces maintenance burden on core team while preserving optionality

Labels

• type: documentation
• priority: low

[neuromechanist]

Obsolete as written: the Ollama path was removed in #162, so there is no local LLM to document as a community-maintained alternative. Local inference would first need the modular OpenAI/Anthropic-compatible endpoint from #163; if that lands, this becomes documentation for that path instead. Closing or folding into #163 is the owner's call.