feat: add OrcaRouter as a provider for the LLM analysis pass

[kuswardhanietidims-svg]

What & why

The AI optimization pass currently routes through Anthropic only — you set ANTHROPIC_API_KEY and --run-llm-analysis evaluates content with claude-3-sonnet-20240229. This PR adds a second first-class provider: set ORCAROUTER_API_KEY instead and the same analysis runs through OrcaRouter, an OpenAI- and Anthropic-compatible AI gateway.

OrcaRouter is a single endpoint in front of many models (OpenAI, Anthropic, Google, DeepSeek, and more) plus adaptive routing, automatic failover, zero-markup inference, observability, guardrails, and agent-tool governance. It also runs gateway-level, zero-trust security for AI agents on the same endpoint — screening every prompt/response and governing every tool call on a default-deny basis, with no application code changes. Because it speaks the Anthropic protocol, this project's existing ChatAnthropic wiring works unchanged: LLMSEOEnhancer just points anthropic_api_url at OrcaRouter's endpoint and uses OrcaRouter's vendor/model namespace.

That means this project's users get the whole OrcaRouter stack without treating it as an anonymous custom base URL — the same way Anthropic is already a named provider here.

What changed

• pyseoanalyzer/llm_analyst.py — when ORCAROUTER_API_KEY is set, LLMSEOEnhancer builds a ChatAnthropic instance pointed at https://api.orcarouter.ai, defaulting to the anthropic/claude-sonnet-4.6 model (configurable via ORCAROUTER_MODEL, e.g. openai/gpt-4o). Otherwise the original Anthropic path is used, unchanged.
• README.md — documents the new env var in the AI Optimization section and adds Docker --run-llm-analysis examples for both pre-built and locally-built images.
• tests/test_llm_analyst.py — covers both provider paths and the ORCAROUTER_MODEL override.

Verification

• Existing test suite: no regressions (the pre-existing failures on main — the max_pages mock mismatch in test_analyzer.py, the crawler-access HTTP mocks in test_website.py, and the LLM integration tests that need a real key — are unchanged; the LLMSEOEnhancer init tests now pass).
• Live test through the new code path with ORCAROUTER_API_KEY: the full enhance_seo_analysis pipeline (entity, credibility, conversation, platform, recommendations) completed end-to-end with openai/gpt-4o via OrcaRouter, and a raw ChatAnthropic invoke against anthropic/claude-sonnet-4.6 returned a valid completion.

Discord: discord.gg/YEubt8enRA · X: https://x.com/OrcaRouter

I'm an engineer on the OrcaRouter team.