Robin Files

Release v2.9 — live model registry, honest results, easier first run

• Models are now discovered from each provider at startup instead of a hardcoded list, so new releases appear on their own and retired ones disappear. Mistral added as a provider — [#140], [#142]
• Search results no longer include the search engine's own navigation and footer links, and when nothing relevant is found Robin says so instead of summarizing whatever it happened to scrape — [#146]
• Follow-up chat and suggested pivots now receive the scraped page content they were missing, so grounded questions about an investigation actually work
• Cyrillic, CJK and Arabic titles, and results served at /index.php, are no longer discarded at the search layer
• New sidebar control for how much of each page the model reads (default raised from 2,000 to 8,000 characters), and Ollama's context window is now set explicitly so local models stop silently truncating
• Much easier first run: .env is just API keys and one is enough, everything else has a working default, plus a new TROUBLESHOOTING.md and clearer errors