jcode

feat(context): classify open-weight model families for static context limits

fix: seed global context-limit cache from named provider config (#366)

fix(onboarding): No on OpenAI login now skips to the normal screen with a /login hint

fix(tui): recolor fuzzy autocomplete matches in-palette instead of underline

test(onboarding): assert OpenRouter key is actually persisted, not just loop-free