bce62350 · zzmq.dev@gmail.com/zz
Commit
bce62350411fd92b93226b28ad3fc6b7825c0c65
zz ladder build --provider: use Groq (or gemini/cloudflare) instead of only local Ollama The onboarding/ingestion pipeline (zz ladder build) could only ever summarize through a local Ollama server -- no --provider flag existed, unlike zz work. Fixed by extracting the --provider/--model -> ChatBackend resolution out of work.ts into a shared cli/lib/resolveChatBackend.ts (work.ts and workAssist.ts now both import it from there instead of duplicating it), and adding createChatSummaryBackend to engine/model.ts: a thin adapter from the richer ChatBackend interface (chat/tool-calling) down to this module's own summarize-only one, which is all ladder-build's ingestion ever needed. zz ladder build now takes the same --provider/--model flags zz work does. Omitting --provider keeps the exact previous behavior (Ollama if reachable, else mechanical text, zero config needed) -- a non-ollama provider skips that reachability dance and fails fast on a real error (e.g. a missing API key) instead of silently degrading, since a provider the caller explicitly asked for should error loudly, not quietly fall back. Verified live: resolveModelBackend("groq") resolves to a real Groq backend and a real summarize() call returns the model's actual output. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Q8tksS4bucUkJguqksSVYR
Troy Howard <thoward37@gmail.com> -- 2026-10-13T22:35:12.000Z