roampal-ai/roampal-core

Summary

Fixes critical sidecar robustness issues and achieves full benchmark alignment for tag extraction.

Key Changes

• Wire TagService to sidecar LLM — v0.4.8 had extract_tags() in code but never called it. Tags were silently falling back to regex. Now properly wired for OpenCode sidecar scoring.
• Remove regex tag extraction — LLM-only tags matching benchmark behavior. If LLM fails, returns [] (no fallback). Use roampal retag to clean up existing memories.
• Robust backend selection — Health tracking, circuit breakers, priority-based fallbacks for sidecar scoring models.
• Unified onboarding — One-step model picker for roampal init and roampal sidecar setup. Auto-detects Ollama/LM Studio models, shows RAM requirements.
• Better failure handling — Once-per-session hints instead of repetitive popups when sidecar unavailable.
• roampal retag command — Re-extract tags on existing memories using your sidecar LLM. Supports --dry-run, --limit, --collection.
• Security hardening — SSL/TLS verification on external calls, input validation on /api/retag, URL validation for custom sidecar endpoints.

Stats

• 489 tests passing
• 6 files changed, +3447/-1223 lines