fix: recreate OpenClaw after bootstrap model upgrade

bootstrap-upgrade.sh updates .env with the new model name but only
restarts LiteLLM and DreamForge — it never recreates OpenClaw.
inject-token.js reads GGUF_FILE from env vars baked in at container
creation time, so a restart still uses the stale bootstrap model name
(extra.Qwen3.5-2B-Q4_K_M.gguf) causing HTTP 404 on every request.

Use `docker compose up -d --force-recreate openclaw` so the container
picks up the updated GGUF_FILE and LLM_MODEL from .env.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: Lightheartdevs

dream-server/scripts/bootstrap-upgrade.sh

@@ -394,6 +394,18 @@ LITELLM_UPGRADE_EOF
             log "Restarting DreamForge to pick up model change..."
             docker restart dream-dreamforge 2>&1 || log "WARNING: DreamForge restart failed (non-fatal)"
         fi
+        # Recreate OpenClaw so inject-token.js picks up the new GGUF_FILE/LLM_MODEL
+        # from .env. A restart alone won't work — env vars are baked in at container
+        # creation time, and inject-token.js builds the Lemonade model name from them.
+        if docker ps --filter name=dream-openclaw --format '{{.Names}}' 2>/dev/null | grep -q dream-openclaw; then
+            log "Recreating OpenClaw to pick up model change..."
+            if [[ ${#COMPOSE_ARGS[@]} -gt 0 ]]; then
+                docker compose "${COMPOSE_ARGS[@]}" up -d --force-recreate openclaw 2>&1 || \
+                    log "WARNING: OpenClaw recreate failed (non-fatal)"
+            else
+                log "WARNING: No compose args — cannot recreate OpenClaw. Restart manually."
+            fi
+        fi
     else
         log "WARNING: llama-server health check timed out. The model may still be loading."
         log "Check: docker logs dream-llama-server"