narrator/run_server.sh

#!/usr/bin/env bash
# Start the LOCAL narrator model server (llama.cpp llama-server).
# Localhost only — bound to 127.0.0.1 so no trace content can leave the machine.
# This server is used ONLY by the narrator, ONLY for prose.
set -euo pipefail

# Repo-relative default so it works on any machine (the ./her runner also passes
# NARRATOR_MODEL explicitly). Override with NARRATOR_MODEL / NARRATOR_PORT.
HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
MODEL="${NARRATOR_MODEL:-$HERE/../models/Qwen3.5-9B-UD-Q4_K_XL.gguf}"
HOST="127.0.0.1"
PORT="${NARRATOR_PORT:-12781}"

if ! command -v llama-server >/dev/null 2>&1; then
  echo "llama-server not found on PATH (install llama.cpp)." >&2
  exit 127
fi

if [ ! -f "$MODEL" ]; then
  echo "Model file not found: $MODEL" >&2
  exit 1
fi

exec llama-server \
  -m "$MODEL" \
  --host "$HOST" \
  --port "$PORT" \
  -c 8192 \
  --jinja