chore: add evaluation scripts, CI workflow and dev tooling

scripts/eval/run_model_matrix.sh

@@ -0,0 +1,31 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+BASE_URL="${PULSE_BASE_URL:-http://127.0.0.1:7655}"
+EVAL_USER="${PULSE_EVAL_USER:-admin}"
+EVAL_PASS="${PULSE_EVAL_PASS:-admin}"
+SCENARIO="${EVAL_SCENARIO:-matrix}"
+REPORT_DIR="${EVAL_REPORT_DIR:-tmp/eval-reports}"
+MODEL_LIST="${EVAL_MODELS:-}"
+WRITE_DOC="${EVAL_WRITE_DOC:-1}"
+
+MODEL_ARGS=("-auto-models")
+if [[ -n "${MODEL_LIST}" ]]; then
+  MODEL_ARGS=("-models" "${MODEL_LIST}")
+fi
+
+echo "Running eval scenario '${SCENARIO}' against ${BASE_URL}"
+EVAL_REPORT_DIR="${REPORT_DIR}" \
+go run ./cmd/eval \
+  -scenario "${SCENARIO}" \
+  "${MODEL_ARGS[@]}" \
+  -url "${BASE_URL}" \
+  -user "${EVAL_USER}" \
+  -pass "${EVAL_PASS}"
+
+if [[ "${WRITE_DOC}" != "0" ]]; then
+  python3 scripts/eval/render_model_matrix.py "${REPORT_DIR}" --write-doc docs/AI.md
+  echo "Updated docs/AI.md model matrix."
+else
+  echo "Skipping docs/AI.md update (EVAL_WRITE_DOC=0)."
+fi