fix: 07-rag-setup.sh — riktig import-endepunkt /ragFiles:import (ikke :importRagFiles)

infrastructure/07-rag-setup.sh

@@ -5,6 +5,9 @@
 # Uploads ALL docs/*.md + corpus-seed/*.md + protocols/sessions/*.md to GCS
 # Idempotent — safe to run multiple times
 # Source .env before running: source .env
+#
+# VERIFIED WORKING ENDPOINT:
+#   POST /v1beta1/projects/{id}/locations/{region}/ragCorpora/{id}/ragFiles:import
 
 set -euo pipefail
 
@@ -28,12 +31,7 @@ gcloud services enable aiplatform.googleapis.com \
   --project="${PROJECT_ID}" --quiet
 echo "✓ APIs enabled"
 
-# ── Hent project-number (kræves av importRagFiles-endepunktet) ────────────────
+# ── GCS bucket (idempotent) ─────────────────────────────────────────────────
-PROJECT_NUMBER=$(gcloud projects describe "${PROJECT_ID}" \
-  --format='value(projectNumber)' 2>/dev/null)
-echo "  Project number: ${PROJECT_NUMBER}"
-
-# ── GCS bucket (idempotent) ─────────────────────────────────────────────
 CORPUS_BUCKET="${PROJECT_ID}-agent-corpus"
 if ! gsutil ls -b "gs://${CORPUS_BUCKET}" &>/dev/null; then
   gsutil mb -l "${REGION}" -b on "gs://${CORPUS_BUCKET}"
@@ -42,7 +40,7 @@ else
   echo "✓ GCS corpus bucket exists: gs://${CORPUS_BUCKET}"
 fi
 
-# ── Upload alle kildedokumenter til GCS ──────────────────────────────────────
+# ── Upload alle kildedokumenter til GCS ─────────────────────────────────────
 UPLOAD_COUNT=0
 
 SEED_DIR="${REPO_ROOT}/docs/corpus-seed"
@@ -76,7 +74,7 @@ if [[ -f "${REPO_ROOT}/MASTERPLAN.md" ]]; then
 fi
 echo "  Totalt ${UPLOAD_COUNT} filer synkronisert"
 
-# ── REST: hent/opprett corpus (bruker project-ID for listing/oppretting) ────────
+# ── REST: hent/opprett corpus ───────────────────────────────────────────────
 TOKEN=$(gcloud auth print-access-token)
 BASE_URL="https://${RAG_REGION}-aiplatform.googleapis.com/v1beta1"
 PARENT="projects/${PROJECT_ID}/locations/${RAG_REGION}"
@@ -109,7 +107,7 @@ else
   for i in $(seq 1 30); do
     sleep 10
     LRO=$(curl -sf -H "Authorization: Bearer ${TOKEN}" \
-      "https://${RAG_REGION}-aiplatform.googleapis.com/v1beta1/${OPERATION}" 2>/dev/null || echo '{}')
+      "${BASE_URL}/${OPERATION}" 2>/dev/null || echo '{}')
     DONE=$(echo "${LRO}" | python3 -c "import sys,json; print(json.load(sys.stdin).get('done',False))" 2>/dev/null || echo 'False')
     if [[ "${DONE}" == "True" ]]; then
       CORPUS_NAME=$(curl -sf -H "Authorization: Bearer ${TOKEN}" \
@@ -128,34 +126,58 @@ for c in data.get('ragCorpora',[]):
   [[ -z "${CORPUS_NAME}" ]] && { echo "ERROR: timed out"; exit 1; }
 fi
 
-# Hent corpus-ID (siste del av ressursstien)
+# ── Import: riktig endepunkt er /ragFiles:import (bekreftet 2026-05-24) ────────
-CORPUS_ID=$(echo "${CORPUS_NAME}" | rev | cut -d'/' -f1 | rev)
+# IKKE :importRagFiles på corpus-nivå — det gir alltid 404 i europe-west4
-
+IMPORT_BASE="${BASE_URL}/${CORPUS_NAME}/ragFiles:import"
-# ── Import: bygg URL med project-NUMBER (ikke project-ID) ───────────────────────
+echo "  Import URL: ${IMPORT_BASE}"
-# importRagFiles krever project-number i stien
-IMPORT_URL="https://${RAG_REGION}-aiplatform.googleapis.com/v1beta1/projects/${PROJECT_NUMBER}/locations/${RAG_REGION}/ragCorpora/${CORPUS_ID}:importRagFiles"
-echo "  Import URL: ${IMPORT_URL}"
 
 IMPORT_OK=0
+OPS=()
 for GCS_PATH in "seed/" "docs/" "sessions/" "MASTERPLAN.md"; do
   GCS_URI="gs://${CORPUS_BUCKET}/${GCS_PATH}"
   if gsutil ls "${GCS_URI}" &>/dev/null; then
-    HTTP_STATUS=$(curl -s -o /tmp/import_response.json -w "%{http_code}" -X POST \
+    RESP=$(curl -s -o /tmp/import_resp.json -w "%{http_code}" -X POST \
       -H "Authorization: Bearer ${TOKEN}" \
       -H "Content-Type: application/json" \
-      "${IMPORT_URL}" \
+      "${IMPORT_BASE}" \
       -d '{"importRagFilesConfig":{"gcsSource":{"uris":["'"${GCS_URI}"'"]},"ragFileChunkingConfig":{"chunkSize":512,"chunkOverlap":50}}}')
-    if [[ "${HTTP_STATUS}" == "200" ]]; then
+    if [[ "${RESP}" == "200" ]]; then
-      echo "✓ Import trigget: ${GCS_URI}"
+      OP_NAME=$(python3 -c "import sys,json; print(json.load(open('/tmp/import_resp.json')).get('name',''))" 2>/dev/null || true)
+      echo "✓ Import trigget: ${GCS_URI}  [op: ${OP_NAME##*/}]"
+      OPS+=("${OP_NAME}")
       IMPORT_OK=$((IMPORT_OK + 1))
     else
-      echo "  ADVARSEL: Import feilet (HTTP ${HTTP_STATUS}): ${GCS_URI}"
+      echo "  ADVARSEL: Import feilet (HTTP ${RESP}): ${GCS_URI}"
-      cat /tmp/import_response.json 2>/dev/null || true
+      cat /tmp/import_resp.json 2>/dev/null || true
     fi
   fi
 done
 echo "  ${IMPORT_OK}/4 import-jobber trigget (async — indeksering tar noen minutter)"
 
+# ── Vent på at alle LRO-er er ferdig (valgfri polling) ──────────────────────
+if [[ ${#OPS[@]} -gt 0 ]]; then
+  echo "  Venter på indeksering..."
+  for OP in "${OPS[@]}"; do
+    for i in $(seq 1 20); do
+      sleep 15
+      LRO=$(curl -sf -H "Authorization: Bearer ${TOKEN}" \
+        "${BASE_URL}/${OP}" 2>/dev/null || echo '{}')
+      DONE=$(echo "${LRO}" | python3 -c "import sys,json; print(json.load(sys.stdin).get('done',False))" 2>/dev/null || echo 'False')
+      if [[ "${DONE}" == "True" ]]; then
+        echo "  ✓ Ferdig: ${OP##*/}"
+        break
+      fi
+      echo "  ... ${OP##*/} (${i}/20)"
+    done
+  done
+fi
+
+# ── Verifiser: tell antall indekserte filer ──────────────────────────────────
+RAG_FILE_COUNT=$(curl -sf -H "Authorization: Bearer ${TOKEN}" \
+  "${BASE_URL}/${CORPUS_NAME}/ragFiles" 2>/dev/null \
+  | python3 -c "import sys,json; d=json.load(sys.stdin); print(len(d.get('ragFiles',[])))" 2>/dev/null || echo '?')
+echo "  Indekserte RAG-filer i corpus: ${RAG_FILE_COUNT}"
+
 # ── Output ─────────────────────────────────────────────────────────────────
 echo "${CORPUS_NAME}" > /tmp/rag_corpus_name.txt
 echo ""