fix: 07-rag-setup.sh — bruk project-number (ikke project-ID) i importRagFiles URL

infrastructure/07-rag-setup.sh

@@ -28,6 +28,11 @@ gcloud services enable aiplatform.googleapis.com \
   --project="${PROJECT_ID}" --quiet
 echo "✓ APIs enabled"
 
+# ── Hent project-number (kræves av importRagFiles-endepunktet) ────────────────
+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://${CORPUS_BUCKET}" &>/dev/null; then
@@ -69,10 +74,9 @@ if [[ -f "${REPO_ROOT}/MASTERPLAN.md" ]]; then
   echo "✓ MASTERPLAN.md → gs://${CORPUS_BUCKET}/MASTERPLAN.md"
   UPLOAD_COUNT=$((UPLOAD_COUNT + 1))
 fi
-
 echo "  Totalt ${UPLOAD_COUNT} filer synkronisert"
 
-# ── REST: hent/opprett corpus ───────────────────────────────────────────────────
+# ── REST: hent/opprett corpus (bruker project-ID for listing/oppretting) ────────
 TOKEN=$(gcloud auth print-access-token)
 BASE_URL="https://${RAG_REGION}-aiplatform.googleapis.com/v1beta1"
 PARENT="projects/${PROJECT_ID}/locations/${RAG_REGION}"
@@ -99,10 +103,8 @@ else
     -H "Content-Type: application/json" \
     "${BASE_URL}/${PARENT}/ragCorpora" \
     -d '{"displayName":"'"${RAG_CORPUS_DISPLAY_NAME}"'","ragEmbeddingModelConfig":{"vertexPredictionEndpoint":{"publisherModel":"publishers/google/models/text-embedding-005"}},"ragVectorDbConfig":{"ragManagedDb":{}}}')
-
   OPERATION=$(echo "${RESPONSE}" | python3 -c "import sys,json; print(json.load(sys.stdin).get('name',''))" 2>/dev/null || true)
   [[ -z "${OPERATION}" ]] && { echo "ERROR: ${RESPONSE}"; exit 1; }
-
   echo "  Venter på LRO..."
   for i in $(seq 1 30); do
     sleep 10
@@ -126,15 +128,19 @@ for c in data.get('ragCorpora',[]):
   [[ -z "${CORPUS_NAME}" ]] && { echo "ERROR: timed out"; exit 1; }
 fi
 
-# ── Import: CORPUS_NAME er full ressurssti — bruk direkte som URL-path ─────────
-# Korrekt URL: https://{region}-aiplatform.googleapis.com/v1beta1/{corpus_name}:importRagFiles
-IMPORT_URL="https://${RAG_REGION}-aiplatform.googleapis.com/v1beta1/${CORPUS_NAME}:importRagFiles"
+# Hent corpus-ID (siste del av ressursstien)
+CORPUS_ID=$(echo "${CORPUS_NAME}" | rev | cut -d'/' -f1 | rev)
+
+# ── Import: bygg URL med project-NUMBER (ikke project-ID) ───────────────────────
+# 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
 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 /dev/null -w "%{http_code}" -X POST \
+    HTTP_STATUS=$(curl -s -o /tmp/import_response.json -w "%{http_code}" -X POST \
       -H "Authorization: Bearer ${TOKEN}" \
       -H "Content-Type: application/json" \
       "${IMPORT_URL}" \
@@ -144,10 +150,11 @@ for GCS_PATH in "seed/" "docs/" "sessions/" "MASTERPLAN.md"; do
       IMPORT_OK=$((IMPORT_OK + 1))
     else
       echo "  ADVARSEL: Import feilet (HTTP ${HTTP_STATUS}): ${GCS_URI}"
+      cat /tmp/import_response.json 2>/dev/null || true
     fi
   fi
 done
-echo "  ${IMPORT_OK} import-jobber trigget (async — indeksering tar noen minutter)"
+echo "  ${IMPORT_OK}/4 import-jobber trigget (async — indeksering tar noen minutter)"
 
 # ── Output ─────────────────────────────────────────────────────────────────
 echo "${CORPUS_NAME}" > /tmp/rag_corpus_name.txt