feat: 07-rag-setup.sh — upload alle docs/*.md + corpus-seed/*.md til GCS automatisk
This commit is contained in:
parent
b2323fe476
commit
06767f8806
|
|
@ -2,6 +2,7 @@
|
|||
# 07-rag-setup.sh — Create Vertex AI RAG Engine corpus in Serverless mode
|
||||
# Uses REST API directly — no SDK version dependency
|
||||
# RAG_REGION defaults to europe-west4 (Serverless available, no allowlist)
|
||||
# Uploads ALL docs/*.md + docs/corpus-seed/*.md to GCS on every run
|
||||
# Idempotent — safe to run multiple times
|
||||
# Source .env before running: source .env
|
||||
|
||||
|
|
@ -14,6 +15,8 @@ set -euo pipefail
|
|||
# europe-west4: Serverless RAG available, no allowlist, gemini-2.0-flash + 2.5-pro present
|
||||
RAG_REGION="${RAG_REGION:-europe-west4}"
|
||||
|
||||
REPO_ROOT="$(cd "$(dirname "$0")/.." && pwd)"
|
||||
|
||||
echo "=== 07: Setting up Vertex AI RAG Engine (Serverless mode) ==="
|
||||
echo " Project : ${PROJECT_ID}"
|
||||
echo " Region : ${RAG_REGION} (override with RAG_REGION= if needed)"
|
||||
|
|
@ -36,21 +39,57 @@ else
|
|||
echo "✓ GCS corpus bucket exists: gs://${CORPUS_BUCKET}"
|
||||
fi
|
||||
|
||||
# ── Upload seed docs ──────────────────────────────────────────────────────────
|
||||
SEED_DIR="$(dirname "$0")/../docs/corpus-seed"
|
||||
if [[ -d "${SEED_DIR}" ]] && ls "${SEED_DIR}"/*.md &>/dev/null; then
|
||||
# ── Upload docs: corpus-seed/*.md + docs/*.md (top-level only) ───────────────────
|
||||
UPLOAD_COUNT=0
|
||||
|
||||
# 1. corpus-seed/ — kuraterte seed-filer
|
||||
SEED_DIR="${REPO_ROOT}/docs/corpus-seed"
|
||||
if [[ -d "${SEED_DIR}" ]] && compgen -G "${SEED_DIR}/*.md" > /dev/null 2>&1; then
|
||||
gsutil -m cp "${SEED_DIR}"/*.md "gs://${CORPUS_BUCKET}/seed/" 2>/dev/null || true
|
||||
echo "✓ Seed documents uploaded to gs://${CORPUS_BUCKET}/seed/"
|
||||
COUNT=$(ls "${SEED_DIR}"/*.md 2>/dev/null | wc -l | tr -d ' ')
|
||||
echo "✓ corpus-seed: ${COUNT} filer lastet opp til gs://${CORPUS_BUCKET}/seed/"
|
||||
UPLOAD_COUNT=$((UPLOAD_COUNT + COUNT))
|
||||
else
|
||||
echo " No seed documents in docs/corpus-seed/ — skipping"
|
||||
echo " Ingen filer i docs/corpus-seed/ — hopper over"
|
||||
fi
|
||||
|
||||
# ── REST: check if corpus exists ───────────────────────────────────────────
|
||||
# 2. docs/*.md — alle toppnivå docs (GCP-kunnskap, arkitektur, regler osv.)
|
||||
DOCS_DIR="${REPO_ROOT}/docs"
|
||||
if compgen -G "${DOCS_DIR}/*.md" > /dev/null 2>&1; then
|
||||
gsutil -m cp "${DOCS_DIR}"/*.md "gs://${CORPUS_BUCKET}/docs/" 2>/dev/null || true
|
||||
COUNT=$(ls "${DOCS_DIR}"/*.md 2>/dev/null | wc -l | tr -d ' ')
|
||||
echo "✓ docs/: ${COUNT} filer lastet opp til gs://${CORPUS_BUCKET}/docs/"
|
||||
UPLOAD_COUNT=$((UPLOAD_COUNT + COUNT))
|
||||
else
|
||||
echo " Ingen .md-filer i docs/ — hopper over"
|
||||
fi
|
||||
|
||||
# 3. protocols/sessions/*.md — sesjonlogger (hele reisen)
|
||||
SESSIONS_DIR="${REPO_ROOT}/protocols/sessions"
|
||||
if [[ -d "${SESSIONS_DIR}" ]] && compgen -G "${SESSIONS_DIR}/*.md" > /dev/null 2>&1; then
|
||||
gsutil -m cp "${SESSIONS_DIR}"/*.md "gs://${CORPUS_BUCKET}/sessions/" 2>/dev/null || true
|
||||
COUNT=$(ls "${SESSIONS_DIR}"/*.md 2>/dev/null | wc -l | tr -d ' ')
|
||||
echo "✓ protocols/sessions/: ${COUNT} filer lastet opp til gs://${CORPUS_BUCKET}/sessions/"
|
||||
UPLOAD_COUNT=$((UPLOAD_COUNT + COUNT))
|
||||
else
|
||||
echo " Ingen sesjonlogger i protocols/sessions/ — hopper over"
|
||||
fi
|
||||
|
||||
# 4. MASTERPLAN.md (rot)
|
||||
if [[ -f "${REPO_ROOT}/MASTERPLAN.md" ]]; then
|
||||
gsutil cp "${REPO_ROOT}/MASTERPLAN.md" "gs://${CORPUS_BUCKET}/" 2>/dev/null || true
|
||||
echo "✓ MASTERPLAN.md lastet opp"
|
||||
UPLOAD_COUNT=$((UPLOAD_COUNT + 1))
|
||||
fi
|
||||
|
||||
echo " Totalt ${UPLOAD_COUNT} filer synkronisert til gs://${CORPUS_BUCKET}/"
|
||||
|
||||
# ── REST: token + endpoint ────────────────────────────────────────────────────
|
||||
TOKEN=$(gcloud auth print-access-token)
|
||||
API="https://${RAG_REGION}-aiplatform.googleapis.com/v1beta1"
|
||||
PARENT="projects/${PROJECT_ID}/locations/${RAG_REGION}"
|
||||
|
||||
echo " Checking for existing corpus..."
|
||||
echo " Sjekker om corpus finnes..."
|
||||
EXISTING=$(curl -sf -H "Authorization: Bearer ${TOKEN}" \
|
||||
"${API}/${PARENT}/ragCorpora" 2>/dev/null || echo '{}')
|
||||
|
||||
|
|
@ -64,9 +103,9 @@ for c in data.get('ragCorpora', []):
|
|||
" 2>/dev/null || true)
|
||||
|
||||
if [[ -n "${CORPUS_NAME}" ]]; then
|
||||
echo "✓ RAG corpus already exists: ${CORPUS_NAME}"
|
||||
echo "✓ RAG corpus finnes allerede: ${CORPUS_NAME}"
|
||||
else
|
||||
echo " Creating corpus '${RAG_CORPUS_DISPLAY_NAME}' in ${RAG_REGION} (Serverless/REST)..."
|
||||
echo " Oppretter corpus '${RAG_CORPUS_DISPLAY_NAME}' i ${RAG_REGION} (Serverless/REST)..."
|
||||
|
||||
RESPONSE=$(curl -sf -X POST \
|
||||
-H "Authorization: Bearer ${TOKEN}" \
|
||||
|
|
@ -84,16 +123,15 @@ else
|
|||
}
|
||||
}')
|
||||
|
||||
# REST create returns an LRO — poll until done
|
||||
OPERATION=$(echo "${RESPONSE}" | python3 -c "import sys,json; print(json.load(sys.stdin).get('name',''))" 2>/dev/null || true)
|
||||
|
||||
if [[ -z "${OPERATION}" ]]; then
|
||||
echo "ERROR: No operation returned. Response:"
|
||||
echo "ERROR: Ingen operation returnert. Response:"
|
||||
echo "${RESPONSE}"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
echo " Waiting for LRO: ${OPERATION}"
|
||||
echo " Venter på LRO: ${OPERATION}"
|
||||
for i in $(seq 1 30); do
|
||||
sleep 10
|
||||
LRO=$(curl -sf -H "Authorization: Bearer ${TOKEN}" \
|
||||
|
|
@ -103,9 +141,8 @@ else
|
|||
CORPUS_NAME=$(echo "${LRO}" | python3 -c "
|
||||
import sys, json
|
||||
d = json.load(sys.stdin)
|
||||
print(d.get('response', {}).get('name', '') or d.get('metadata', {}).get('genericMetadata', {}).get('updateTime', ''))
|
||||
print(d.get('response', {}).get('name', ''))
|
||||
" 2>/dev/null || true)
|
||||
# Fallback: re-list to get name
|
||||
if [[ -z "${CORPUS_NAME}" ]] || [[ "${CORPUS_NAME}" != *ragCorpora* ]]; then
|
||||
CORPUS_NAME=$(curl -sf -H "Authorization: Bearer ${TOKEN}" \
|
||||
"${API}/${PARENT}/ragCorpora" 2>/dev/null \
|
||||
|
|
@ -118,32 +155,39 @@ for c in data.get('ragCorpora', []):
|
|||
break
|
||||
" 2>/dev/null || true)
|
||||
fi
|
||||
echo "✓ RAG corpus created: ${CORPUS_NAME}"
|
||||
echo "✓ RAG corpus opprettet: ${CORPUS_NAME}"
|
||||
break
|
||||
fi
|
||||
echo " ... still waiting (${i}/30)"
|
||||
echo " ... venter (${i}/30)"
|
||||
done
|
||||
|
||||
if [[ -z "${CORPUS_NAME}" ]]; then
|
||||
echo "ERROR: Corpus creation timed out or failed."
|
||||
echo "ERROR: Corpus-oppretting timet ut eller feilet."
|
||||
exit 1
|
||||
fi
|
||||
fi
|
||||
|
||||
# ── Import seed docs via REST ─────────────────────────────────────────────
|
||||
GCS_URI="gs://${CORPUS_BUCKET}/seed/"
|
||||
curl -sf -X POST \
|
||||
-H "Authorization: Bearer ${TOKEN}" \
|
||||
-H "Content-Type: application/json" \
|
||||
"${API}/${CORPUS_NAME}:importRagFiles" \
|
||||
-d '{
|
||||
"importRagFilesConfig": {
|
||||
"gcsSource": { "uris": ["'"${GCS_URI}"'"] },
|
||||
"ragFileChunkingConfig": { "chunkSize": 512, "chunkOverlap": 50 }
|
||||
}
|
||||
}' &>/dev/null && echo "✓ Seed import triggered (async)" || echo " WARNING: Seed import skipped (non-fatal)"
|
||||
# ── Import alle GCS-mapper til corpus ───────────────────────────────────────────
|
||||
for GCS_PATH in "seed/" "docs/" "sessions/" ""; do
|
||||
GCS_URI="gs://${CORPUS_BUCKET}/${GCS_PATH}"
|
||||
# Sjekk at det finnes noe der
|
||||
if gsutil ls "${GCS_URI}" &>/dev/null; then
|
||||
curl -sf -X POST \
|
||||
-H "Authorization: Bearer ${TOKEN}" \
|
||||
-H "Content-Type: application/json" \
|
||||
"${API}/${CORPUS_NAME}:importRagFiles" \
|
||||
-d '{
|
||||
"importRagFilesConfig": {
|
||||
"gcsSource": { "uris": ["'"${GCS_URI}"'"] },
|
||||
"ragFileChunkingConfig": { "chunkSize": 512, "chunkOverlap": 50 }
|
||||
}
|
||||
}' &>/dev/null \
|
||||
&& echo "✓ Import trigget (async): ${GCS_URI}" \
|
||||
|| echo " ADVARSEL: Import hoppet over (ikke fatal): ${GCS_URI}"
|
||||
fi
|
||||
done
|
||||
|
||||
# ── Write output ──────────────────────────────────────────────────────────────────
|
||||
# ── Skriv output ──────────────────────────────────────────────────────────────────
|
||||
echo "${CORPUS_NAME}" > /tmp/rag_corpus_name.txt
|
||||
|
||||
echo ""
|
||||
|
|
|
|||
Loading…
Reference in New Issue
Block a user