# 🔄 Migration Guide: Adding tenant_id to Existing Data
## Overview
After applying the security fixes, all existing data in the vector store needs `tenant_id` added to metadata.
## Option 1: Manual Migration (Recommended for Small Datasets)
### Step 1: Export Current Data
```python
# Run this script to export all chunks
from app.rag.vectorstore import get_vector_store
vector_store = get_vector_store()
all_data = vector_store.collection.get(include=["documents", "metadatas"])
# Save to JSON for backup
import json
with open("backup_chunks.json", "w") as f:
json.dump({
"ids": all_data["ids"],
"documents": all_data["documents"],
"metadatas": all_data["metadatas"]
}, f)
```
### Step 2: Add tenant_id to Metadata
```python
# For each chunk, determine tenant_id based on user_id or kb_id
# This depends on your business logic
for i, metadata in enumerate(all_data["metadatas"]):
# Determine tenant_id (example: extract from user_id or kb_id)
user_id = metadata.get("user_id", "")
kb_id = metadata.get("kb_id", "")
# YOUR LOGIC HERE: Map user_id/kb_id to tenant_id
tenant_id = determine_tenant_id(user_id, kb_id) # Implement this
metadata["tenant_id"] = tenant_id
```
### Step 3: Re-import with tenant_id
```python
# Delete old collection
vector_store.clear_collection()
# Re-add with tenant_id
vector_store.add_documents(
documents=all_data["documents"],
embeddings=[...], # Re-generate or store separately
metadatas=all_data["metadatas"], # Now includes tenant_id
ids=all_data["ids"]
)
```
## Option 2: Gradual Migration (For Large Datasets)
1. Keep old data accessible temporarily
2. New uploads include tenant_id
3. Migrate old data in batches
4. Verify no data loss
5. Remove old data
## Option 3: Fresh Start (If Data is Test/Dev)
Simply clear the collection and re-upload all documents with tenant_id:
```python
from app.rag.vectorstore import get_vector_store
vector_store = get_vector_store()
vector_store.clear_collection()
```
Then re-upload all documents through the API with tenant_id.
## Verification Script
```python
# Verify all chunks have tenant_id
from app.rag.vectorstore import get_vector_store
vector_store = get_vector_store()
all_data = vector_store.collection.get(include=["metadatas"])
missing_tenant = []
for i, metadata in enumerate(all_data["metadatas"]):
if "tenant_id" not in metadata:
missing_tenant.append(all_data["ids"][i])
if missing_tenant:
print(f"⚠️ {len(missing_tenant)} chunks missing tenant_id")
else:
print("✅ All chunks have tenant_id")
```
