Migrating Weaviate Data

migrating-weaviate-data.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Migrating data in Weaviate\n",
    "\n",
    "Note this method describes a way to migrate data using raw queries and is intended for relatively small datasets. There are a couple other options:\n",
    "\n",
    "1. Use of backups feature https://weaviate.io/developers/weaviate/current/configuration/backups.html.\n",
    "\n",
    "2. Use of upcoming scroll api https://github.com/semi-technologies/weaviate/issues/2302. This feature is planned for 1.18 (so relatively soon) and will remove the limit of query maximum results below allowing for paging out efficiently millions of records.\n",
    "\n",
    "## Process\n",
    "\n",
    "1. Increase your `QUERY_MAXIMUM_RESULTS` size of your source environment. This is set as an environment variable in `docker-compose.yml`. This is needed for large pagination.\n",
    "\n",
    "```yaml\n",
    "QUERY_MAXIMUM_RESULTS: 100000\n",
    "```\n",
    "\n",
    "### Retrieve all data from existing class and write to new class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "import weaviate\n",
    "from tqdm import tqdm\n",
    "\n",
    "client = weaviate.Client(\"http://localhost:8080\")\n",
    "LIMIT = 100000\n",
    "\n",
    "def pull_data_for_class(client: weaviate.Client, class_name: str):\n",
    "    schema = client.schema.get(class_name)\n",
    "    fields = [field[\"name\"] for field in client.schema.get(class_name=class_name)[\"properties\"]]\n",
    "    data = client.query.get(class_name=class_name, properties=fields).with_additional(['vector', 'id']).with_limit(LIMIT).do()\n",
    "    if \"errors\" in data:\n",
    "        raise Exception(data[\"errors\"])\n",
    "    return schema, data[\"data\"][\"Get\"][class_name]\n",
    "\n",
    "def push_data(client: weaviate.Client, class_name: str, schema : dict, data: list):\n",
    "    schema[\"class\"] = class_name\n",
    "    client.schema.create_class(schema)\n",
    "    with client.batch as batch:\n",
    "        for record in tqdm(data):\n",
    "            vector = record[\"_additional\"][\"vector\"]\n",
    "            uuid = record[\"_additional\"][\"id\"]\n",
    "            del record[\"_additional\"]\n",
    "            if vector == []:\n",
    "                batch.add_data_object(data_object=record, class_name=class_name, uuid=uuid)\n",
    "            else:\n",
    "                batch.add_data_object(data_object=record, class_name=class_name, uuid=uuid, vector=vector)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 100000/100000 [00:00<00:00, 285620.39it/s]\n"
     ]
    }
   ],
   "source": [
    "schema, data = pull_data_for_class(client, \"Item2\")\n",
    "\n",
    "# write to new client to migrate data\n",
    "push_data(client, \"Item4\", schema, data)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.0"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "5c7b89af1651d0b8571dde13640ecdccf7d5a6204171d6ab33e7c296e100e08a"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}