langchain/docs/docs/how_to/vectorstore_retriever.ipynb

{
 "cells": [
  {
   "cell_type": "raw",
   "id": "ee14951b",
   "metadata": {},
   "source": [
    "---\n",
    "sidebar_position: 0\n",
    "---"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "105cddce",
   "metadata": {},
   "source": [
    "# How to use a vectorstore as a retriever\n",
    "\n",
    "A vector store retriever is a retriever that uses a vector store to retrieve documents. It is a lightweight wrapper around the vector store class to make it conform to the retriever interface.\n",
    "It uses the search methods implemented by a vector store, like similarity search and MMR, to query the texts in the vector store.\n",
    "\n",
    "In this guide we will cover:\n",
    "\n",
    "1. How to instantiate a retriever from a vectorstore;\n",
    "2. How to specify the search type for the retriever;\n",
    "3. How to specify additional search parameters, such as threshold scores and top-k.\n",
    "\n",
    "## Creating a retriever from a vectorstore\n",
    "\n",
    "You can build a retriever from a vectorstore using its [.as_retriever](https://python.langchain.com/v0.2/api_reference/core/vectorstores/langchain_core.vectorstores.VectorStore.html#langchain_core.vectorstores.VectorStore.as_retriever) method. Let's walk through an example.\n",
    "\n",
    "First we instantiate a vectorstore. We will use an in-memory [FAISS](https://python.langchain.com/v0.2/api_reference/community/vectorstores/langchain_community.vectorstores.faiss.FAISS.html) vectorstore:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "174e3c69",
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain_community.document_loaders import TextLoader\n",
    "from langchain_community.vectorstores import FAISS\n",
    "from langchain_openai import OpenAIEmbeddings\n",
    "from langchain_text_splitters import CharacterTextSplitter\n",
    "\n",
    "loader = TextLoader(\"state_of_the_union.txt\")\n",
    "\n",
    "documents = loader.load()\n",
    "text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0)\n",
    "texts = text_splitter.split_documents(documents)\n",
    "embeddings = OpenAIEmbeddings()\n",
    "vectorstore = FAISS.from_documents(texts, embeddings)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6f6e65a1-5eb4-4165-b06b-9bb40624a8d8",
   "metadata": {},
   "source": [
    "We can then instantiate a retriever:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "52df5f55",
   "metadata": {},
   "outputs": [],
   "source": [
    "retriever = vectorstore.as_retriever()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "08f8b820-5912-49c1-9d76-40be0571dffb",
   "metadata": {},
   "source": [
    "This creates a retriever (specifically a [VectorStoreRetriever](https://python.langchain.com/v0.2/api_reference/core/vectorstores/langchain_core.vectorstores.VectorStoreRetriever.html)), which we can use in the usual way:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "32334fda",
   "metadata": {},
   "outputs": [],
   "source": [
    "docs = retriever.invoke(\"what did the president say about ketanji brown jackson?\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fd7b19f0",
   "metadata": {},
   "source": [
    "## Maximum marginal relevance retrieval\n",
    "By default, the vector store retriever uses similarity search. If the underlying vector store supports maximum marginal relevance search, you can specify that as the search type.\n",
    "\n",
    "This effectively specifies what method on the underlying vectorstore is used (e.g., `similarity_search`, `max_marginal_relevance_search`, etc.)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "b286ac04",
   "metadata": {},
   "outputs": [],
   "source": [
    "retriever = vectorstore.as_retriever(search_type=\"mmr\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "07f937f7",
   "metadata": {},
   "outputs": [],
   "source": [
    "docs = retriever.invoke(\"what did the president say about ketanji brown jackson?\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6ce77789",
   "metadata": {},
   "source": [
    "## Passing search parameters\n",
    "\n",
    "We can pass parameters to the underlying vectorstore's search methods using `search_kwargs`.\n",
    "\n",
    "### Similarity score threshold retrieval\n",
    "\n",
    "For example, we can set a similarity score threshold and only return documents with a score above that threshold."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "dbb38a03",
   "metadata": {},
   "outputs": [],
   "source": [
    "retriever = vectorstore.as_retriever(\n",
    "    search_type=\"similarity_score_threshold\", search_kwargs={\"score_threshold\": 0.5}\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "56f6c9ae",
   "metadata": {},
   "outputs": [],
   "source": [
    "docs = retriever.invoke(\"what did the president say about ketanji brown jackson?\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "329f5b26",
   "metadata": {},
   "source": [
    "### Specifying top k\n",
    "\n",
    "We can also limit the number of documents `k` returned by the retriever."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "d712c91d",
   "metadata": {},
   "outputs": [],
   "source": [
    "retriever = vectorstore.as_retriever(search_kwargs={\"k\": 1})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "a79b573b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "docs = retriever.invoke(\"what did the president say about ketanji brown jackson?\")\n",
    "len(docs)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.10.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}