> ## Documentation Index
> Fetch the complete documentation index at: https://docs-v1.latitude.so/llms.txt
> Use this file to discover all available pages before exploring further.

# RAG retrieval

> Learn how to use RAG retrieval with the Latitude SDK

## Prompt

In your prompt, you can define a tool that will be used to retrieve information.

<CodeGroup>
  ```markdown example theme={null}
  ---
  provider: Latitude
  model: gpt-4o-mini
  temperature: 0.7
  tools:
    - get_answer:
        description: Ask this tool for the answer when user do a question.
        parameters:
          type: object
          additionalProperties: false
          required: ['question']
          properties:
            question:
              type: string
              description: Question to ask
  ---

  Give user's question {{ question }} a concise answer.
  ```
</CodeGroup>

## Code

Performing RAG retrieval with the Latitude SDK simply involves defining a tool in your prompt. In your code, you can then get the results from the RAG solution you use.

<CodeGroup>
  ```typescript Typescript theme={null}
  import { Pinecone } from '@pinecone-database/pinecone'
  import OpenAI from 'openai'
  import { Latitude } from '@latitude-data/sdk'

  type Tools = { get_answer: { question: string } }

  const PINECODE_INDEX_NAME = 'geography-quizz-index'

  async function run() {
    const sdk = new Latitude(process.env.LATITUDE_API_KEY, {
      projectId: Number(process.env.PROJECT_ID),
      versionUuid: 'live',
    })
    const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY })
    const pinecone = new Pinecone({ apiKey: process.env.PINECONE_API_KEY })
    const pc = pinecone.Index(PINECODE_INDEX_NAME)

    const question = 'What is the deepest ocean in the world?'

    console.log('Question: ', question)
    console.log('\nSearching...\n')

    const result = await sdk.prompts.run<Tools>('rag-retrieval/example', {
      parameters: { question },
      tools: {
        get_answer: async ({ question }) => {
          // Do the embedding
          const embedding = await openai.embeddings
            .create({
              input: question,
              model: 'text-embedding-3-small',
            })
            .then((res) => res.data[0].embedding)

          // Query your RAG backend. In this case, Pinecone
          const queryResponse = await pc.query({
            vector: embedding,
            topK: 10,
            includeMetadata: true,
          })

          const first = queryResponse.matches[0]
          return first?.metadata?.answer
        },
      },
    })

    console.log('Answer: ', result.response.text)
  }

  run()
  ```
</CodeGroup>