Chapter 2: Building the Knowledge Base (RAG)

In this chapter, we will build the “brain” of our Level 1 Agent, which will answer general questions. This process converts a document into a database that the AI can consult.

Required Components

• Google Drive File Manager
• Parse Data
• Split Text
• Google Generative AI Embeddings
• Chroma DB

---

Step-by-Step Construction and Configuration

Add the following components to the canvas and connect them as indicated:

1. Google Drive File Manager

   • Purpose: To load the document with your company’s policies.
   • Configuration:
     • Credential: Select My Drive Credential.
     • Operation: Get.
     • Mode of Input: By Selection.
     • Select file: Choose the PDF or document containing your information.

2. Parse Data

   • Purpose: To extract the text from the document.
   • Connection: Connect the Data output of the Google Drive to the data input of this component.

3. Split Text

   • Purpose: To divide the text into manageable fragments (chunks).
   • Connection: Connect the processed_data output of the Parse Data to the data input.
   • Configuration:
     • Chunk Size: 1000.
     • Chunk Overlap: 200.

4. Google Generative AI Embeddings

   • Purpose: To prepare the model to convert text into vectors (embeddings).
   • Configuration:
     • Credential: Select Mi Credencial de Gemini.

5. Chroma DB

   • Purpose: To store the text fragments as vectors in the Vector Store.
   • Connections:
     • Connect the chunks output of the Split Text to the ingest_data input.
     • Connect the embeddings output of the Google Generative AI Embeddings to the embedding input.
   • Configuration:
     • Operation: Add.
     • Collection Name: ecommerce_policies (save this name, you will need it).

By running this flow, you will have created your knowledge base.