​
Video Tutorial


Check out this Github Repository for a quick overview on how to Implement RAG Capabilities.

​
Understanding RAG Implementation

Retrieval-Augmented Generation (RAG) enhances your flows with specific domain knowledge. The implementation process involves three main stages:
  1. 🗂️ Create a Dataset: Establish a knowledge base that contains specialized information your flow will reference during execution.
  2. ➕ Add Data Sources: Populate your dataset with relevant information from various supported sources.
  3. 🔗 Link Dataset to Flow: Connect the dataset to your flow, enabling it to leverage this information during processing.

​
Creating Your Dataset

Begin by establishing your knowledge base:
Python
from mira_sdk import MiraClient

client = MiraClient(config={"API_KEY": "YOUR_API_KEY"})        # Initialize client

# Create a new dataset
client.dataset.create(
    "author/dataset_name",                                     # Unique identifier
    "Description of your knowledge base"                       # Dataset purpose
)

​
Adding Data Sources

Populate your dataset with information from various supported formats:
Python
# Add a PDF document as data source
client.dataset.add_source("author/dataset_name", file_path="document.pdf")

# Add and URL as data source
client.dataset.add_source("author/dataset_name", url="https://example.com/data")

# Add multiple URL sources via a CSV file
client.dataset.add_source("author/dataset_name", file_path="sources.csv")

​
Linking Dataset to Flow

Connect your dataset to an existing flow by modifying its configuration:
.yaml
# Additional flow configuration with RAG

dataset:
  source: "author/dataset_name"         # Link to your dataset

# Rest of your flow configuration remains unchanged

​
Example

Implement RAG capabilities

Complete example showcasing how to set up and configure datasets for Retrieval Augmented Generation (RAG) capabilities.