Datasets and RAG

Datasets in Mira Flows support Retrieval-Augmented Generation (RAG) to enhance flows with specific knowledge. This capability allows flows to leverage custom knowledge bases for improved accuracy and contextual awareness in their responses. 🤖📈

Dataset Attributes

Component	Description	Required	Example
Name	Unique identifier for dataset	Yes	`"author/dataset_name"`
Description	Purpose and content overview	Yes	`"Optional description"`
Source Type	Type of data being added	Yes	`"url"` or `"file_path"`
Source Path	Location of source data	Yes	`"example.com"` or `"path/to/file"`
Author	Creator’s username	Yes	`"your-username"`

Supported File Formats

Format	Description	Processing Method
PDF (.pdf)	Document files	Text extraction from document
Markdown (.md)	Formatted text	Text extraction from document
URL	Web content	Web content scraping
CSV (.csv)	URL listings	URL extraction and content scraping
Text (.txt)	Plain text	Direct text extraction

Creating and Configuring Datasets

Creating a Dataset

Python

from mira_sdk import MiraClient

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

# Create dataset
client.dataset.create("author/dataset_name", "Optional description")

Adding Data Sources

Python

# Add URL to your dataset
client.dataset.add_source("author/dataset_name", url="example.com")

# Add file to your dataset
client.dataset.add_source("author/dataset_name", file_path="path/to/my/file.csv")

Linking Dataset with Flow

Add the following configuration to your flow.yaml file:

.yaml

dataset:
  source: "author/dataset_name"

Get Started

Core Concepts

How To Guides

Datasets and RAG

Dataset Attributes

Supported File Formats

Creating and Configuring Datasets

Creating a Dataset

Adding Data Sources

Linking Dataset with Flow

​Dataset Attributes

​Supported File Formats

​Creating and Configuring Datasets

​Creating a Dataset

​Adding Data Sources

​Linking Dataset with Flow

Dataset Attributes

Supported File Formats

Creating and Configuring Datasets

Creating a Dataset

Adding Data Sources

Linking Dataset with Flow