AI Refinery™¶

AI Refinery™ by Accenture is an AI platform for developing and executing AI multi-agent solutions. It is designed to help organizations:

Adopt and customize large language models (LLMs) to meet specific business needs.
Integrate generative AI across various enterprise functions using a robust AI stack.
Foster continuous innovation with minimal human intervention.

This platform ensures seamless integration and ongoing advancements in AI capabilities within your organization.

Flexible Agentic Teams

Enable agents to autonomously perform tasks, make decisions, and interact with other agents and systems.

Explore Agent Library
Comprehensive Model Catalog

Our diverse model catalog includes LLMs, VLLMs, rerankers, and more. Choose models to power your agents.

View Model Catalog
Scalable Distiller Framework

Designed to streamline complex workflows by orchestrating various agents that handle different tasks.

Learn More
Agent Memory

Allow your agents to retain context, personalize interactions, and provide coherent responses over time.

View Memory Tutorial

What AI Refinery Offers¶

AI Refinery offers a powerful suite of AI tools to enhance productivity and innovation. It offers various agents, customizable integrations, a robust model catalog, and essential safety features. Advanced functionalities like self-reflection and shared memory optimize efficiency, while APIs for audio, chat, embedding, and knowledge etc., ensure seamless workflow integration. AI Refinery empowers users to effectively leverage AI with secure and ethical standards.

Agentic Core Features¶

Built-in Utility Agents¶

AI Refinery offers a diverse selection of built-in utility agents engineered to streamline tasks such as Retrieval Augmented Generation (RAG), data analytics, and image generation. These ready-to-deploy agents optimize workflows, enabling you to achieve project goals more efficiently. Below, you'll find a straightforward example demonstrating how easily you can configure these powerful agents using YAML and deploy them using minimal Python code.

The YAML configuration for a sample project that employs SearchAgent to perform web searches and respond to user queries:

utility_agents: # configure your utility agents in this list
  - agent_class: SearchAgent # The class of the agent
    agent_name: "Search Agent" # A name that you choose for the agent

orchestrator:
  agent_list: # list the configured agents here
    - agent_name: "Search Agent"

The python code to deploy the agent based on this YAML configuration:

import asyncio
import os

from air import DistillerClient, login
from dotenv import load_dotenv

load_dotenv() # This loads your ACCOUNT and API_KEY from your local '.env' file

# login to AI Refinery with your credentials
login(
    account=str(os.getenv("ACCOUNT")),
    api_key=str(os.getenv("API_KEY")),
)

async def search_demo():
    # Instantiate a DistillerClient to communicate
    # with AI Refinery platform
    distiller_client = DistillerClient()

    # upload your config file to register a new AI Refinery project
    distiller_client.create_project(config_path="example.yaml", project="example") # assuming the YAML is stored in the file "example.yaml"

    # connect to the created project
    async with distiller_client(
        project="example",
        uuid="test_user"
    ) as dc:
        responses = await dc.query(query="Who won the FIFA world cup 2022?") # send a query to project
        async for response in responses:
            print(response['content']) 

if __name__ == "__main__":
    asyncio.run(search_demo())

The example demonstrates a single agent in the project. However, you can set up additional agents under utility_agents and include them in the agent_list within the orchestrator section to develop a multi-agent solution.

The list of currently available built-in utility agents can be seen below:

Analytics Agent: Streamlines data analysis tasks for insightful decision-making.
Author Agent: Enhances writing processes with AI-driven content creation.
Composite Agent: Combines responses from multiple agents into a comprehensive answer.
Critical Thinker Agent: Analyzes conversations to identify issues and provide insights.
Image Generation Agent: Creates high-quality images (both text-to-image & image-to-image).
Image Understanding Agent: Analyzes and interprets visual data for deeper insights.
MCP Agent: Integrates Model Context Protocol (MCP) support for dynamic tool discovery and invocation via MCP servers.
Planning Agent: Designs realistic plans by analyzing user interactions and goals.
Research Agent: Handles complex queries using RAG via web search and vector search methods.
Search Agent: Answers queries by searching the internet, specifically using Google.
Tool Use Agent: Interacts with external tools to perform tasks and deliver results.

Built-in Super Agents¶

Super Agents are engineered to handle complex tasks by orchestrating multiple utility agents, creating dynamic and powerful collaborations. Explore the built-in Super Agents in the AI Refinery SDK:

Base Super Agent: Decomposes a complex task into several subtasks, assigning each to the appropriate utility agents.
Flow Super Agent: Executes a deterministic workflow configured by the user among utility agents.
Evaluation Super Agent: Systematically assesses the performance of utility agents based on predefined metrics and sample queries, providing a structured approach to improving agent performance.

Trusted Agent Huddle¶

Explore third-party integrations to extend the AI Refinery SDK's capabilities. By incorporating third-party agents into the agentic workflow, users can enhance their existing systems alongside other utility agents, creating more robust and versatile AI solutions:

Amazon Bedrock Agent: Hosted on AWS, it utilizes the reasoning of foundation models, APIs, and data to break down user requests, gathers relevant information, and efficiently completes tasks.
Azure AI Agent: Cloud-hosted on Microsoft Azure, it interprets queries, invokes tools, executes tasks, and returns results.
Salesforce Agent: Hosted on Salesforce, it routes cases, provides order details, extends databases, and responds to queries.
Writer AI Agent: From Writer.com, it generates, refines, and structures content using integrated tools and customizable guidelines.

Custom Agent Integration¶

A custom agent is defined as a Python function that processes a string query to return an output. You can design your custom agent to handle tasks from simple LLM responses to complex operations like using RAG for data gathering and analytics. These agents can be seamlessly integrated into AI Refinery's multi-agent workflow alongside other utility and super agents.

Custom Tools Integration¶

Easily integrate your custom Python functions as tools to be used by the Tool Use Agent. This flexibility empowers users to extend the agent's capabilities, from simple calculations to complex data processing, by incorporating their own functions alongside built-in tools.

Comprehensive Model Catalog¶

Our model catalog offers a wide range of AI solutions for text, image, and speech processing. Models can be used through the agentic workflow or directly via API calls. Below is a list of the model types currently available:

LLMs & VLMs: For text and image input processing.
Embedding Models: For embedding textual data.
Compressors: For prompt compression.
Rerankers: For optimizing search result rankings.
Diffusers: For image generation tasks.
Segmentation Models: For high-quality image segmentation.
Text-to-Speech Models: For converting text to speech.

Safety Features¶

AI Refinery prioritizes the importance of safety in AI applications, offering key features to ensure ethical and secure interactions. Our platform currently provides two safety features:

PII Masking: Safeguards personally identifiable information by masking sensitive data like emails and phone numbers before they reach backend systems or AI agents. This feature is configurable, reversible, and toggleable, aligning with global data protection standards.
Responsible AI (RAI): Applies safety and policy rules to user queries handled by Large Language Models (LLMs). It includes default rules for filtering illegal, harmful, and discriminatory content, while allowing users to create custom rules for specific needs, ensuring ethical AI operations.

These features are crucial for maintaining privacy and promoting responsible AI usage across various applications.

Agentic Advanced Features¶

Agents' Shared Memory¶

Agents' Shared Memory allows multiple AI agents to access and utilize common memory resources, enhancing collaboration for more coherent and contextually aware responses. Key memory types include:

Chat History Module: Stores and retrieves chat conversations efficiently, allowing agents to maintain context across interactions.
Relevant Chat History Module: Fetches and summarizes the most pertinent past conversations, focusing on key insights and themes.
Variable Memory Module: Manages key-value pairs for storing and updating user-specific data to improve personalization and continuity.

Prompt Compression¶

Prompt Compression reduces the size of input prompts while retaining essential information, enabling faster and more cost-effective processing. It focuses on streamlining content from top-ranked documents to enhance efficiency in generating comprehensive responses.

Reranking¶

Reranking improves the relevance of retrieved documents by reordering them based on their pertinence to the query. It prioritizes the most relevant information, ensuring that the agent provides precise and meaningful responses to user queries.

Self-reflection¶

Self-Reflection enables Utility Agents to iteratively refine responses by evaluating and regenerating them until they meet quality standards. It ensures responses are correct and relevant, using strategies like selecting the best attempt or aggregating information for the final output.

APIs¶

The AI Refinery platform offers a comprehensive suite of APIs to enhance AI application development, from processing audio to utilizing machine learning models. Each API focuses on a specific area to meet diverse project needs.

Audio: Tools for audio processing and analysis, including speech recognition.
Chat Completion: Generates responses using LLMs supported by AI Refinery.
Distiller: Enables agentic project creation and access to other AI Refinery features.
Embeddings: Creates the embedding of textual data using embedding models supported by AI Refinery.
Images: Provides image generation and segmentation capabilities.
Knowledge: Offers knowledge extraction and knowledge graph functionalities.
Models: Access the list of models currently supported by AI Refinery.
Physical AI (preview): Provides advanced tools for video-based understanding, simulation, and synthesis of the physical world.