Lyzr
The simplest agent framework to build ‘Secure, Safe and Responsible’ GenAI apps faster
Introducing Lyzr
Lyzr is an agent framework with an agentic approach to building generative AI applications. Its highly configurable agents come with pre-built RAG pipelines, safe and responsible features, allowing you to build and launch enterprise ready agents in minutes. Agents can be built and tested on Lyzr’s Studio or can be directly used with the help of the REST APIs that are available.
With Lyzr, you can build chatbots, knowledge search, data analysis, RAG-powered apps, and multi-agent workflow automation with minimal effort.
Lyzr’s Safe and Responsible AI Architecture
Lyzr’s framework incorporates Lyzr’s proprietory Safe and Responsible AI features within the agent inferencing flow. This allows every agent built on the platform to take advantage of these protections and deliver results with high accuracy.
Who is Lyzr for?
For Developers - you will love the simplicity of the framework and appreciate the ability to build generative AI apps rapidly.
For CTOs, CPOs - integrate generative AI features into your apps seamlessly with configurable agents with a vast array of features, all with your in-house tech team. The required learning curve to build on Lyzr is literally just a few minutes.
For CIOs - introduce generative AI to your enterprise with Lyzr’s proprietory Safe and Responsible AI architecture, ensuring security, privacy and accuracy. And Lyzr’s AI Management System (AIMS) makes it easy to manage agents, monitor events logs, build using AI studios, and even help your team learn generative AI with the in-built Lyzr academy.
What can you build on Lyzr’s agent framework?
- Lyzr Agents: Deploy in minutes
Create agents in Lyzr Studio or directly through the APIs. The agents’ rich feature set and tool capabilities allows you to create your own tailor made agent.
| Use Case | Description |
|---|---|
| Chat Agent | State-of-the-art chatbots with short-term and long-term memory, RLHF (Reinforced Learning Human Feedback), a customizable system prompt, and seamless switches between LLMs. Try Lyzr Chat Agent |
| Knowledge Search | Bring perplexity-style document and data search to your organization. Lyzr’s document search can handle 100,000 documents and more, retrieving information accurately every time. And you get a citation to the source along with fetching the source file. Try Lyzr Knowledge Search |
| RAG Powered Apps | It is probably the most comprehensive RAG agent ever built. Just by tuning the parameters, you can choose up to 630 different RAG pipelines. You can also integrate with any LLM, vector store, embedding model, reranker, or parser of your choice. Read Blog |
| QA Bot | A simplified version of Lyzr Chatbot—an accurate question-answering agent to power your help and FAQs section. QA Bots are also the best fit for handling one-way conversations like employee handbook searches, law lookup searches, etc. |
| Text-to-SQL | Want to run an exploratory analysis on your database? Run Lyzr’s Text-to-SQL agent to convert your text queries to SQL code that can handle complex joins, extract information, and build insights and recommendations. |
- Lyzr Multi-Agent Orchestration
Lyzr provides two modes of orchestration
- Orchestration with Directed Acyclic Graphs - In cases where the workflow is well defined with multiple dependent tasks, the entire flow can be orchestrated using Lyzr’s DAG Agent Orchestration. An example would be Lyzr’s AI SDR. A task can be dependent on multiple other tasks or multiple tasks can be dependent on a single tasks or anything in between. Lyzr’s orchestration will ensure all dependencies are resolved and executed in perfect order. All this while ensuring complete traceability.
- Managerial orchestration - This style of orchestration is more suited for dynamic workflows. Depending on the end goal, the manager agent breaks down the task and invokes the agents available to it to perform a subtask. At every step, the manager observes and determines the next course of action. This could be retrying in case of faults or invoking another agent.
How is Lyzr different from competitors?
Langchain - building blocks way of building generative AI apps
DSPy - programmatic way of building generative AI apps
LlamaIndex - modular way of building generative AI apps with organizations data
Lyzr - agentic way of building generative AI apps