Oracle AI Optimizer and Toolkit
Simplify and accelerate development of AI-powered applications using retrieval-augmented generation and agentic use cases through rapid experimentation and evaluation.
Why Oracle AI Optimizer and Toolkit?
Organizations are moving beyond generative AI experimentation to agentic AI applications that can reason over enterprise data, retrieve trusted information, use tools, and take action. The challenge is finding the right combination of models, data, tools, and configurations to deliver business value.
The Oracle AI Optimizer and Toolkit makes it easier to experiment with and evaluate AI applications grounded in enterprise data. Its no-code environment lets you connect your data to the language and embedding models of your choice, with control over prompts, model parameters, chunking strategies, and semantic search. Explore RAG and agentic workflows, connect agents to enterprise data and tools through Model Context Protocol (MCP), then use built-in evaluation to compare approaches and find the configuration that works best for your use case.
Once you have validated an approach, move from experimentation to production without starting over. Deploy it as a scalable service using the included API server, or generate production-ready microservices in Java with Spring AI or Python with LangChain. Oracle AI Database provides the scalable data foundation for agentic AI, while Oracle AI Optimizer and Toolkit provides the environment to rapidly experiment, evaluate, and turn ideas into working applications.
Free and Open Source
Free to use
Licensed under the Universal Permissive License v1.0, a permissive, OSI- and FSF-approved, GPL-compatible license.
Explore, learn, and contribute
Explore the code to see how each feature is implemented, learn from it for your own projects. Contributions are welcome.
Easy Configuration
Use the models that fit your use case
Choose the language and embedding models that work best for your data and goals, whether provided as a service or self-hosted.
Tune model behavior
Adjust model parameters to control how models respond, reduce hallucinations, and meet the needs of your use case.
Customize model instructions
Use predefined instructions or create your own to guide how the model behaves and responds.
Extensible Knowledge
Ground AI in your enterprise data
Use retrevial augmented generation and natural language to SQL to give models access to relevant, trusted information from your Oracle AI Database.
Work with structured and unstructured data
Use structured data alongside unstructured content to provide richer context for AI applications and agents.
Use Oracle AI Database for vector search
Store embeddings in Oracle AI Database and use scalable vector search to find relevant information based on meaning.
Automatic Testing
Generate test data automatically
Generate question-and-answer test sets to evaluate your chatbot and automatically test each iteration.
Evaluate responses automatically
Use a second model to evaluate your application’s responses and assess their quality.
Model Context Protocol (MCP) integration
Extend with MCP
Use the built-in MCP server to connect AI Optimizer and Toolkit with external tools and services.
Connect to Oracle AI Database through SQLcl MCP
Use Oracle SQLcl MCP as a proxy to connect agents to Oracle AI Database, including natural language to SQL capabilities.
Benefits of Oracle AI Optimizer and Toolkit
Explore without writing code
Quickly explore AI applications grounded in your enterprise data without writing code. Work with structured and unstructured data, experiment with RAG and natural language to SQL, and see how different approaches perform directly in the user interface.
Use the models that fit your needs
Choose the language and embedding models that best fit your use case. Connect to hosted models or use models within your own environment when you need greater control over where your data is processed.
Generate and run automatic tests
Generate test data from your own knowledge sources instead of building test sets manually. Refine and reuse those tests to compare configurations, evaluate responses, and see whether changes improve results.
Move from experimentation to production
Turn a tested configuration into a production application without rebuilding it from scratch. Deploy it using the OpenAI-compatible API server, or generate a Spring AI microservice to run on Kubernetes platforms such as Oracle Backend for Microservices and AI.
Integrated Oracle Technologies
- Agentic Systems
- Natural Language to SQL
- Context Management