OpenLink Software: AI Layer

OpenLink AI Layer

Smart Agent Layer that leverages the GPT family of Large Language Models (LLMs) for providing a new generation software application experience that's driven by a natural language-based conversational interface. This solution takes natural language input (text or voice) which is then translated into knowledge base lookups using a variety of declarative query languages such as SQL, SPARQL, SPASQL, and GraphQL.

The OpenLink AI Layer (OPAL) breathes new life into the quest for high-performance, secure data access, integration, and management.

OPAL offers a comprehensive solution for advanced data interaction by combining the strengths of Large Language Models (LLMs), Knowledge Graphs, and RAG, and utilizing declarative query languages such as SPARQL, SQL, and SPASQL (SPARQL inside SQL).

Benefits

Built-in Retrieval Augment Generation

Facilitates Retrieval Augmented Generation using various declarative query languages in distributed or federated modes, ensuring high-performance access to external databases and knowledge graphs.

http://www.openlinksw.com/data/gifs/opal-built-in-rag-virtuoso-support-assistant.gif

Built-in Retrieval Augment Generation

External OpenAPI Service Support

Enhances functionality by integrating any Web Service described via an OpenAPI document. Each Web Service adds a new action to prompt response production pipelines.

http://www.openlinksw.com/data/screenshots/opal-external-function-config2.png

External OpenAPI Service Support

User-Controlled Mapping of Identity to Session Transcripts

Empowers users to control the mapping of their identity to session transcripts, ensuring privacy and security.

http://www.openlinksw.com/data/gifs/opal-user-controlled-session-mapping.gif

User-Controlled Mapping of Identity to Session Transcripts

Supports Fine-Tuning Templates

Provides integration capabilities with various structured data sources, such as databases and knowledge graphs, through fine-tuning templates.

http://www.openlinksw.com/data/screenshots/opal-predefined-query-templates.png

Supports Fine-Tuning Templates

Shareable Sessions, Prompts, and Responses

Enables the sharing of session transcripts, individual prompts, and responses through hyperlinks, facilitating future reuse and recall.

http://www.openlinksw.com/data/gifs/opal-session-promp-response-sharing.gif

Shareable Sessions, Prompts, and Responses

Multi-Protocol Authentication

Supports data access control through loosely-coupled data access policies, compatible with a variety of authentication protocols, ensuring secure access.

http://www.openlinksw.com/data/screenshots/multi-protocol-authentication-opal.png

Multi-Protocol Authentication

Exportable Session Logs

Offers the functionality to export session data to user-designated storage, allowing for easy data management and retrieval.

http://www.openlinksw.com/data/gifs/opal-exportable-session-logs.gif

Exportable Session Logs

Custom Configuration

Enables behavior control through JSON-based configuration documents, incorporating user preferences and fine-tuning templates for optimal performance.

http://www.openlinksw.com/data/screenshots/opal-llm-and-assitant-selection-config.png

Custom Configuration

Loosely Coupled Architecture

Demonstrates a powerful loosely-coupled architecture through external service integration, custom configurations, and fine-tuning templates, enhancing usage and integration options. The conversational user interface is also fully re-skinable.

http://www.openlinksw.com/data/gifs/opal-loosely-coupled-architecture-larger.gif

Loosely Coupled Architecture

Query Execution

Allows direct execution of queries using SQL, SPARQL, SPASQL, and GraphQL, including access to any query service endpoint for SPARQL and GraphQL.

http://www.openlinksw.com/data/screenshots/custom-gpt-sample-sql-sparql-query.png

Query Execution

Use Cases

Data Space driven Retrieval Augmented Generation (RAG)

Constraining hallucinations by Loosely coupling Large Language Models (LLMs) with data managed by database managements, file systems, and knowledge graphs. .

https://www.openlinksw.com/data/gifs/cheapest-virtuoso-offer-purchase-1.gif

Open Conversational Data Connectivity

Enabling the direct and safe execution of queries against, and across, disparate data sources using declarative query languages such as SQL, SPARQL, SPASQL (SPARQL inside SQL), and GraphQL from within Large Language Models e.g., the GPT and Mistral families of language models. For instance, directly query disparate line-of-business data sources as part of a AI-driven systems of intelligence.

https://www.openlinksw.com/data/gifs/data-twingler-sql-example.gif

Enhanced Customer, Partner, and Prospect Support

Enabling the use of natural language for finding information with precision and then acting on it immediately.

https://www.openlinksw.com/data/gifs/uda-oracle-odbc-lite-driver-purchase-1.gif

Enhance Interaction with Core Functionality

Use natural language to execute actions such as integrating virtual or physical Knowledge Graphs generated from disparate data sources e.g., collection of CVS documents or ODBC- or JDBC-accessible data.

https://www.openlinksw.com/data/gifs/opal-fint-virtual-kg-test.gif

Frequently Asked Questions

What is OPAL?

OPAL (OpenLink AI Layer) is a powerful AI-oriented middleware layer that enables the creation, deployment, and use of Smart Agents. It achieves this by loosely coupling Large Language Models (LLMs), OpenAPI-compliant Web Services, and Data Spaces, which include databases, knowledgebases, knowledge graphs, and file systems.

What are the main benefits of OPAL?

<ul><li>Easy development and deployment of cutting-edge Smart Agents or Assistants.</li><li>Mitigation of hallucination effects in AI models through loose coupling with data spaces.</li><li>Enhanced user interaction and experience via natural language interfaces.</li><li>Flexible deployment of Smart Agents across both public and private networks.</li></ul>

What is the OpenLink Personal Assistant (OPAL)?

OPAL is a service hosted on Virtuoso, integrating ChatGPT with additional services for a comprehensive smart agent experience. It supports natural language interactions, offering an enriched Large Language Model (LLM)-based interface for users.

How does OPAL enhance AI model responses?

OPAL includes a built-in Retrieval Augmented Generation (RAG) feature, also known as GraphRAG. This feature enhances LLM response specificity by providing high-performance access to external databases and knowledge graphs using declarative query languages such as SQL, SPARQL, SPASQL, and GraphQL.

How does OPAL handle session logs?

OPAL provides exportable and importable session logs, which aids in session sharing, analysis, and troubleshooting.

Can sessions, prompts, and responses be shared in OPAL?

Yes, OPAL allows the sharing of session transcripts, individual prompts, and responses via hyperlinks, facilitating future reuse and recall.

How do I replay a session in animated playback mode?

Add the <code>&t</code> parameter to a permalink, just before the <code>#</code> fragment identifier. Example: <a href="http://netid-qa.openlinksw.com:8443/chat/?chat_id=s-2ooxbBh4WrhnXRjStCr6pgjYb2EjsUoFM4Ws7wpZZ8zC&t=30#asi-12719">http://netid-qa.openlinksw.com:8443/chat/?chat_id=s-2ooxbBh4WrhnXRjStCr6pgjYb2EjsUoFM4Ws7wpZZ8zC&t=30#asi-12719</a>

How do I know what identities have been used in successful OPAL connections?

Execute the following SPARQL query: <code>SELECT ?s FROM urn:chat:users WHERE { ?s foaf:nick [] }</code>

How do I interact with OPAL using natural language?

OPAL supports natural language interactions through its integration with LLMs like ChatGPT. You can ask questions, give commands, and interact with Smart Agents as if you were conversing with a human assistant.

What are Smart Agents and how do they work in OPAL?

Smart Agents in OPAL are advanced AI assistants that use LLMs to understand and respond to user queries. They can access and process data from various data spaces, providing relevant and accurate information or actions.

Can I personalize the behavior of Smart Agents?

Yes, Smart Agents can be personalized through custom configurations and fine-tuning templates, allowing them to better meet individual user needs and preferences.

How do I troubleshoot issues with Smart Agents?

For troubleshooting, you can export and analyze session logs, consult OPALÃ¢ÂÂs documentation, or reach out to support. Detailed logs help in identifying and resolving issues.

What kind of data sources can OPAL connect to?

OPAL can connect to a wide range of data sources including databases, knowledgebases, knowledge graphs, and file systems. It supports standard protocols like HTTP, ODBC, and JDBC for data access and connectivity.

How can OPAL assist with customer, partner, and prospect support?

OPAL offers new solutions that leverage direct conversational interactions with product support knowledge usually held captive by monolithic applications or websites. This enables direct access to relevant information with precision, the ability to perform actions like opening a support ticket, and sharing sessions via hyperlinks.

How can OPAL assist with customer, partner, and prospect sales?

OPAL provides solutions that leverage direct conversational interactions with product sales-related knowledge, offering direct access to relevant sales information, the ability to perform actions like purchasing offers, and sharing sessions via hyperlinks.