All Connector Updates

Merlin Guardian Generally Available

Now Generally available!

You can use Merlin Guardian to

1. Tokenize sensitive PII data

2. Identify the presence of PII data

3. Unmask tokenized payloads generated from AI inference for use in downstream automations

4. and more

Read more in the docs

Merlin Intelligent Document Processing (IDP) Generally Available

Now Generally available!

You can use Merlin IDP to

1. Convert files to markdown and make unstructured data available to AI services and Agents

2. Run queries against invoices, receipts, contracts, and more to get specific data points with confidence intervals

3. and more

Read more in the docs

Merlin IDP (Beta)

We are migrating all functionality from the Merlin Extract connector to this new connector and deploying two new operations: Convert to Markdown and Convert to HTML

Use Composite AI to extract important data from PDFs and Images.

Merlin IDP brings native AI Capabilities to the Tray platform. Use this tooling to extract important information from contracts, invoices, receipts and more.

Each Extraction job returns the data you queried and a confidence interval for the returned information.

All of our native AI capabilities are private and secure because they run directly in our cloud infrastructure within the region of your Tray instance.

Read more in the docs

Merlin Text Generation (Beta)

We are making our Merlin Native AI capabilities easier to find and use within the app. Text generation and summarization operations are the first two capabilities in this connector.

Read more in the docs

Merlin Text Analysis (Beta)

We are making our Merlin Native AI capabilities easier to find and use within the app. Sentiment analysis and classification operations are the first two capabilities in this connector.

Read more in the docs

AWS Bedrock Connector (GenerativeAI)

Version: 1.0

Release Date: June 14th, 2024

AWS Bedrock offers a choice of high-performing foundation models (FMs) from leading AI companies like AI21 Labs, Anthropic, Cohere, Meta, Mistral AI, Stability AI, and AWS through a single API, along with a broad set of capabilities you need to build generative AI applications with security, privacy, and responsible AI.

In this version we focused on delivering the best functionality for infusing AI into your business process. We used a specific set of endpoints to make prompt engineering and model interoperability as simple as possible so you can focus on performance, not prompt syntax.

• AI Generation (Run Inference):

  • Prompt the exact same way across all your models. If a particular model doesn't support a parameter you get an error informing you so you don't have to spend time reading the docs for each model.

  • All models available on Bedrock are immediately available. When Bedrock adds more models, those are AUTOMATICALLY added to our connector. You still need to approve models for use within the AWS console in order to use it in Tray.

  • You can leverage AWS Bedrock Guardrails to configure privacy and security features in their console and then apply them to all the Generative AI work you do right within Tray.

• Create Embeddings:

  • Two model families to choose from. Cohere has batch embedding options and Titan is a reliable model at a bargain.

We can't wait to see what you build on Tray with AWS Bedrock.

OpenAI - Connector v3.0 with Improved User Experience and New Operations

Version: 3.0

Release Date: May 16th, 2024

The new GPT-4o model is here. We added a few new operations and made the connector much easier to work with thanks to your feedback!

What's New?

Copy

• Create Chat Completion:

  • Now with GTP-4o, OpenAI's frontier model which is 2x faster and 50% the cost of GPT-4-turbo.

  • We simplified the interface of the connector so that you can get started faster.

  • JSON mode - perfect for AI infusion in your automations - is easy to use and has improved error messaging if you forget to include "JSON" in your prompt (required by the OpenAI API)

• Create Embeddings:

  • Use the newest embedding models and experience our simplified embeddings operation that helps you get started faster.

• Create Speech:

  • Generate an audio file from text. Taking content generation to new heights. Everything you write can now become audio content.

• HTTP Operation:

  • Access any part of the OpenAI API through the Raw operation.

For detailed implementation guides and more information, refer to Tray Documentation - OpenAI.

New Connector - Cohere

We're excited to introduce our latest connector addition - Cohere.

Customers can use this connector to leverage Cohere's advanced large language models (LLMs) and Retrieval-Augmented Generation (RAG) capabilities, designed for enterprise applications.

Key features:

Copy

The Cohere connector simplifies the creation and management of text and document embeddings, enabling powerful similarity searches and conversational AI interactions. Supported operations include:

• Generating Text Embeddings:

  Create vector representations of text for nuanced similarity searches.

• Batch Document Embeddings:

  Process and generate embeddings for large batches of documents, optimising for efficiency in data-heavy scenarios.

• Query Embeddings Generation:

  Turn incoming queries into embeddings, enhancing search accuracy within vector databases.

• Chat Generation:

  Leverage similarity searches to produce model responses for chat interactions, facilitating more engaging and context-aware conversational experiences.

For detailed implementation guides and more information, please visit the Tray Documentation.