You are here:
Create a Intelligent Context Configuration
Creating a Intelligent Context configuration includes these key steps: uploading sample documents for evaluation, configuring the Intelligent Context configuration, testing the results, iterating on the configuration, and publishing to a search index.
Required Editions
| Available in: All Editions supported by Data 360. See Data 360 edition availability. |
| User Permissions Needed | |
|---|---|
| To create a Intelligent Context configuration: | |
Enable Agentforce
Intelligent Context requires generative AI features available through Agentforce. If you haven't already, enable Agentforce in your org.
Create a New Intelligent Context Configuration
To get started with a new Intelligent Context configuration, you need to upload some documents. You can upload up to five documents.
- Navigate to Data Cloud | Process Content | Intelligent Context, and click New Configuration.
- In the New Search Configuration dialog, choose a data space and name for s, and click Save.
- From the Intelligent Context page, select your new project.
- Click Upload Files, or drop files into the content space.
- Select a file to review its content in the Configuration Preview.
Configure with AI Smart Defaults
Generate a configuration using AI smart defaults settings.
- In the agent chat, click Set up my configuration using smart defaults.
-
Review the configuration summary, and choose to accept or modify it.
- If you accept the configuration, continue to Test and Iterate On a Intelligent Context Configuration.
- If you want to modify the configuration, continue to Modify an Intelligent Context Configuration.
Modify an Intelligent Context Configuration
The configuration preview shows the generated chunks created from your source documents. Modify these configuration settings as necessary.
- Click the Edit Configuration tab..
-
To change the search type, click Select Search Index Type, and choose vector
or hybrid search.
Vector Index content based on semantic meaning. Hybrid Index content based on semantic meaning and keywords. -
To change how your content is parsed, click Set Parsing Options.
You cannot select both LLM-based parsing and LLM-based Visual Data Preprocessing options . See LLM-based parsing and preprocessing for more information about these options.
-
Choose one of these options.
Default Parser Extract text with built-in settings. LLM-based Parser Extract all text, images, and other visual elements using an LLM. Docling Parser Extract text from images, tables, and document layouts using Docling Parser. -
To change how your content is preprocessed for chunking multimodal elements, such as
images and tables, click Set Preprocessing Options.
Preprocessing extracts context from images and extracts data from tables while maintaining the context and relationships within the table data.
-
From the Preprocessing dropdown list, choose one of the following:
No preprocessing Default setting. LLM-based preprocessing Use the default prompt as is, or edit it for your specific use case. - To extract text from tables, forms, and image captions, enable the toggle for Image Processing.
- Click Select an Embedding Model, and choose an embedding model from the list.
- To change how your content is chunked, click Set Data Chunking Rules, and choose a chunking strategy.
-
Adjust the token settings as needed.
Chunking Strategy Description Section-aware chunking Uses title and heading elements to chunk documents. Additional settings:
- Max Tokens: Combine adjacent small sections into larger chunks by adjusting the token size.
- Overlap Tokens: Use only if you create multiple chunks from a single section. Specify the number of tokens to add from the end of each chunk to the beginning of the next chunk.
Passage extraction Uses the semantic meaning inherent in HTML tags to chunk a document into passages. Additional settings:
- Max Tokens: Combine adjacent small sections into larger chunks by adjusting the token size.
-
Click Apply Changes.
If you make any configuration changes, Data Cloud generates a new set of content chunks in the Configuration Preview.
- Continue to test and evaluate your configuration.
Test and Iterate On a Intelligent Context Configuration
Test your configuration by asking it questions in the agent chat window or compare the generated chunks with your source document. Each generated chunk contains a confidence score, a source file reference, and a citation.
- In the agent chat, select a suggested question or write your own into the chat box.
- To open the source file in the preview, click a generated chunk. You can also search and filter chunks.
- To compare the generated chunks with the source document, click Compare with Source.
- To make further changes to your configuration, follow the steps in Modify an Intelligent Context Configuration.
- Continue to publish your search index configuration.
Publish the Search Index Configuration
Before you can publish your configuration, you need to set up an unstructured data connection to ingest your source files. Set up an unstructured data connector to ingest your source files into an unstructured data lake object (UDLO) in Data Cloud. See Unstructured Data File Formats and Connectors for a list of connectors.
After you test and modify your configuration, the final step is to publish it to make it available for use in your org.
- From Intelligent Context, click Publish.
- In the Select an Unstructured Data Model Object dialog, select an unstructured data model object (UDMO) to apply the configuration to.
- Click Next.
- In the Create Search Index dialog, confirm or change the index name, and click Publish.
- Navigate to the Content Indexing or Search Index pages to review the results.
