Creating an Orchestration Job

A Knowledge Orchestration job processes your transcript or query data through the full Knowledge Orchestration pipeline — ingestion, embedding, clustering, topic hierarchy generation, and instant answer generation.

To create an orchestration job:

In the Navigation Menu, browse to Knowledge Orchestration.
Click + Create Job in the upper right of the dashboard to get started.
The Create Job dialog offers four input methods depending on where your source data resides. Select the tab that matches your data source:
- Upload Files: Drag and drop your file into the upload area, or select Select File to browse for it.
  - Supported formats are JSON files or a ZIP archive of JSON files, with a maximum of 10,000 JSON files at 40 KB per file.
  - From the Source format dropdown, select the vendor format of your source data. The available options are the same across the Upload Files and S3/Server Path tabs — see Source format options below.
- S3/Server Path: Enter the full S3 URI or directory path to your JSON files in the S3 URI or path field — for example, s3://your-bucket/path/to/transcripts/.
  - Verify that the service account has read access to the specified location before submitting.
  - Use the Source format dropdown to select the vendor format of your source data (see Source format options below).
  Select Sample Transcript Format at the bottom of the dialog to review the expected JSON structure before uploading files. The reference panel provides two views:
  - Schema: Shows the full JSON schema for the transcript file, including the required fields (filename, channel_type, language, text), their expected data types, and descriptions for each property.
  - Sample JSON: Shows a populated example transcript, with values for filename, ref_id, channel_type, language, intent, created_date, and a text array containing the individual caller and agent turns of the conversation.
  Ensure your files conform to this structure before submitting. Malformed or incomplete files may cause the job to fail or produce unreliable clustering results.
- Keywords & Queries: Paste your list of keywords or search queries into the Keywords and search queries field. Use this method to audit portal coverage against a known set of terms without requiring transcript data.
  
  Select Sample format at the bottom of the dialog to open the Supported input formats reference. You can paste your keywords or queries in any one of the following formats:
  - Plain text: One value per line, for example:
```
How do I reset my password?
What are your business hours?
Return policy
```
  - Single-column CSV: A header row is optional; for example, export the first column from a spreadsheet keyword list.
  - JSON array of strings: The pasted text must start with [, for example:
```
[
  "How do I reset my password?",
  "What are your business hours?",
  "Return policy"
]
```
  Limits: up to 1,000 items, with a maximum of 1,000 characters per item.
- Analytics: Use this tab to generate a job directly from portal search analytics instead of uploading transcripts or queries.
  - Select a Date range — Last 1 month, Last 3 months, Last 6 months, or Custom — to define the analytics period the job should pull from.
  - If you select Custom, enter a Start date and End date.
  - The job pulls the portal's search terms along with their successful and failed search counts from analytics, normalizes them using an LLM, and processes them through the Knowledge Orchestration pipeline to identify knowledge gaps based on the selected portal's scope. Use these gaps as the basis for creating new articles.
Source format options (Upload Files and S3/Server Path tabs):
- Native (Transcript JSON): Use this option for eGain format transcripts.
- Amazon Connect (Contact Lens)
- Genesys Cloud
- Salesforce (LiveChatTranscript)
- Twilio (Conversation Intelligence)
- Plain text / WebVTT / SRT / CSV / Infomine
Provide the below information for the selected data source:
- Which Portal Would You Like to Generate Instant Answers For?: Select the portal that the tool should query when auditing answers. Only portals with Use Generative Instant Answer enabled appear in this list.
- Language: Select the language of the source content. The dropdown displays all KB primary languages configured for the partition.
- Job Name: Enter a descriptive name for the job.
- Job Description (optional): Add context to help identify the job later.
Once these are filled in, click Start Job.

Creating an Orchestration Job

To create an orchestration job:

Related Topics