Summarize
v2.0.0
The Summarize LOP sends text or table data to an LLM and returns a condensed summary. It supports four summary styles — brief, detailed, bullet points, and action items — and can accept input from a connected DAT or directly from a text parameter.
Input/Output
Section titled “Input/Output”Inputs
Section titled “Inputs”- Input DAT: Connect a Table DAT or Text DAT. If the table has
roleandmessagecolumns, the operator automatically detects it as a conversation and formats it accordingly. Otherwise the table contents are passed as raw text.
Outputs
Section titled “Outputs”- Summary DAT: A Text DAT containing the generated summary. The summary is also stored in the read-only Summary parameter on the Summary page.
Usage Examples
Section titled “Usage Examples”Summarizing a Conversation Log
Section titled “Summarizing a Conversation Log”- Connect a Table DAT with
roleandmessagecolumns to the Summarize operator’s input. - On the Summary page, set Input Type to “From Input DAT”.
- Set Summary Type to “Brief” for a concise overview, or “Action Items” to extract follow-up tasks.
- Pulse Call Summary to generate the summary.
- The result appears in the Summary parameter and in the output DAT.
Summarizing Arbitrary Text
Section titled “Summarizing Arbitrary Text”- On the Summary page, set Input Type to “From Text Parameter”.
- Enter or paste your content into the Input Text field.
- Choose a Summary Type (Brief, Detailed, Bullet Points, or Action Items).
- Pulse Call Summary.
Using a Custom Prompt
Section titled “Using a Custom Prompt”- Enter additional instructions in the Custom Prompt field — for example, “Focus on budget-related decisions only.”
- The custom prompt is appended to the summary style instructions, giving you control over what the LLM emphasizes.
Structured JSON Output
Section titled “Structured JSON Output”- Enable JSON Mode on the Summary page.
- The LLM is instructed to return the summary as
{"summary": "..."}, which the operator parses automatically. - This is useful when feeding summaries into downstream operators that expect structured data.
Best Practices
Section titled “Best Practices”- Use Brief for quick overviews and Action Items for meeting notes or task extraction.
- Set Temperature to 0 for consistent, deterministic summaries. Increase it for more varied phrasing.
- When summarizing large tables, the operator concatenates all rows into a text block. Keep input size reasonable relative to your model’s context window.
- Max Tokens controls the maximum length of the generated summary. Set it to 0 to let the model decide.
Troubleshooting
Section titled “Troubleshooting”- “No input DAT found”: The operator expects a connected DAT when Input Type is set to “From Input DAT”. Wire a Table DAT or Text DAT to the input.
- “No text provided”: When using “From Text Parameter” mode, the Input Text field must not be empty.
- Truncated summaries: Increase the Max Tokens value to allow longer responses.
Parameters
Section titled “Parameters”Summary
Section titled “Summary” Call Summary (Call)
op('summarize').par.Call Pulse - Default:
False
Active (Active)
op('summarize').par.Active Toggle - Default:
False
Autocall (Autocall)
op('summarize').par.Autocall Toggle - Default:
False
Input Text (Inputtext)
op('summarize').par.Inputtext Str - Default:
"" (Empty String)
Custom Prompt (Customprompt)
op('summarize').par.Customprompt Str - Default:
"" (Empty String)
Max Tokens (Maxtokens)
op('summarize').par.Maxtokens Int - Default:
0- Range:
- 0 to 1
- Slider Range:
- 0 to 1
Temperature (Temperature)
op('summarize').par.Temperature Float - Default:
0.0- Range:
- 0 to 1
- Slider Range:
- 0 to 1
JSON Mode (Jsonmode)
op('summarize').par.Jsonmode Toggle - Default:
False
Summary (Summary)
op('summarize').par.Summary Str - Default:
"" (Empty String)
Model Controller (Modelcontroller)
op('summarize').par.Modelcontroller OP - Default:
"" (Empty String)
Search Models (Search)
op('summarize').par.Search Toggle - Default:
False
Model Search (Modelsearch)
op('summarize').par.Modelsearch Str - Default:
"" (Empty String)
Provider Model Documentation
Consult the documentation for your chosen provider to find supported models, API key information, and usage limits.
View LiteLLM Supported Providers →
Callbacks
Section titled “Callbacks” Available Callbacks:
onTaskStartonTaskComplete
Example Callback Structure:
def onTaskStart(info):
# Called when the summarization API call begins
pass
def onTaskComplete(info):
# Called when summarization finishes
# The summary text is available in the Summary parameter and summary_dat output
passChangelog
Section titled “Changelog”v2.0.02025-07-30
completely rewritten and sleeker
v1.0.02024-11-06
Initial release