feat(assistant): improve prompt summary (#2435)

This pull request improves the prompt summary in the assistant code. It
updates the map_template and reduce_template to provide clearer
instructions for identifying main themes, key points, and important
information in each section of a document. The reduce_template also
specifies that the final summary should include the main themes, key
points, and important information of the entire document.

<!--
ELLIPSIS_HIDDEN
-->


----

| <a href="https://ellipsis.dev" target="_blank"><img
src="https://avatars.githubusercontent.com/u/80834858?s=400&u=31e596315b0d8f7465b3ee670f25cea677299c96&v=4"
alt="Ellipsis" width="30px" height="30px"/></a> | 🚀 This PR
description was created by [Ellipsis](https://www.ellipsis.dev) for
commit 456736a41d. |
|--------|--------|

### Summary:
This PR enhances the document summarization process in the
`SummaryAssistant` class by providing clearer instructions for
identifying main themes, key points, and important information in each
section of a document.

**Key points**:
- Updated `map_template` and `reduce_template` in `process_assistant`
function of `SummaryAssistant` class.
- Changes provide clearer instructions for summarizing documents.
- Final summary should include main themes, key points, and important
information of the entire document.


----
Generated with ❤️ by [ellipsis.dev](https://www.ellipsis.dev)



<!--
ELLIPSIS_HIDDEN
-->

backend/modules/assistant/ito/summary.py

@@ -87,18 +87,19 @@ class SummaryAssistant(ITO):
 
         llm = ChatLiteLLM(model="gpt-3.5-turbo")
 
-        map_template = """The following is a set of documents
+        map_template = """The following is one document to summarize that has been split into multiple sections:
         {docs}
-        Based on this list of docs, please identify the main themes 
-        Helpful Answer:"""
+        Based on the section, please identify the main themes, key points, and important information in each section.
+        Helpful Knowledge:"""
         map_prompt = PromptTemplate.from_template(map_template)
         map_chain = LLMChain(llm=llm, prompt=map_prompt)
 
         # Reduce
-        reduce_template = """The following is set of summaries:
+        reduce_template = """The following is set of summaries for each section of the document:
         {docs}
-        Take these and distill it into a final, consolidated summary of the main themes. 
-        Helpful Answer:"""
+        Take these and distill it into a final, consolidated summary of the document. Make sure to include the main themes, key points, and important information.
+        Use markdown, headings, bullet points, or any other formatting to make the summary clear and easy to read.
+        Summary:"""
         reduce_prompt = PromptTemplate.from_template(reduce_template)
 
         # Run chain