feat(server): faster reranking based on confidence (#12957)

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

## Summary by CodeRabbit

* **New Features**
* Improved document reranking with a more streamlined and accurate
scoring system.
* Enhanced support for binary ("yes"/"no") document relevance judgments.

* **Improvements**
* Simplified user prompts and output formats for reranking tasks, making
results easier to interpret.
  * Increased reliability and consistency in document ranking results.

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

packages/backend/server/src/plugins/copilot/prompt/prompts.ts

@@ -342,57 +342,11 @@ Convert a multi-speaker audio recording into a structured JSON format by transcr
     messages: [
       {
         role: 'system',
-        content: `Evaluate and rank search results based on their relevance and quality to the given query by assigning a score from 1 to 10, where 10 denotes the highest relevance.
-
-Consider various factors such as content alignment with the query, source credibility, timeliness, and user intent.
-
-# Steps
-
-1. **Read the Query**: Understand the main intent and specific details of the search query.
-2. **Review Each Result**:
-   - Analyze the content's relevance to the query.
-   - Assess the credibility of the source or website.
-   - Consider the timeliness of the information, ensuring it's current and relevant.
-   - Evaluate the alignment with potential user intent based on the query.
-3. **Scoring**:
-   - Assign a score from 1 to 10 based on the overall relevance and quality, with 10 being the most relevant.
-   - Each chunk returns a score and should not be mixed together.
-
-# Output Format
-
-Return a JSON object for each result in the following format in raw:
-{
-  "scores": [
-    {
-      "reason": "[Reasoning behind the score in 20 words]",
-      "chunk": "[chunk]",
-      "targetId": "[targetId]",
-      "score": [1-10]
-    }
-  ]
-}
-
-# Notes
-
-- Be aware of the potential biases or inaccuracies in the sources.
-- Consider if the content is comprehensive and directly answers the query.
-- Pay attention to the nuances of user intent that might influence relevance.`,
+        content: `Judge whether the Document meets the requirements based on the Query and the Instruct provided. The answer must be "yes" or "no".`,
       },
       {
         role: 'user',
-        content: `
-<query>{{query}}</query>
-<results>
-{{#results}}
-<result>
-<targetId>{{targetId}}</targetId>
-<chunk>{{chunk}}</chunk>
-<content>
-{{content}}
-</content>
-</result>
-{{/results}}
-</results>`,
+        content: `<Instruct>: Given a web search query, retrieve relevant passages that answer the query\n<Query>: {query}\n<Document>: {doc}`,
       },
     ],
   },