update: live event prompt

app.py

@@ -343,7 +343,7 @@ def get_initial_analysis(
 Your task is to analyze this street interview transcript and identify the final/best timestamps for each topic or question discussed. Here are the key rules:
 
 1. For any topic/answer that appears multiple times in the transcript (even partially):
-   - The LAST occurrence is always considered the best version
+   - The LAST occurrence is always considered the best version. If the same thing is said multiple times, the last time is the best, all previous times are considered as additional takes.
    - This includes cases where parts of an answer are scattered throughout the transcript
    - Even slight variations of the same answer should be tracked
    - List timestamps for ALL takes, with the final take highlighted as the best answer
@@ -359,6 +359,8 @@ Your task is to analyze this street interview transcript and identify the final/
    - If a question or introduction is revisited later at any point, please use the later timestamp
    - Track partial answers to the same question across the transcript
 
+You need to make sure that any words that are repeated, you need to pick the last of them.
+
 Return format:
 
 [Question Title]
@@ -385,17 +387,19 @@ Total takes: 2
 Format the output as follows:
 
 **Speaker Name**
-1. [Topic title <div id='topic' style="display: inline"> 15s at 12:30 </div>]({link_start}://{{origin}}/collab/{{cid}}/{{rsid}}?st={{750}}&et={{765}})
-2. [Topic title <div id='topic' style="display: inline"> 30s at 14:45 </div>]({link_start}://{{origin}}/collab/{{cid}}/{{rsid}}?st={{885}}&et={{915}})
+1. [Topic title <div id='topic' style="display: inline"> 12s at 12:30 </div>]({link_start}://{{origin}}/collab/{{cid}}/{{rsid}}?st={{750}}&et={{762}})
+2. [Topic title <div id='topic' style="display: inline"> 33s at 14:45 </div>]({link_start}://{{origin}}/collab/{{cid}}/{{rsid}}?st={{885}}&et={{918}})
 3. [Topic title <div id='topic' style="display: inline"> 1m 10s at 16:20 </div>]({link_start}://{{origin}}/collab/{{cid}}/{{rsid}}?st={{980}}&et={{1050}})
+4. [Topic title <div id='topic' style="display: inline"> 25s at 14:00 </div>]({link_start}://{{origin}}/collab/{{cid}}/{{rsid}}?st={{840}}&et={{865}})
 
 Requirements:
 - Only speaker names should be bold using markdown **Name**
 - Topic titles should be in normal case (i.e first word of Sentence capital rest small)
 - Use timestamp and duration format: HH:MM (Xs) where X is seconds
 - Each line should be a clickable link including timestamp and title
+- When selecting timestamps, strictly limit the start and end times to contain only the current speaker's continuous dialogue. The timestamp should begin exactly when the current speaker starts talking about that specific topic and end when they finish their point or another speaker begins. Do not include any portions of previous or subsequent speakers' dialogue in the selected time range.
 - Convert display timestamps to seconds for URL parameters
-    Example: 15s at 10:13 in URL would be st=613&et=628
+    Example: 21s at 10:13 in URL would be st=613&et=634
 
 URL format: {link_start}://{{origin}}/collab/{{cid}}/{{rsid}}?st={{start_time_in_sec}}&et={{end_time_in_sec}}