I'm using GPT-4o mini to find duplicates from requests, but the prompt executes 60 times instead of once. Despite correct results, AI builder shows 60 executions for 2 requests. Any help to resolve this would be appreciated.
Initial Execution:
1. Fetch data from the 'Test Requests' Dataverse table where the 'Created On' column is within the last 72 hours.
2. Store the current execution time.
3. **Ensure this step is executed only once per batch of requests.**
Subsequent Executions:
1. Fetch data from the 'Test Requests' Dataverse table where the 'Created On' column is after the last stored execution time.
2. Exclude EAM Work Requests that have already been processed.
3. **Ensure this step is executed only once per batch of requests.**
Data Analysis:
1. Analyze the specified column details updated in the last 72 hours to identify duplicate Test Requests.
2. Preprocess the text data by converting to lowercase, removing punctuation, and standardizing spelling.
3. Tokenize the sentences into individual words.
4. Expand terms using a list of synonyms.
5. Apply lemmatization to reduce words to their base form.
6. Use NLP techniques to compare the 'What needs to be done?' column, considering semantic similarity and cosine similarity to identify potential duplicates.
7. Check the 'Owner' column:
- If the 'Owner' detail is the same for all Test Requests, they are not duplicates.
- If the 'Owner' data is different, they are duplicates.
8. Mark the Test Requests created first as Original and mark the later ones as duplicates.
9. **Ensure this step is executed only once per batch of requests.**
Output:
1. Specifically show the 'Test Requests' unique ID and 'requestor' columns data that are identified as duplicates.
2. Ensure not to show the 'What needs to be done?' column or data in the output.
3. Show accuracy as a percentage.
4. **Ensure this step is executed only once per batch of requests.**
Model Response:
1. Just show Output and Accuracy.
2. Do not include other responses.
3. **Ensure this step is executed only once per batch of requests.**
