Filter BOT/AI Pull Request Reviews

[samad-yar-khan]

Why do we need this ?

• We currently include Bot/AI reviewers while considering first response time. This decreases the actual First Response time of the PR due to instant bots reveiws.

Acceptance Criteria

• Identify Reviews by AI/Bots
• While calculating Pull Request Cycle Time metrics (first response time and overall cycle time) exclude Bot/AI reviews.

Further Comments / References

[thesujai]

Hi @samad-yar-khan Hope you are well:)
For identifying bots, a new field called actor_type (which indicates either "bot" or "user") should be added to the PullRequestEvent model.
How will we know whether the review is from a bot or not?
For GitHub, we can leverage the github.PullRequest.get_reviews() API, where each review's user object also includes a type indicating whether it is a bot or a user. So, just after this line, we will retrieve the actor type as:

actor_type=actor.get("type", "")

Then, we add it to the PullRequestEvent.

I'm not sure how GitLab processes this, but the second part will be the same for both platforms:

The second part is straightforward: simply filter out bot events before any operation in this method. At the beginning of the function, the code will look like:

pr_events = [
            e
            for e in pr_events
            if not e.actor_type.lower() == "bot" 
        ]

[Ayyanaruto]

Can I work on this issue

[samad-yar-khan]

@Ayyanaruto sure, give it a go. Lets discuss how you are planning to solve this ?

[Ayyanaruto]

Hi @samad-yar-khan sorry for late reply as per my deduction the cycle time is calculated here backend/analytics_server/mhq/service/code/sync/etl_code_analytics.py In the Class etl_code_analytics I will check in the api response includes bot in username or if type is bot or author_association=BOT and filter out if present

[Ayyanaruto]

Hi @samad-yar-khan,
I’ve created the PR for this issue. Sorry about the earlier PR — I had pushed the branch with the wrong logic and wasn’t aware of it. All the tests were failing, and I got a bit confused.
I’ve closed that PR and created a fresh one. When you’re free, please take a look!