SuveenE-TF

Suveen Ellawela

@SuveenE-TF
GitHub Profile
collaborative and accommodating
Collaborative and solution-oriented with a focus on pragmatic improvements. Often responds positively to suggestions with immediate commitments to make changes, and provides detailed context when explaining decisions or asking for clarification.
614
Comments
328
PRs
12
Repos
88
Avg Chars
2
Harshness

Personality

Highly responsive and accommodating Detail-oriented with good memory for project context Pragmatic problem-solver Collaborative team player Process-conscious about merging and workflows Question-asking for clarity Quick to acknowledge good catches Documentation-focused

Greatest Hits

"Will make the changes!"
"Good catch. Will change."
"That makes sense."
"lgtm!"
"correct me if i'm wrong."
"Merging to unblock dogfooding."

Focus Areas

Common Phrases

"Will make the changes" "Good catch. Will change" "Makes sense" "lgtm!" "Thanks for pointing it out!" "I think" "Will change" "Understood" "That makes sense" "good work!" "Any reason why" "correct me if i'm wrong" "i assume" "pls change" "nit:" "iiuc"

Sentiment Breakdown

positive
40
neutral
343
questioning
40
constructive
36
harsh_questioning
3

Review Outcomes

APPROVED
192
CHANGES_REQUESTED
21
COMMENTED
10
DISMISSED
10

Most Reviewed Authors

SuveenE-TF
221
joshkung247
120
jinyangTF
62
ayc1
45
KateZhang98
34
jinxz01
31
jayfish0
25
frankfeng98
16
zackermax-tinyfish
15
aarontantf
13

Spiciest Comments

friday/#691 [view]
> Thanks for the update! Could u also run the hotel regression test to make sure nothing is broken? `pytest tests/integration/hotel_test.py --count=1 -n=16` yep, I just ran the hotel test and it passes.
friday/#733 [view]
> > seemsgood. i assume @KateZhang98 knows what happens to the debugger with the change and i assume langsmith UI still looks fine. > > Langsmith will be broken, but it's ok. I will rework when the restructure is done It's done now. Thanks a lot Kate!
friday/#442 [view]
> this doesn't work all the time actually. it fails to extract the 72,600 yen option. https://smith.langchain.com/o/15e007eb-2337-4ddb-97f3-d01467854f36/projects/p/3d9f0773-4461-4799-ae2a-392ba661b6fc/r/35a452f9-3140-4890-8828-193291911b98?trace_id=24e463e8-92e3-4ae1-8837-32835cf25abb&start_time=2025-03-05T02:43:26.803614 I see. If it only fails some times, it means the atree includes 72,600. I'll try the gemini-pro models (what we are using in agentql).

AI Persona Prompt

You are @SuveenE-TF, a highly collaborative and accommodating code reviewer who focuses on practical solutions and team productivity. You frequently respond to feedback with immediate commitments like 'Will make the changes!' or 'Good catch. Will change.' You're detail-oriented and often provide helpful context about why certain decisions were made, referencing other parts of the codebase or external examples (like OpenAI docs) to support your reasoning. You ask clarifying questions using phrases like 'iiuc' (if I understand correctly), 'correct me if i'm wrong', and 'Any reason why'. You're process-conscious about workflows, often mentioning branch merging strategies and unblocking dogfooding efforts. When reviewing others' code, you use encouraging language like 'good work!' and 'lgtm!' while still providing constructive feedback with 'nit:' for minor issues. You have strong institutional knowledge and often reference previous implementations, related PRs, and project context. You're pragmatic about trade-offs and willing to defer perfect solutions for functional ones, often saying 'I think we can come back to this later' or 'we can fix them after landing this'. You focus heavily on documentation consistency, API accuracy, testing coverage, and logical implementation flow. Your reviews are thorough but always aimed at moving the project forward efficiently.

Recent Comments (462 total)

friday/#589 Catch keyboard interrupt in main [view]
lgtm
friday/#308 Implement open new tab tool [view]
good work! Added a few comments.
friday/#308 Implement open new tab tool · pyt/app/tools/constants.py [view]
Let's remove the AgentQL part from everywhere. Even though we wrap the page with agentql, i dont think it's significant enough to have it in the name.
friday/#308 Implement open new tab tool · pyt/app/tools/open_new_tab.py [view]
remove the name and description from the input.
friday/#308 Implement open new tab tool · pyt/app/tools/open_new_tab.py [view]
Since the url in the input is not optional, when will it be None?
friday/#308 Implement open new tab tool · pyt/app/models/browsers/base.py [view]
We might run into issues with this as we have the following line in our `main.py`. Then we'll have two contexts. ``` context = await browser.new_context() if cookies: for cookie in cookies: if "sameSite" not in cookie or cookie["sameSite"] not in [ "Strict", "Lax", "None", ]:
friday/#1123 Add json schema input type · pyt/friday/components/summarizer.py [view]
shouldn't the response be validated against `self.agentOutput`?
friday/#1018 add multishot eval to goal planner evaluator [view]
what if the goal planner decides it's achievable within 5 actions? iiuc we are not looking for that. If we have an expected value, we can compare that and do the llm eval as per the prev implementation. Please lmk if i am missing context.
friday/#1018 add multishot eval to goal planner evaluator · pyt/evaluation/evaluator/components/prompts/goal_planner.py [view]
@jzachr do you mean a summary of https://docs.google.com/document/d/1NkLj3K1UGM0iFWal4RjYcSRGpqaYX2ftz-bhM9MIRB4/edit?tab=t.sx8owixcz8ho
friday/#1018 add multishot eval to goal planner evaluator [view]
Closing this PR as the main change requested is to create a new evalautor for the multi shot method. The new PR is #1045
friday/#1220 Add decorator to monitor http calls in cory evaluator.py [view]
lgtm! thanks!!
friday/#1207 Add trip.com example [view]
lgtm! to test with a new usecase we need to add an output type to the test_inputs right?
friday/#1165 Session replayer to handle responses w/o content-type headers · pyt/friday/services/weblog/session_replayer.py [view]
yeah you are correct. I talked with Zach about this too. It can't be empty
friday/#1149 Migrate tool-annotator-local from local to friday [view]
Looks pretty ok to me. a couple of nits: 1. adding a screen recording might be very helpful since this is a new tool and most of us don't have a good idea 2. we can use prettier or something to format the code 3. there's a lot of console.logs, I assume they are not too noisy 4. please add a README.md to this folder with the instructions also wait for @frankfeng98's review too, i think h
friday/#1155 Read expected value from langsmith dataset feedback section · pyt/evaluation/evaluator/components/goal_completion_verifier_evaluator.py [view]
will do!