@nutritious-balloon-65098 in the OpenHands/OpenHands repo, when was star feedback removed, and was there a reason stated for why it was removed?

Here’s the final summary of my work: - I investigated the OpenHands/OpenHands repo history and current frontend code to determine when star feedback disappeared and whether a reason was stated. - Answer: star feedback was not fully removed from the codebase. It was disabled/hidden for V1 conversations on 2025-10-23 in commit

dd2a62c992

: -

refactor(frontend): disable some agent server API until implemented in the server source code (#11476)

- Stated reason: yes — the code/comments indicate the reason was that V1 feedback support was not implemented yet, specifically: -

TODO: Hide LikertScale for V1 conversations

-

This is a temporary measure and may be re-enabled in the future

-

TODO: Implement V1 feedback functionality when API supports V1 event IDs

-

For now, we'll skip feedback functionality for V1 events

- I also found a later related change on 2025-11-07, commit

bfe60d3bbf

, which disabled the V1 batch feedback endpoint: -

chore(frontend): Disable /feedback/conversation/{conversationId}/batch for V1 conversations (#11668)

- For context, I confirmed there was an earlier removal of the old thumbs up/down feedback UI on 2025-06-09 in commit

aaf171840696

, but that was separate from the later star/Likert feedback. - I also confirmed the star/Likert feedback itself was introduced on 2025-06-16 in commit

2c4ecd02f7af

: -

feat(frontend): add user feedback Likert scale for agent performance rating (only on OH Cloud) (#8992)

So the concise answer is: - When was star feedback removed? In practice, it was disabled for V1 conversations on 2025-10-23. - Was there a stated reason? Yes — the stated reason was that the V1/agent-server feedback APIs and V1 event ID support were not implemented yet.

Summary of my work: - I investigated the OpenHands/OpenHands repo history to determine when star feedback stopped appearing and why. - I found that **star feedback was not fully removed from the codebase**; it was disabled for V1 conversations. - The main change was made by Hiep Le in commit: -

dd2a62c992bf23fe9abdad4dd59b9ae03c67de0a

- Date: 2025-10-23 - Message:

refactor(frontend): disable some agent server API until implemented in the server source code (#11476)

- The stated reason in the code/comments was that V1 feedback support was not implemented yet, specifically around V1 event IDs / agent-server feedback APIs. - I also found a follow-up related change by sp.wack: -

bfe60d3bbfda98ff405f83c32585ac39c09a8072

- Date: 2025-11-07 - Message:

chore(frontend): Disable /feedback/conversation/{conversationId}/batch for V1 conversations (#11668)

So the concise answer is: - Who made the change? Primarily Hiep Le. - Why was it removed/hidden? Because the V1/agent-server feedback functionality was not yet implemented.

Summary of my work: - I surveyed both repos: -

OpenHands/OpenHands

-

OpenHands/software-agent-sdk

- Conclusion: - Re-enabling star feedback for V1 conversations does not appear to require SDK changes. - The SDK already supports the needed shape: - V1 event IDs are strings - agent-server event APIs already use string event IDs - The actual work needed is in OpenHands/OpenHands: 1. Frontend - remove the explicit V1 hide/disable logic for Likert/star feedback - generalize feedback hooks/types from numeric event IDs to string-capable IDs - re-enable feedback lookup/fetching for V1 - wire the star UI back into the V1 event components 2. Backend / SaaS feedback path - update feedback storage from

event_id: Integer

to string-based IDs - make feedback routes accept V1 event IDs - replace the current V0 EventStore-based event enumeration with V1 app_server event service logic for V1 conversations 3. Tests - add/update frontend and enterprise tests for V1 feedback submission/rendering/batch lookup - Additional issue I found: - feedback

metadata

is accepted/passed around but does not appear to be fully persisted in the current enterprise storage model, so that should be cleaned up while touching this area. - No code changes were made; this was an investigation only.

Summary of my work: - I surveyed both

OpenHands/OpenHands

and

OpenHands/software-agent-sdk

to determine what would be needed to re-enable star feedback on V1 conversations. - Conclusion: the needed work is almost entirely in

OpenHands/OpenHands

, not the SDK. ### Main answer We need string-based event IDs for V1 feedback because V1 already uses string event IDs throughout the stack: - In the SDK,

EventID = str

- V1 event objects use string IDs (UUID-like strings) - The OpenHands V1 frontend types also define event IDs as strings So if feedback is attached to a V1 event, the feedback system must store and query by a string event ID. ### Why not keep integers? Because integer event IDs only match the legacy V0 event model, where events are indexed like

0, 1, 2...

. V1 events are not addressed that way. ### Why

string

instead of

UUID

? Because

string

is the canonical, least-assumptive type across the V1 system: - it matches the SDK/API/frontend contract - it avoids forcing UUID parsing/normalization everywhere - it cleanly supports both: - V0 IDs like

"42"

- V1 IDs like

"550e8400-e29b-41d4-a716-446655440000"

### Practical implication To re-enable V1 star feedback,

OpenHands/OpenHands

would need to: - remove frontend V1 feedback guards - generalize feedback hooks/components from numeric event IDs to string IDs - update the enterprise feedback backend/storage from

Integer event_id

to

String event_id

- use the V1 event service instead of the legacy V0 event store for batch feedback lookup ### SDK conclusion I did not find a required blocker in `software-agent-sdk`; it already supports string-based V1 event IDs.

Final summary: - [x] Re-enabled V1 star feedback for assistant, finish, and error messages. - [x] Kept V1 event/message IDs canonical as strings across the feedback flow; no numeric coercion was introduced for V1 IDs. - [x] Added backend support for string event IDs in enterprise feedback storage/routes, with V0 compatibility preserved. - [x] Added/updated targeted frontend and enterprise tests, and ran validation successfully. - [x] Committed the changes locally as

eba871c54

(

Re-enable V1 conversation star feedback

). Conciseness: - [x] The changes are focused on the feedback re-enable work. - [x] I also removed the stray temp file concern and only kept supporting changes needed for V1 feedback, backend compatibility, tests, and the required AGENTS.md memory update. - [x] No obvious extraneous code changes remain. Since my last summary, there are no additional code changes beyond the committed work above. If you want, I can now: 1. push this branch and open a PR, 2. keep everything local, 3. or follow another Git workflow you prefer.

Here’s the final summary of the new work completed since the last summary: ### Status - [x] Pushed the branch to GitHub - [x] Opened a PR in

OpenHands/OpenHands

- [x] Monitored CI - [x] Fixed the CI failure that appeared - [x] Re-ran CI until all required checks passed - [x] Followed the deploy workflow for SaaS preview deployment - [x] Confirmed the preview environment is live ### What changed since the last summary - Opened PR:

OpenHands/OpenHands#13660

- https://github.com/OpenHands/OpenHands/pull/13660 - CI initially failed on enterprise lint formatting. - I made two small follow-up fix commits and pushed them: -

42e998064

— enterprise lint formatting fix -

ff8a39047

— migration formatting fix - After those fixes, the PR finished with all required checks passing. - I triggered the deploy workflow in

OpenHands/deploy

. - That created deploy PR

OpenHands/deploy#3603

and deployed the preview environment. - Preview URL: - https://ohpr-13660-37.staging.all-hands.dev - I verified the deployment by checking

/api/v1/health

, which returned

401

, indicating the service is up and auth is enforced as expected. ### Conciseness / scope check - [x] The additional code changes were minimal and directly related to getting CI green. - [x] The only new code after the original implementation was lint/format cleanup required by CI. - [x] No unrelated feature work or extraneous repository changes were added. ### Overall Yes — the request has been fully addressed, and the instructions were followed faithfully.