Repository navigation
[release] v0.128.9 - #8198
Merged
Merged
[release] v0.128.9#8198
Conversation
added 30 commits
November 5, 2026 22:15
…e under beta Entire-Checkpoint: 01M6PS8K1SVF63XJW5N6CJZJBN
…e item, pretty JSON, auto-refresh)
Entire-Checkpoint: 01M6PV17PVDNMHNEDNQ62J68AX
…tarter columns and search in the /m test set editor - Hide only internal data keys (__* and testcase_dedup_id) from test set columns and dirty checks; meta, tags, flags and id are user columns inside testcase.data. - Read a collapsed group cell by path, so customer.orders shows its object. - Open a new test set with input and correct_answer and one empty row, as /w does. - Add a search box over the /m test set grid that filters loaded rows.
Entire-Checkpoint: 01M6PVH75XXEBJYXMXAYPKDSAM
- Derive columns from every row, so a search does not hide columns. - Clear the selection when the search changes, so a bulk delete only hits visible rows. - A testcase with empty data gives no columns, not its own fields.
…fresh skips busy ticks, copy visible on touch
Entire-Checkpoint: 01M6PW2YSYNKWHZ97P9V9N8QDD
Entire-Checkpoint: 01M6PWM0MEQTJ9TCN447PKY42C
…t set has no rows The starter columns go into the column list too, as in /w, so a row added after every row was deleted still gets the input and correct_answer fields. The grid toolbar shows above the table while it has no rows, because the table hides its own toolbar then.
…llapsed values exactly - Bulk delete takes only selected rows the grid still shows. - Column discovery adds rows a search matches past the first 20. - The data controller reads each row as data, so user columns named id or data stay. - A collapsed group path resolves keys that hold a dot at any depth.
… /m now has Entire-Checkpoint: 01M6PYCR846B1BQQ0G7V00STBD
…sh shares one pending guard and re-arms after each refresh
…at/m-parity-app-menu
…o feat/m-parity-app-menu
…views Drop the last per-view stage check on the developer side. Add tests that beta developer mode shows every view and that stages and Preferences switches stay as they are without it. Entire-Checkpoint: 01M6QZJP4R7HRYSQG9TAXD1DQE
…n group, Observability, Invite Teammate
Column add, rename and delete in the /m test set editor now wait for one Commit with a message, together with row edits, as in /w. Discard drops them. - @agenta/entities: column reducers keep the pending operations in the form the server applies them (removals and renames read on the saved names), so a renamed then deleted column is removed, and a deleted then re-added column comes back empty. The column edit atoms behind revisionMolecule.tableReducers rewrite edited and new rows at once; saved rows get the operations when read. - @agenta/entities: commitPreviewAtom, a readable commit preview with no row ids or internal fields. - /m: useTestsetEditor stages column changes instead of saving them; the dialogs no longer promise a save; Import stays on with unsaved changes; the commit dialog shows the preview beside the message, like /w.
…empty states Entire-Checkpoint: 01M6R0W9KQEFHPPQK5F29H3ZW0
…app font in /m Entire-Checkpoint: 01M6R1CZ9QKZNK08CEY3XJJXR0
Entire-Checkpoint: 01M6R1FEV76ZMS9CY3AEXPGC4R
…p names in tab titles In beta developer mode the developer list and detail pages use the full width, the test set and evaluation results headers are compact, the app overview tab names the app, and the evaluations tab says Evaluations. Without beta developer mode nothing changes. Entire-Checkpoint: 01M6R1VPFTV881EK864HM18ABC
…table Entire-Checkpoint: 01M6R2999R63CM3PJV8Z7KCD5C
The new evaluation sheet in /m now uses the /w layout: the name on top, a step list on the left with a check and chips for each pick, and the open step's table on the right. Each step shows the /w columns. - Application: Name, Kind, Type, Created At, a type filter and a search. - Revision: one row per variant with version, deployments, model, date, author and commit notes; older revisions open under it. - Test set: an inline table with Name, Commit Message, Date Created and Created by (a name, not a user id). A pick takes the latest revision and moves to Evaluators. - Evaluators: the /w columns, the Feedback column for human evaluators, older versions, and Create new evaluator (opens Evaluators in a new tab). - Close, Escape, overlay click and Cancel ask before they drop a draft with input; an untouched draft closes at once. The playground opens the same sheet, so it gets the same steps. Entire-Checkpoint: 01M6R2JESW6N3AG1ZFAC6A5Z3T
… invite link in and out of beta developer mode
…nd an app switcher
feat(frontend): Evaluators list with one row per evaluator, select all and coloured tags in /m
…rity-human-evaluator-form # Conflicts: # web/mobile/src/features/evaluators/EvaluatorRowCells.tsx # web/mobile/src/features/evaluators/EvaluatorsScreen.tsx # web/mobile/src/features/evaluators/EvaluatorsTable.tsx
…-form feat(frontend): create and edit human evaluators in /m
…drawer feat(frontend): /w steps and tables in the /m new evaluation sheet
…t-header feat(frontend): match /w on the /m evaluations list header, menu and empty states
fix(frontend): playground message editors fit their text and use the app font in /m
…rity-page-frame # Conflicts: # web/mobile/src/features/app-overview/AppOverviewScreen.tsx # web/mobile/src/features/archived-apps/ArchivedAppsScreen.tsx # web/mobile/src/features/endpoints/EndpointsScreen.tsx # web/mobile/src/features/prompts/PromptsScreen.tsx # web/mobile/src/features/variants/VariantsScreen.tsx
feat(frontend): full-width developer pages and compact detail header in /m
fix(frontend): Prompts rename blocks duplicate app names
…rity-prompts-list # Conflicts: # web/mobile/src/features/prompts/PromptsScreen.tsx
feat(frontend): Open in the Prompts row menu and red confirm buttons
…rity-create-prompt # Conflicts: # web/mobile/src/features/prompts/PromptsScreen.tsx
feat(frontend): create a prompt from the /m Prompts page
feat(frontend): named, linked reference chips in the /m trace drawer
…rity-obs-table # Conflicts: # web/mobile/tests/unit/observabilityScreenAppFilter.render.test.tsx
feat(frontend): /m observability toolbar matches /w (refresh, filter icons, export rule, tab in URL)
feat(frontend): /m traces table polish (role colors, row lines, rate-limit state)
feat(frontend): pick the test set revision when adding traces in /m
feat(frontend): shared evaluation logic for /m, and the live card shows the run value
feat(frontend): /m live evaluations like /w
fix(frontend): default the invite role to one the plan offers
chore(frontend): test-design skill and Testing Library for /m tests
…oy button and tests - Clear the drawer's committed revision when the host asks for another revision. - Make the drawer config read-only while the next revision loads. - Keep Deploy off until the revision loads, and read the variant label live. - Use the destructive tokens for the solid red confirm button. - Give the editable drawer tests a timeout longer than their polling. - Expect Stop, not Pause, in the run details tests. - Reword two test-design rules on mocks.
Contributor
Author
Railway Preview Environment
Dependent tests finished. Post /preview for manual review. Environment absence verified; retained storage requires separate provider accounting. Post |
fix(frontend): settle v0.128.9 review threads on the /m drawers, deploy button and tests
…ry and costs A stopped live run reads Stopped in the runs list and the overview, from one rule that keeps a final status. The overview reads the query revision the run stored. The per-trace cost and tokens of several evaluators add their totals and divide by the trace count, and a run with two app steps reads the first one. The solid red confirm button has white text in both themes, as /w.
…no count A total with no trace count says nothing per trace. When every evaluator has a count, the totals add up as before; otherwise the means add up, and with a missing mean there is no average.
…irm-text fix(frontend): settle v0.128.9 review threads on live run status, query and costs
mmabrouk
approved these changes
Nov 6, 2026
mmabrouk
left a comment
mmabrouk
left a comment
Member
There was a problem hiding this comment.
🤖 The AI agent says: Approved on Mahmoud's instruction after the v0.128.9 production deploy (workflow 29 runs 42262732173 for US and 42264050434 for EU) passed the live checks.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
New version v0.128.9 in
Contents
Batch 1 of the /m developer mode parity work: 36 PRs (#8133 to #8178, see the merged PRs on this branch). Almost all of it sits behind beta developer mode (
AGENTA_BETA_DEVELOPER_MODE), which stays off in production. Two PRs change things outside it: #8162 fixes the OSS invite dialog in /w so it defaults to a role the plan offers, and #8163 adds a test-design skill and two Testing Library devDependencies to /m. Release fixes: #8200 and #8202 (review threads, run status wording "Stopped", stored query revision for run overviews, run metric totals, white text on red confirm buttons).Final QA
53942d0d: pass, app shell 884.2 KB.Deploy steps
None. Frontend only (web and web-mobile). No migration, no new env var. Production keeps
AGENTA_BETA_DEVELOPER_MODEunset.