automation: Docs update for OpenHands v1.43.0 release

Abierto
#1,170 0 comentarios 0 reacciones 0 asignados Ver en GitHub

Los mantenedores suelen responder en 1 día

Nadie ha tomado este issue todavía.

Evaluación

Dificultad
3/5
Tiempo estimado
1-2 días
Aptitud para principiantes
65/100
Tipo de issue
Documentación
Claridad
Bastante claro
Estado de actividad
Activo

Línea de trabajo

Start with the Settings overview and Agent Canvas automation guide mentioned in the checklist. Update the Settings section table to include Experimental after Secrets, and document that the page is local-backend only; review the automation Activity Log section for the conversation indicator. Confirm each addition against the release notes and linked pull requests, and check that the docs distinguish wizard-created automations from imported ones.

Escrito por el modelo de indexación a partir del texto del issue.

Descripción

agent-in-progress

Release

Release summary

OpenHands v1.43.0 is a routine release covering v1.42.0...v1.43.0. It contains five features, two bug fixes, one performance change, three internal verify-openhands documentation commits, and a group of dependency and pinned-component bumps (agent-server to 1.63.1, automation to 1.34.1, extensions to 0.43.0), plus CI and end-to-end test changes.

The user-facing product changes are concentrated in the Agent Canvas UI:

  • The automation setup wizard now creates the automation already enabled instead of leaving it disabled (PR #19769).
  • A new Settings > Experimental page exposes per-user yes/no feature flags, with Conversation reactions as the first flag (PR #19503). The page reads its list from the agent server and is described in the PR as local-backend only — Cloud lists no features.
  • Three chat changes are described in their PRs as experimental and flag-gated: decision cards for the SDK's ask_user questions (PR #19789), display of the agent's own message edits (PR #19794), and emoji reactions on messages (part of PR #19503).
  • Automation run rows show a speech-bubble indicator when the run has an associated conversation (PR #19903).

The two fixes restore intended UI behavior: a setup conversation keeps its form when you switch conversations and come back (PR #19860), and Experimental is listed last in the settings navigation, after Secrets (PR #19881). The performance change virtualizes the conversation message list so mounted rows stay bounded on long conversations (PR #17820).

Documentation review checklist

  • Create automations enabled from the setup wizard (PR #19769) — probably no documentation change. The wizard's final create now sends enabled: true for both preset and custom automations; server-draft saves and preflight validation still send enabled: false. /openhands/usage/automations/creating-automations already says a created automation "starts enabled" by default, and /openhands/usage/agent-canvas/managing-automations says only that the automation "then appears in the automation list like any other". Confirm no page still tells readers to enable a wizard-created automation as a second step, and that the "Imported automations are created disabled" statement is not confused with wizard creation. Source: https://github.com/OpenHands/OpenHands/pull/19769
  • Settings > Experimental page for user feature flags (PR #19503) — documentation needed. Adds a settings screen at /settings/experimental that lists the agent server's experimental features as switches, saves each toggle immediately, and shows an empty list against an agent server without the endpoint. The page is local-backend only; the PR states Cloud has no Features entry. The settings overview page in the docs lists the Settings sections and does not mention Experimental. Source: https://github.com/OpenHands/OpenHands/pull/19503
  • Conversation reactions (first flag in PR #19503) — documentation needed, gated on the flag. While the flag is on, the agent's reactions appear as emoji chips inside the message they target, the agent's messages get a small picker, clicking your own chip removes it, and the react/unreact tool calls are hidden from the chat (a failed call stays visible). With the flag off the picker is gone and the reactions endpoint answers 409, but existing chips stay. No docs page currently describes reactions. Source: https://github.com/OpenHands/OpenHands/pull/19503
  • Experimental decision cards (PR #19789, fixes #19773) — needs maintainer confirmation. Renders the SDK's ask_user questions as cards with choices, consequences, a recommendation and optional notes, answered through the conversation service; answers survive reloads, and Markdown/HTML transcript exports keep the question and answer. The PR describes it as under an experimental flag and shows the flag as Settings > Experimental > Decision cards. Docs currently mention ask_user nowhere. Decide whether experimental UI gets documented now or after it graduates. Source: https://github.com/OpenHands/OpenHands/pull/19789
  • Agent message edits, experimental (PR #19794, fixes #19793) — needs maintainer confirmation. Agent messages and finish answers show the edited text in place with an "edited" mark; each edit is a new MessageEditEvent in an append-only log, and the text renders as Markdown. The PR notes edits display whether or not the flag is currently on, because an edit exists only if the conversation was started with the flag on. Source: https://github.com/OpenHands/OpenHands/pull/19794
  • Conversation indicator on automation run rows (PR #19903) — documentation needed (small). Each run row is already a link to the run's conversation; a non-interactive speech-bubble icon now marks rows whose run reported a conversation_id, with the tooltip and accessible name "Conversation started". Runs without a conversation show no icon. The Activity Log section of /openhands/usage/agent-canvas/managing-automations explains that script runs have no conversation and points at View logs; it could name the icon so readers can tell the two kinds of row apart. Source: https://github.com/OpenHands/OpenHands/pull/19903
  • Keep the setup form when switching conversations (PR #19860, fixes #19871) — probably no documentation change. A setup conversation now keeps its form for the rest of the page session instead of rebuilding a blank draft on return; the New Automation page still starts clean, and newer unsaved edits win over the saved copy. This matches what /openhands/usage/agent-canvas/managing-automations already promises about saved drafts; no new user action is introduced. Source: https://github.com/OpenHands/OpenHands/pull/19860
  • List Experimental last, after Secrets (PR #19881, fixes #19880) — documentation needed, together with the Settings > Experimental page. Changes the settings nav and command-menu order so Experimental follows Secrets; order only, and Cloud still hides Experimental. The Settings section table in the docs should list Experimental in that position. Source: https://github.com/OpenHands/OpenHands/pull/19881
  • Virtualize the conversation message list (PR #17820) — probably no documentation change. Bounds the mounted rows on long conversations, keeps lazy older-page loading with scroll anchoring, autoscrolls the streamed tail only while pinned to the bottom, and removes the development-only ?virtualize=0 escape hatch. No new setting, option, or user-visible control is described, so there is nothing to document beyond observed behavior. Source: https://github.com/OpenHands/OpenHands/pull/17820
  • verify-openhands map maintenance and mapping of Experimental (PRs #19854, #19873, #19882) — probably no documentation change. Changes are confined to the internal .agents/skills/verify-openhands/ feature map in OpenHands/OpenHands and are not part of the published docs site. Source: https://github.com/OpenHands/OpenHands/pull/19882
  • Dependency and component bumps (PRs #19866–#19870, #19876, #19879, #19884, #19883, #19895, plus CI PRs #19690 and #19615) — probably no documentation change. agent-server rises to 1.63.1, automation to 1.34.1, extensions stays at 0.43.0; the rest are npm dev/runtime bumps. No docs page states these pinned versions as a requirement, and the only version gate the docs describe (agent-server 1.31.0 for a toggle in /openhands/usage/agent-canvas/agent-profiles) is unaffected. Source: https://github.com/OpenHands/OpenHands/compare/v1.42.0...v1.43.0
  • End-to-end test changes (PRs #19900, #19906) — probably no documentation change. Release-harness alignment with SDK 1.63.1 and an e2e assertion change; no user-facing behavior described. Source: https://github.com/OpenHands/OpenHands/pull/19906

Pull request review

  • https://github.com/OpenHands/OpenHands/pull/19503 — 35 changed files, all under src/, __tests__/, .pr/ and package-lock.json; no docs files. Adds the experimental-features service, the /settings/experimental route, a settings-nav entry, the reactions service and the message reaction chips. Most documentation-relevant PR in the release. Two points the docs must state exactly: the feature list comes from GET /api/settings/experimental (a new feature needs no Canvas change), and the page is local-backend only. The PR body also records that the reaction picker is not offered on planning-agent messages.
  • https://github.com/OpenHands/OpenHands/pull/19881 — 4 changed files (src/components/features/command-menu/command-menu-items.tsx, src/constants/settings-nav.tsx, one test, one screenshot); no docs files. Pure ordering change for Secrets and Experimental. Docs impact is the Settings table order only.
  • https://github.com/OpenHands/OpenHands/pull/19789 — 20 changed files under src/, __tests__/ and .pr/; no docs files. Adds the decision-card component, an answerAskUser call on the conversation client, grouping and export changes. Documentation-relevant only if experimental chat UI is in scope; the PR itself describes the flag as Settings > Experimental > Decision cards.
  • https://github.com/OpenHands/OpenHands/pull/19794 — 17 changed files under src/, __tests__/ and .pr/; no docs files. Adds the message-edits component and an "edited" mark in 15 languages. Its notes contain the one non-obvious rule for docs: edits render even when the flag is now off, because they exist only in conversations started with the flag on.
  • https://github.com/OpenHands/OpenHands/pull/19903 — 4 changed files (activity-log-item.tsx, one test, translation.json, one screenshot); no docs files. Adds a non-interactive MessageSquare icon with the accessible name "Conversation started" when run.conversation_id is set. Small, precise docs opportunity in the Activity Log section.
  • https://github.com/OpenHands/OpenHands/pull/19769 — 8 changed files (automation-setup-panel.tsx, one test, .pr/ screenshots and a capture script); no docs files. Only the create call's enabled flag changes; drafts and preflight keep enabled: false. The docs statements about drafts and about newly created automations remain correct.
  • https://github.com/OpenHands/OpenHands/pull/19860 — 11 changed files (automation-setup-panel.tsx, a new automation-form-session.ts, tests, an e2e spec); no docs files. Restores intended behavior for returning to a setup conversation. No documentation impact found beyond the existing draft-persistence text.
  • https://github.com/OpenHands/OpenHands/pull/17820 — 65 changed files, almost all tests plus chat rendering internals and package.json; no docs files. No user-facing configuration is added, and the development-only ?virtualize=0 parameter was deleted, so there is nothing to document as a feature.
  • verify-openhands PRs #19854, #19873, #19882 and the dependency, CI and e2e PRs: changes confined to .agents/skills/verify-openhands/, lockfiles, workflows and test harnesses; no docs-site files and no user-facing behavior described. No documentation impact found.

Human testing required

  • Settings > Experimental on a local backend (PR #19503, PR #19881): open Settings on a local agent-server backend. Expected: an Experimental entry listed last, after Secrets, in both the settings nav and the command menu, and a page that lists the server's flags with Conversation reactions present. Toggle it on, reload, and expected: it stays on. Then switch to a Cloud backend and expected: no Experimental entry.
  • Reactions with the flag on (PR #19503): start a new conversation after enabling the flag and ask the agent to react to your message. Expected: an emoji chip appears inside your message, no Reacted tool card is shown, and the agent's messages offer the picker. Click your own chip and expected: the reaction is removed and stays removed after a reload. Turn the flag off and expected: the picker disappears, existing chips remain, and new conversations have no reaction tools. The PR reports a 409 from the reactions endpoint with the flag off; confirm what the UI shows in that case.
  • Decision cards with the flag on (PR #19789): enable Decision cards, then ask the agent a question that makes it call ask_user with two choices. Expected: a card with the choices, consequences and a recommendation; selecting the non-recommended choice with a note resumes the agent, and the choice and note persist after a reload at desktop and phone widths without a generated answer bubble. Also check the Markdown and HTML transcript exports contain the question, options and note once, with tool details both shown and hidden.
  • Message edits with the flag on (PR #19794): with Message edits enabled, ask the agent a question, answer it, and ask the agent to correct its earlier message. Expected: the message shows the new text with an "edited" mark and no tool card; after a reload the text and a single mark remain. Then disable the flag and reopen that conversation: expected: the edit still displays.
  • Wizard-created automation starts enabled (PR #19769): create an automation through the setup view with a schedule. Expected: the automation appears enabled in the automation list and runs at the next scheduled time, with no separate enable step. Also confirm an imported automation is still created disabled, as documented.
  • Setup form survives switching conversations (PR #19860): open a setup conversation, let the agent fill the form, edit a field yourself, switch to another conversation and back. Expected: your newer edit is still there and the draft still shows as saved.
  • Run-row conversation indicator (PR #19903): open an automation's run list containing one run that started a conversation and one script run that did not. Expected: the first row shows a speech-bubble icon with the tooltip and accessible name "Conversation started"; the script run shows no icon and still offers View logs.
  • Long conversation behavior (PR #17820): open a conversation with a few hundred events. Expected: scrolling stays responsive, scrolling to the top loads older pages with the visible content staying put, and a streamed event autoscrolls only when the view is pinned to the bottom. Also expand a tool group and expected: the rows below re-lay out without overlapping. The PR's own measurements come from an agent-server 1.53.0 stack, so confirm on this release's agent-server 1.63.1.

Suggested documentation locations

  • /openhands/usage/settings/experimental-settings (new page, following the settings/*-settings.mdx naming) — the natural home for the feature-flag page: what Settings > Experimental lists, that toggles save immediately, that the list comes from the agent server, that it is local-backend only, and each documented flag with its effect. Requires a docs.json entry in the Settings group.
  • /openhands/usage/agent-canvas/customize-and-settings — its Settings section table lists every settings section and currently omits Experimental; add the row in the new position, after Secrets, and link the new page.
  • /openhands/usage/agent-canvas/conversations — already the home for message-level behavior (branching, timestamps, transcripts, attachments); the place for reaction chips and the picker, and for decision cards and the "edited" mark if the experimental features are documented now.
  • /openhands/usage/agent-canvas/managing-automations — the Activity Log section describes runs without a conversation and points at View logs; add the "Conversation started" indicator there, and confirm the setup-assistant section still matches PRs #19769 and #19860.
  • /openhands/usage/automations/creating-automations — verify the "It starts enabled" statement under After Creation still matches the wizard path changed by PR #19769.
  • /openhands/usage/agent-canvas/release-notes/v1.43.0 (new page) plus the Release Notes group in docs.json — repository convention keeps the six newest release pages in the sidebar and moves the oldest to previous.mdx, so adding v1.43.0 moves v1.37.0 to the previous-release page.

Source links

This issue was created by an AI agent (OpenHands) on behalf of the user.

Lenguaje dominante
MDX
Estrellas
19
Forks
104
Merge medio
2 d 12 h
PR fusionados (30 d)
80

Preparar el entorno

  • Sin Dockerfile ni archivo de Docker Compose
  • Tiene una plantilla de pull request
  • Sin guía de contribución

Primeros pasos

  1. Lee el issue completo y luego la guía de contribución del proyecto.
  2. Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
  3. Haz un fork del repositorio y trabaja en una rama.
  4. Abre un pull request que haga referencia al número del issue.

Más de OpenHands/docs

Todos los issues de OpenHands/docs

Issues similares

Más issues de Documentation

Recibe los nuevos issues en tu correo

Un resumen breve de issues de GitHub para principiantes.