Docs: Child Conversations incorrectly implies the child's result is returned to the parent
Los mantenedores suelen responder en 1 día
Evaluación
- Dificultad
- 2/5
- Tiempo estimado
- 1-3 horas
- Aptitud para principiantes
- 82/100
- Tipo de issue
- Documentación
- Claridad
- Bien especificado
- Estado de actividad
- Activo
- Stack tecnológico
- markdown
- Área
- documentation
Línea de trabajo
Lee la sección Child Conversations en openhands/usage/agent-canvas/conversations.mdx y compara su redacción con la descripción de la herramienta launch_child_conversation mencionada. Actualiza la sección para distinguir la confirmación de inicio de la respuesta final del hijo y explicar que el hijo se ejecuta de forma independiente, sin bloquear al padre. El trabajo estará terminado cuando quede claro que la respuesta final no se entrega automáticamente y que debe leerse abriendo o consultando la conversación del hijo.
Escrito por el modelo de indexación a partir del texto del issue.
Descripción
Summary
The Child Conversations section of openhands/usage/agent-canvas/conversations.mdx states that a child conversation's result is returned to the parent conversation. In the current product this is not accurate: children launched via the launch_child_conversation tool run independently, and the child's final response is never delivered to the parent automatically. The parent only receives a launch acknowledgement (the child's id/URL/initial status), and a user has to open the child conversation to read its output.
This wording has already caused user confusion in the community (a user read it as "the child's answer is pushed back to the parent" and spent time chasing a handoff that does not exist).
Current text
openhands/usage/agent-canvas/conversations.mdx (Child Conversations section):
The child remains linked to its parent, and its result is returned to the parent conversation. Agent Canvas validates the launch inputs before creating the child conversation.
Why it is confusing
Two separate things are being conflated:
- The launch result / link is returned. When the agent calls
launch_child_conversation, the frontend posts a machine-generated follow-up message beginning with[child-conversation]back into the parent, carrying the child's conversation id, URL, target and initial status (or corrective guidance on a malformed call). This happens at launch time — not when the child finishes. - The child's final answer is NOT returned. The child runs on its own ("It does not block you, you do not see its output", per the tool description). Its final response is not forwarded to the parent on completion. The
parent_conversation_id/sub_conversation_idsrelationship is persisted server-side but is used for discovery, recovery and deletion bookkeeping, not for message delivery.
The phrase "its result is returned to the parent conversation" reads naturally as (2), when what actually happens is only (1).
Contrast this with the SDK Task tool (software-agent-sdk), where delegation is blocking and the sub-agent's final response genuinely is returned in the tool observation. Because the docs use the same word "result" for both, readers reasonably assume the Canvas child behaves the same way — it does not.
Acceptance criteria
- The Child Conversations section no longer states or implies that a child's final response/output is automatically returned to the parent.
- The section distinguishes the two events described above:
- the automatic launch acknowledgement posted back into the parent (child id/URL/initial status, the
[child-conversation]follow-up), and - the child's final result, which is not auto-delivered and requires the parent (or user) to open/query the child conversation.
- the automatic launch acknowledgement posted back into the parent (child id/URL/initial status, the
- The section explicitly states that a child launched this way is non-blocking and independent — the parent does not wait for it and does not automatically see its output (consistent with the
launch_child_conversationtool description inOpenHands/OpenHands: "The child runs on its own. It does not block you, you do not see its output..."). - If a contrast with the SDK Task tool is added, it correctly notes that Task tool delegation is blocking and does return the sub-agent's final response, so the two are not confused.
- Wording is consistent with the child-conversation behavior implemented in
OpenHands/OpenHands(src/api/launch-child-conversation-client-tool.ts,src/services/child-conversation-launch.ts) andOpenHands/software-agent-sdk(agent-server child linking,openhands-tools/openhands/tools/task).
References
openhands/usage/agent-canvas/conversations.mdx— Child Conversations section (line 12)OpenHands/OpenHandssrc/api/launch-child-conversation-client-tool.ts— tool description and "What comes back" textOpenHands/OpenHandssrc/services/child-conversation-launch.ts—reportLaunchResult(launch receipt only)OpenHands/software-agent-sdkopenhands-agent-server/openhands/agent_server/models.py—parent_conversation_id/sub_conversation_ids(link, not delivery)- Community thread: user reported with Agent Canvas 1.25 that the child's response did not return to the parent, citing this documentation line.
This issue was drafted by an AI agent (OpenHands) on behalf of a community member.
- Lenguaje dominante
- MDX
- Estrellas
- 16
- Forks
- 71
- Merge medio
- 2 d 14 h
- PR fusionados (30 d)
- 72
Preparar el entorno
- Sin Dockerfile ni archivo de Docker Compose
- Tiene una plantilla de pull request
- Sin guía de contribución
Primeros pasos
- Lee el issue completo y luego la guía de contribución del proyecto.
- Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
- Haz un fork del repositorio y trabaja en una rama.
- Abre un pull request que haga referencia al número del issue.
Más de OpenHands/docs
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 84/100
Los mantenedores suelen responder en 1 día
-
Moonshot provider documentation references a deprecated model, setup fails per doc recommendationsAbierto
Dificultad 1/5 Menos de una hora Aptitud para principiantes 88/100
Los mantenedores suelen responder en 1 día
-
enhancement
Dificultad 2/5 1-3 horas Aptitud para principiantes 68/100
OpenHands/docs#658 · 2 comentarios ·
Los mantenedores suelen responder en 1 día
-
Dificultad 3/5 1-2 días Aptitud para principiantes 30/100
Los mantenedores suelen responder en 1 día
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 35/100
Los mantenedores suelen responder en 1 día
Todos los issues de OpenHands/docs
Issues similares
-
Dificultad 1/5 Menos de una hora Aptitud para principiantes 72/100
AFLplusplus/AFLplusplus#2903 ·
-
Dificultad 1/5 Menos de una hora Aptitud para principiantes 90/100
Los mantenedores suelen responder en 1 día
-
infrastructure qa
Dificultad 2/5 1-3 horas Aptitud para principiantes 84/100
khuisman/mcp-gee-sweet#1068 ·
Los mantenedores suelen responder en 1 día
-
Docs: how to get more careful answers (Effort, reasoning models, Reasoning level, Deliberate)Abiertodocumentation priority:p1
Dificultad 2/5 1-3 horas Aptitud para principiantes 78/100
yeixio/toskar-core#697 ·
Los mantenedores suelen responder en 1 día
-
The book has a Nix install section that chelis.ch lacks, and it names two of the three systemsAbiertoarea:ecosystem documentation
Dificultad 2/5 1-3 horas Aptitud para principiantes 78/100
Chelis-Lang/chelis#3866 ·
Los mantenedores suelen responder en 1 día