publish-release.yml pushes ghcr.io/copilotkit/openbot:<version> only. agent-langgraph/Dockerfile builds from the repo root in about 20 s and is what a hosted deployment needs for New coworker to work (docs/deployment.md:85-89: the image does not carry the agent; set MANAGED_AGENT_AG_UI_URL when a Bot is reachable). Today anyone running OpenBot off-Kubernetes has to build and host that image themselves — I currently publish ghcr.io/will-bogusz/openbot-agent-langgraph:v0.0.5 from the tagged source, unchanged, for exactly this reason, and would rather point at yours.
Would you take a PR adding a second build/push job to publish-release.yml for ghcr.io/copilotkit/openbot-agent-langgraph:<version> (same tag scheme, same trigger)? It is platform-neutral: Compose users get the same benefit as the Helm chart, which already assumes a separately built agent image.
publish-release.ymlpushesghcr.io/copilotkit/openbot:<version>only.agent-langgraph/Dockerfilebuilds from the repo root in about 20 s and is what a hosted deployment needs for New coworker to work (docs/deployment.md:85-89: the image does not carry the agent; setMANAGED_AGENT_AG_UI_URLwhen a Bot is reachable). Today anyone running OpenBot off-Kubernetes has to build and host that image themselves — I currently publishghcr.io/will-bogusz/openbot-agent-langgraph:v0.0.5from the tagged source, unchanged, for exactly this reason, and would rather point at yours.Would you take a PR adding a second build/push job to
publish-release.ymlforghcr.io/copilotkit/openbot-agent-langgraph:<version>(same tag scheme, same trigger)? It is platform-neutral: Compose users get the same benefit as the Helm chart, which already assumes a separately built agent image.