馃悶 Bug Report
Describe the bug
When multiple Honcho clients use the same workspace, the OpenCode plugin only recalls conclusions written by its own AI peer. Conclusions created by another configured agent in that workspace cannot be recalled through OpenCode, even when they concern the same user peer.
This prevents a shared workspace from providing cross-agent memory in OpenCode.
Is this a regression?
Unknown. This was observed with the current plugin release.
To Reproduce
- Configure Honcho clients for two agents with:
- the same workspace, for example
shared-test-workspace
- the same user peer, for example
test-user
- different AI peers, for example
agent-a and opencode
- Using
agent-a, create a conclusion for test-user, such as: The test marker is blue.
- Start an OpenCode session configured with AI peer
opencode.
- Ask OpenCode to recall the test marker for
test-user.
Expected behaviour
OpenCode should support recalling conclusions written by selected peer agents in the same workspace, while retaining the originating observer in the returned metadata.
A configuration option analogous to linkedHosts, or an explicit multi-observer query option, would allow users to opt into this behavior.
Media prove
Not applicable.
Your environment
- OS: Linux 6.17.0-20-generic x86_64
- Browser name and version: Not applicable
- Honcho Server Version: Official
latest container image
- Honcho Client Version:
@honcho-ai/opencode-honcho 0.1.3
- OpenCode Version: 1.18.13
Additional context
The conclusion is present when queried directly with the originating observer (agent-a) and the shared workspace. The OpenCode integration appears to query only its own observer (opencode).
馃悶 Bug Report
Describe the bug
When multiple Honcho clients use the same workspace, the OpenCode plugin only recalls conclusions written by its own AI peer. Conclusions created by another configured agent in that workspace cannot be recalled through OpenCode, even when they concern the same user peer.
This prevents a shared workspace from providing cross-agent memory in OpenCode.
Is this a regression?
Unknown. This was observed with the current plugin release.
To Reproduce
shared-test-workspacetest-useragent-aandopencodeagent-a, create a conclusion fortest-user, such as:The test marker is blue.opencode.test-user.Expected behaviour
OpenCode should support recalling conclusions written by selected peer agents in the same workspace, while retaining the originating observer in the returned metadata.
A configuration option analogous to
linkedHosts, or an explicit multi-observer query option, would allow users to opt into this behavior.Media prove
Not applicable.
Your environment
latestcontainer image@honcho-ai/opencode-honcho0.1.3Additional context
The conclusion is present when queried directly with the originating observer (
agent-a) and the shared workspace. The OpenCode integration appears to query only its own observer (opencode).