Context
I'm a heavy user of opencode-mem (v2.19.4) on OpenCode V1 (1.18.3), relying on:
- Cross-session project memory (SQLite + vector search)
- Auto-capture and user profile learning
- memory tool integration in agent workflows
- Web UI for memory browsing
I noticed PR #101 already migrated the internal AI calls to OpenCode V2 SDK (session.prompt), which is great progress. However, OpenCode V2's migration guide states:
V1 plugins will not work in V2. The V2 plugin API is still being finalized during beta.
Questions
-
Is there a planned timeline for opencode-mem to fully support the OpenCode V2 plugin API? Given that the internal SDK is already using V2 APIs, I'm hoping the plugin-level migration is on the near-term roadmap.
-
What will the migration path look like for existing users? Will existing memory data (SQLite databases) transfer seamlessly? Will the config format (opencode-mem.jsonc) remain compatible?
-
Are there any V2-native features you're considering (e.g., deeper integration with V2's session tree, agent-specific memory scopes)?
My Setup
- OpenCode V1 1.18.3 on Windows
- opencode-mem v2.19.4 with custom DeepSeek V4 Flash endpoint for memory extraction
- ~50+ project memory records across multiple Roblox game projects
- Currently blocked from migrating to V2 until opencode-mem is ready
Thanks for the great work — cross-session memory is a critical part of my workflow.
Context
I'm a heavy user of opencode-mem (v2.19.4) on OpenCode V1 (1.18.3), relying on:
I noticed PR #101 already migrated the internal AI calls to OpenCode V2 SDK (session.prompt), which is great progress. However, OpenCode V2's migration guide states:
Questions
Is there a planned timeline for opencode-mem to fully support the OpenCode V2 plugin API? Given that the internal SDK is already using V2 APIs, I'm hoping the plugin-level migration is on the near-term roadmap.
What will the migration path look like for existing users? Will existing memory data (SQLite databases) transfer seamlessly? Will the config format (opencode-mem.jsonc) remain compatible?
Are there any V2-native features you're considering (e.g., deeper integration with V2's session tree, agent-specific memory scopes)?
My Setup
Thanks for the great work — cross-session memory is a critical part of my workflow.