Skip to content

docs(opencode): document integration limitations + manual activation guide (#317)#388

Open
tirth8205 wants to merge 1 commit into
Egonex-AI:mainfrom
tirth8205:docs/opencode-integration-limitations
Open

docs(opencode): document integration limitations + manual activation guide (#317)#388
tirth8205 wants to merge 1 commit into
Egonex-AI:mainfrom
tirth8205:docs/opencode-integration-limitations

Conversation

@tirth8205

Copy link
Copy Markdown
Contributor

Summary

  • Add docs/platforms/opencode.md documenting that OpenCode (1.15.11) runs the deterministic scan/import/batch scripts but skips every LLM subagent phase — because OpenCode does not implement Claude Code's Task tool. Result is a structural-only knowledge graph (~14% coverage vs ~70% in Claude Code).
  • The new page covers: status, per-component compatibility matrix, what you get (file nodes + import edges, no summaries/tour/layers), a step-by-step manual activation procedure (scan-project.mjsextract-import-map.mjscompute-batches.mjs → minimal graph assembly → dashboard launch), optional ~/.config/opencode/command/*.md stubs for slash-command registration, and a future-direction note about Task-equivalent dispatch.
  • Update the README Platform Compatibility table: the OpenCode row now flags "Structural-only" and links to the new page, and a short callout under the table explains the limitation at a glance.

No code changes — docs only.

Test plan

  • Render docs/platforms/opencode.md locally (GitHub markdown / mdcat / preview) and confirm headings, code fences, and tables format correctly.
  • Click through the README link from the Platform Compatibility table to the new doc.
  • Spot-check that the deterministic script paths and CLI signatures match understand-anything-plugin/skills/understand/{scan-project,extract-import-map,compute-batches}.mjs.

Closes #317.

Co-Authored-By: Claude Opus 4.7 (1M context) noreply@anthropic.com

…guide (Egonex-AI#317)

OpenCode 1.15.11 installs the plugin and the deterministic scan/import/batch
scripts work, but it lacks Claude Code's Task tool — so the LLM subagent phases
(file-analyzer, assemble-reviewer, architecture-analyzer, tour-builder,
graph-reviewer) never run. Result is a structural-only graph (~14% coverage).

Add docs/platforms/opencode.md covering:
- Status: structural-only (what runs, what doesn't, why)
- Compatibility matrix per component
- What you get (file nodes + import edges, no summaries/tour/layers)
- Manual activation procedure using scan-project.mjs, extract-import-map.mjs,
  compute-batches.mjs, a minimal graph-assembly snippet, and dashboard launch
- Optional ~/.config/opencode/command/*.md stubs for slash-command registration
- Future direction (Task-equivalent dispatch would unlock the full pipeline)

Link the new page from the README Platform Compatibility table (OpenCode row
now flags "Structural-only") and add a short OpenCode note below the table.

Closes Egonex-AI#317.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

feat(opencode): document integration limitations + provide manual activation guide

1 participant