Skip to content

feat: add continue_run tool for resuming diagram/plot runs#190

Open
claytonlin1110 wants to merge 1 commit into
llmsresearch:mainfrom
claytonlin1110:feat/mcp-continue-run
Open

feat: add continue_run tool for resuming diagram/plot runs#190
claytonlin1110 wants to merge 1 commit into
llmsresearch:mainfrom
claytonlin1110:feat/mcp-continue-run

Conversation

@claytonlin1110
Copy link
Copy Markdown
Contributor

Summary

Adds a new MCP tool so clients (Cursor, Claude Code, etc.) can continue an existing outputs/run_* job the same way the CLI does with --continue-run and optional --feedback, without re-entering full methodology text or re-running from a blank run.

Closes #189

Motivation

generate_diagram / generate_plot always start a new pipeline. Run continuation was already supported in the core pipeline and CLI; MCP users had no supported path, while other MCP flows already expose resume-style parameters for batch/orchestration.

Changes

  • New @mcp.tool continue_run in mcp_server/server.py: load_resume_state(settings.output_dir, run_id) → continue_run(..., user_feedback=feedback).
  • Docs: mcp_server/README.md, root README.md, copilot-instructions.md, server.json.
  • Test: tests/test_utils.py::test_mcp_continue_run_loads_state_and_calls_pipeline with PaperBananaPipeline mocked (skipped when fastmcp is not installed).

@claytonlin1110
Copy link
Copy Markdown
Contributor Author

@dippatel1994 Would you please review?

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.

[Feature]: MCP: expose diagram run continuation (continue_run + optional user feedback)

1 participant