Skip to content

roadmap.md

Latest commit

 

History

History
111 lines (71 loc) · 2.97 KB

roadmap.md

File metadata and controls

111 lines (71 loc) · 2.97 KB

ResearchKit Roadmap

Summary

The next phase should build on the current architecture rather than adding parallel systems. The highest-value work is to deepen the existing research workflow, finish the placeholder agents, and tighten the end-to-end writing loop in the editor.

Near-Term Priorities

1. Deepen the Research Agent

What exists now:

  • literature search
  • citation verification
  • BibTeX output
  • read-only workspace context

What is missing:

  • stronger related-work synthesis tied directly to the current draft
  • better paper ranking and grouping for section-level writing tasks
  • improved transfer of research artifacts back into the Main Agent workflow

Likely implementation areas:

  • services/researchkit/researchkit/agents/research_agent.py
  • services/researchkit/researchkit/literature/
  • frontend rendering for richer research artifacts

2. Finish the Figure Agent

What exists now:

  • delegation hook in the Main Agent
  • placeholder FigureAgent class

What is missing:

  • real execution flow for figure generation
  • file outputs and patch/diff integration
  • support for code-driven plot generation, TikZ, or diagram assets

Likely implementation areas:

  • services/researchkit/researchkit/agents/figure_agent.py
  • runner integration for controlled figure-generation commands
  • frontend handling for non-text artifacts

3. Finish the Review Agent

What exists now:

  • delegation hook in the Main Agent
  • placeholder ReviewAgent class

What is missing:

  • structured review prompts and rubric outputs
  • section-by-section issue reporting
  • links between review findings and actionable edits or citations

Likely implementation areas:

  • services/researchkit/researchkit/agents/review_agent.py
  • UI support for review summaries and issue navigation

4. Strengthen Memory and Draft Understanding

What exists now:

  • section structure
  • abstract
  • venue hint
  • citations
  • summary generation

What is missing:

  • better extraction of research questions and contributions
  • stronger handling of multi-file paper layouts
  • richer style and terminology profiling

Likely implementation areas:

  • services/researchkit/researchkit/memory/

5. Improve Workspace and Runner Ergonomics

What exists now:

  • guarded editor operations
  • runner-backed command execution
  • overlay diff capture

What is missing:

  • clearer workspace setup guidance in the product
  • better handling of nested project directories
  • broader command-result UX for execution-heavy flows

Likely implementation areas:

  • services/researchkit/researchkit/agents/main_agent.py
  • services/researchkit/researchkit/runner/main.py
  • provider settings UI and editor-side action rendering

Product Direction Defaults

  • Keep the Overleaf module thin and let the Python service own agent logic.
  • Reuse the existing runner plus patch pipeline instead of introducing a separate execution path.
  • Treat the Research Agent as the first fully operational specialized agent and use it as the pattern for Figure and Review.