​
AutoFigure & AI-Researcher

BioCortex optionally integrates AutoFigure (methodology/result figures from text or papers) and AI-Researcher (autonomous research pipeline). Both are registered as DomainTools and appear in the tool registry and on the Resources page. The core framework runs without them if not installed.

​
AutoFigure

​
What it does

  • Generate figures from text — User describes a methodology or result; AutoFigure produces a figure.
  • Generate from a paper — Extract methodology from a research paper (PDF or Markdown) and generate a methodology diagram.

​
Integration

  • Module: biocortex/tools/autofigure_tool.py
  • Tools (examples): autofigure_generate_from_text, autofigure_generate_from_paper, etc.
  • Domain: e.g. protocols, literature
  • API key: Configurable via AUTOFIGURE_API_KEY or fallback to BioCortex .env (e.g. BIOCORTEX_CUSTOM_API_KEY).

​
Registration

Registered in biocortex/domains/loader.py behind try/import. If AutoFigure is not installed, the loader skips it and the rest of the framework is unchanged.

​
Dependencies

Optional AutoFigure-related dependencies are listed in requirements.txt (commented or in an optional section). Install them if you want figure-generation tools.

​
AI-Researcher

​
What it does

  • Reference-based mode — Given a list of papers, AI-Researcher generates a novel idea and implements it (literature → idea → implementation).
  • Detailed-idea mode — User provides a full research idea; AI-Researcher runs literature review, implementation, and (optionally) validation/paper writing.
It runs as an external pipeline (subprocess or Docker), with its own env (e.g. OPENROUTER_API_KEY, COMPLETION_MODEL, CATEGORY, INSTANCE_ID).

​
Integration

  • Module: biocortex/tools/ai_researcher_tool.py
  • Tools: ai_researcher_run_detailed_idea, ai_researcher_run_reference_based
  • Domain: literature
  • Invocation: Subprocess (or HTTP if you expose an API) to the AI-Researcher repo/script. Requires AI_RESEARCHER_PATH (repo root). Uses the pipeline’s own .env for models and keys.

​
Registration

Registered in loader.py behind try/except. If the module is missing or AI_RESEARCHER_PATH is not set, the tools are not loaded; when called, they return a clear error.

​
Caveats

  • Long-running: The bridge should return quickly (e.g. job id or output path); the actual run may be consumed later or the Executor timeout increased for that step.
  • Env: Reuse BioCortex’s API key in AI-Researcher’s .env if desired; set the rest (model, category, etc.) in the pipeline’s config.

​
Adding Your Own Integrations

Same pattern:
  1. Implement a module under biocortex/tools/ with get_*_tools() returning list[DomainTool].
  2. Register in biocortex/domains/loader.py with try/except so the core runs without the optional package.
  3. Document env vars and optional dependencies.
See Adding Tools and Agents for the full pattern.