Interfaces & Data

Interfaces

Interface	Type
Summarizer.py	CLI entrypoint
config.toml / config.example.toml	Configuration
ProjectIndex.md	Inclusion authority (the only source of truth for what gets evaluated)
<folder>.manifest.toml discovery	Per-project manifest lookup
Ollama / OpenAI-compatible endpoints	Model backend (via ModelClient.py)
Dashboard/Project-Summaries/raw/	JSON output directory
Dashboard/Dashboard.py	Flask dashboard over the compiled output

Produces

• Per-project JSON summaries (raw, one file per project)
• Compiled index.json — the same shape consumed by this documentation site
• summary.md — compiled Markdown portfolio summary
• Flask-renderable dashboard HTML
• History and raw archives under Project-Summaries

Consumes

• ProjectIndex.md
• Project files and manifests, sampled per project within configured budgets
• config.toml and an environment-held OpenAI API key
• Governing standards files (CTS, DRS, WDS)
• LLM service endpoints and credentials

Depends on

• Python 3.11+ (or a tomli fallback)
• requests, flask
• concurrent.futures threadpool for incremental concurrent processing
• Ollama or OpenAI-compatible API endpoints
• Local filesystem access to every indexed project path

Used by

• The operator (Herb) directly
• Portfolio maintainers and standards reviewers
• The local Flask dashboard
• This documentation site — every CTS/DRS/WDS project page's summary, capabilities, interfaces, and roadmap data traces back to a run of this tool

Notable governance detail

The manifest marks this project stability = experimental and active_development = true, distinct from a tool like CloneCratesio which is paused-but-stable. A known, intentional gap: full-portfolio API evaluation is excluded from local test verification, since it would spend real API credits on every test run.