eyeszik/unbound-innovation-engine

How eyeszik/unbound-innovation-engine works

UNBOUND is a persistent, multi-stage innovation application: an owner submits a brief, a worker runs creative generation through evidence review, evaluation, blueprinting, and validation, then the dashboard exposes progress and exports results. The graph preserves the documented phases and optional integrations; several engine-to-stage calls and UI request details were not excerpted, so those links are shown only where the README or sampled source supports them.

Main entry points: engine.py Phase engine · api.py HTTP API · worker.py Background worker · app.js Run dashboard

Components 23

User experience

Run dashboard unbound/static/app.js
HTTP API unbound/api.py
Report export unbound/export.py

Run orchestration

Background worker unbound/worker.py
Phase engine unbound/engine.py
Run contracts unbound/models.py
Runtime settings unbound/settings.py

Innovation stages

Model providers unbound/providers.py
Intent classification unbound/engine.py
Idea generation unbound/engine.py
Wildcard seeds unbound/engine.py
Recursive synthesis unbound/engine.py
Novelty review unbound/engine.py
Candidate evaluation unbound/engine.py
Experiment blueprint unbound/engine.py
Output validation unbound/validation.py

Evidence and persistence

Evidence review unbound/research.py
Computation queries unbound/research.py
Run store unbound/store.py
Prompt rules unbound/rules.py

Other components

Run owner
Exa search
Wolfram Alpha

Connections 27

Run owner → Run dashboard: submits brief README.md
Run dashboard → HTTP API: starts and polls README.md
HTTP API → Run store: creates and reads runs unbound/api.py
HTTP API → Report export: renders Markdown unbound/api.py
Background worker → Run store: claims and commits unbound/worker.py
Background worker → Phase engine: invokes graph unbound/worker.py
Background worker → Model providers: constructs provider unbound/worker.py
Background worker → Runtime settings: loads settings unbound/worker.py
Phase engine → Run store: reads and commits phases unbound/engine.py
Phase engine → Model providers: requests model output unbound/engine.py
Phase engine → Run contracts: validates phase outputs unbound/engine.py
Phase engine → Intent classification: classifies intent README.md
Phase engine → Idea generation: generates ideas README.md
Phase engine → Wildcard seeds: extracts seeds README.md
Phase engine → Recursive synthesis: builds hybrids README.md
Phase engine → Novelty review: reviews novelty README.md
Phase engine → Evidence review: requests research unbound/engine.py
Phase engine → Computation queries: requests computations unbound/engine.py
Phase engine → Candidate evaluation: evaluates candidates README.md
Phase engine → Experiment blueprint: defines experiment README.md
Phase engine → Output validation: validates completed run unbound/validation.py
Phase engine → Prompt rules: uses prompt rules unbound/engine.py
Evidence review → Exa search: searches evidence unbound/research.py
Computation queries → Wolfram Alpha: queries computation unbound/research.py
HTTP API → Runtime settings: checks provider configuration unbound/api.py
HTTP API → Run dashboard: serves interface unbound/api.py
Run dashboard → Report export: downloads reports

Drawn by GitDiagram on November 13, 2026 from the repository's file tree, README and source files. Read as Markdown ·