Open-source product engineering / Image workflow tooling
Rendriva.
An installable ChatGPT Agent Skill plus deterministic OpenAI Image API batch adapter that turns a creative request into a plan, generation, inspection, repair, and delivery workflow for separate image outputs.
Explore public source ↗01 / THE PROBLEM
Make the process visible.
Batch image requests can collapse into inconsistent outputs, accidental collages, swapped variants, weak product fidelity, or repeated compositions. Rendriva treats the batch as a production system with explicit output cardinality, reference roles, repair logic, and reports.
02 / THE APPROACH
Turn requirements
into useful work.
The workflow plans each requested image independently, protects product and model reference constraints, supports draft selection and targeted repair, and produces separate final files with reusable quality, brand, fidelity, diversity, and campaign reports.
- Agent Skill product design
- Batch workflow architecture
- Reference-fidelity and quality gates
- Deterministic Python adapter
- Platform export and reporting design
03 / SCOPE & EVIDENCE
What you can inspect.
These notes describe an internal product-engineering project, not a client engagement. Scope and technical descriptions come from the existing WebTayo case study; no performance or commercial outcome is claimed.
Review the linked repository for current implementation details, release history, tests, and documented limitations.
Open source repository ↗