작동하는 배합 도구.
개발 중인 AI 워크스페이스.
PISTORA connects ingredient information to recipe development for small food producers. This release is a public, browser-based formulation prototype. It is not a deployed multi-tenant production system.
피스토라는 소규모 식품 제조사의 원료 정보와 배합 작업을 연결합니다. 현재 공개 버전에서 실행 가능한 기능과 향후 개발할 기능을 구분합니다.
Open workspace / 배합 도구 열기 ↗Available in this release / 현재 구현
| 기능 / Feature | 범위 / Scope |
|---|---|
| 21 editable starting recipes | 9 gelato, 7 fruit/coconut formulations, 5 alcohol formulations. Illustrative, not production-certified. |
| Deterministic calculations | Mass scaling, fat, solids, total sugars, MSNF, POD, PAC, ethanol mass fraction and ingredient cost. |
| Ingredient editor | 32 reference records. Source notes, composition, alcohol ABV/density and review status. Add your own ingredients. |
| Recipe workflow | Browser-local save/load, JSON backup/import, CSV export, print-to-PDF and baseline comparison. |
| Claude text extraction | Optional browser-to-API requests using your key. Sonnet 5.5 or Haiku 5.5. Exact quote-presence check; explicit review before creating a separate ingredient record. API calls require a valid key, model access and network connectivity. |
Planned / 개발 예정
- PDF and scanned-document ingestion with page citations.
- Measured evaluation on supplier documents and calibrated uncertainty handling.
- Persistent, versioned human approvals with user identities.
- Task routing, prompt caching and Message Batches integration.
- Source-linked product copy with validation against approved records.
- Multi-user workspaces and server-managed API credentials.
The optional extraction feature uses structured tool calls. A schema constrains format; it does not guarantee factual accuracy. Human review remains necessary. Example output is explicitly hand-authored and does not represent a measured API result.
Private beta target / 비공개 베타 목표
We aim to recruit five independent Korean gelato and dessert producers for a Q1 2027 private beta. This is a recruitment goal, not a claim of five signed customers. Our next objective is to measure time saved and extraction reliability in real workflows.