Camera-first batch capture · all visible records are mock data from a German-configured demo instance
Scope
What this area changes in the data flow.
Open Inventory separates the short interaction at the object from slower server-side processing. Type, location and optional processing settings can be set once for a batch. Resource creation, uploads, analysis and cover generation then run as separate jobs with their own states.
Result 01
Capture several objects without waiting for every upload and processing step
Result 02
Review and correct generated fields before treating them as inventory data
Result 03
Keep original media and structured fields on the same record
Included implementation
Individual capabilities and constraints.
These components are part of the MIT-licensed project. Optional providers are identified where they take part in a workflow.
01
Camera-first batch capture
Capture several objects in sequence. Persistent jobs continue processing earlier records while the camera is available for the next object.
02
Image-derived draft fields
Optional OpenAI or compatible image analysis proposes a name, description, type, tags, alt text and confidence value. The draft is editable before it is accepted.
03
Generated square covers
OpenAI or Google can generate a square cover from a source photo. The generated asset is stored separately and the original photo remains available.
04
Count similar parts from a photo
Optional photo counting marks detected instances and returns a confidence value. The quantity can be corrected before a stock movement is written.
05
Media attached to the record
JPG, PNG, WebP, AVIF, HEIC, MP4, MOV, WebM, PDF and USDZ can be ordered on a record. Images are optimized client-side and EXIF coordinates can be imported.
06
Native iPhone capture
The open SwiftUI app captures photos, scans codes and persists upload stages. Stable operation IDs prevent defined retries from creating duplicate records.
Sequence
A staged queue instead of one long request
Shared context is set before the series. Each object then moves through explicit stages that can fail or be retried independently.
01
Set batch context
Choose the type, location and optional analysis or cover settings once for the series.
02
Take the photos
Capture a primary image and, when useful, additional views or a separate nameplate photo.
03
Process in stages
Resource creation, media uploads, analysis and cover generation expose separate job states. Defined retries use stable idempotency keys.
04
Review the result
Check name, type, tags, quantity, tracking mode and media before relying on the record.
Worked example
Concrete records, states and decisions.
The values are mock data, but the workflow and state changes correspond to operations implemented by Open Inventory.
concrete example
Example · Workshop shelf
Twelve objects with one shared location
The batch sets Tool and Shelf B2 once. Each object then needs its own photos; serial number, condition and other exceptional fields are added only where required.
Shared context
Tool · Shelf B2
Captures
12 objects
Final action
Review suggestions
Real SwiftUI app capture · demo data
Open-source iOS implementation
Capture at the shelf with the native client.
The SwiftUI app takes photos at the object and persists each operation in an Application Support outbox before upload. It connects to an Open Inventory instance selected by the operator.
Take or select up to twelve photos for one record
Resume defined upload stages after a network interruption or restart
Scan a code, open the matching record and verify it on site
Image analysis and cover generation are optional provider-backed operations. The repository shows when data leaves the instance. Original images, provider configuration and the final stored values remain under operator control.
MIT-licensed web application and native iOS client in one repository
Optional external processing rather than a requirement for basic capture
CSV export and a documented REST API for independent workflows
