Upload a product photograph. The pipeline identifies it, extracts its attributes, writes the copy in every language you sell in, generates the imagery, and indexes it for search — then stops and waits for a person to sign it off.
Each capability is a separate model call with its own cost, its own log entry and its own place in the review queue — never a single opaque black box.
Category and brand identification, product attribute extraction, OCR from packaging and labels, barcode extraction for UPC, EAN and ISBN, plus object detection and tagging.
Background removal with automatic provider fallback, multi-angle and novel view generation, 3D assets in GLB, GLTF, USDZ and OBJ for AR, object segmentation, and moderation on every generated image.
Titles, short and long descriptions, feature bullets, highlights, specifications, meta titles, meta descriptions and SEO keywords — delivered as a structured field set, not a wall of text.
Content is generated per language rather than run through a translator, so the output reads natively in each market instead of reading translated.
Semantic search finds products by meaning rather than keyword match. Visual search finds them from an image. Duplicate detection works across both text and imagery.
Product video in multiple formats, built from your catalogue imagery and sized for marketplace listings and social channels.
A photograph, a set of images, or a bulk file of SKUs. That is the whole input — no datasheet to prepare, no template to fill in first.
The pipeline identifies the product, extracts attributes, writes the copy, generates the imagery and builds the search index as a single orchestrated run.
Your team reviews the draft, approves it, and exports it or hands it to your publishing workflow. Nothing reaches a customer that a person has not signed off.
Image and copy generation is table stakes. What a technical evaluation actually tests is whether you can prove who approved a listing, which model produced it, and what stopped it costing more than you agreed.
Every SKU moves through draft, review, approved and published — and the person who approves cannot be the person who created it. Enforced by the system, not by policy. Eleven distinct roles, from catalogue manager to SEO manager to publisher.
Each model call is logged: which model, which version, which prompt version, how long it took, what it cost. When a listing is challenged, you can show exactly how it was produced.
Per-organisation credit budgets, per-service spending caps and daily request quotas resolving across platform, organisation and user level. Budget checks run before the expensive calls, not after the invoice.
Multi-tenancy is PostgreSQL row-level security, not application filtering. A missing WHERE clause cannot leak another tenant's catalogue, because the database refuses the read.
Choose the provider and model for each capability, per organisation. Credentials are encrypted at rest with AES-256-GCM. Your data goes to the provider you selected, not wherever the vendor happens to hold a contract.
Multi-factor authentication, configurable password policy, and GDPR Article 17 erasure that anonymises personal data while preserving the audit record and referential integrity.
Bulk jobs return a cost estimate and wait for confirmation before a rupee is spent. Every call below is captured with its model and prompt version. The run stops at review — publishing is a human action.
Illustrative run record showing the fields captured for every enrichment. Credit values are units of account, not currency.
Registered partners receive an authenticated, read-only feed of approved products — attributes, content and time-limited image URLs — filtered by SKU or by change date for incremental sync.
Draft and archived products are never exposed.
Talk to us about API access →Marketplace compliance rules are evaluated before publishing, with pass, warn and fail outcomes and per-marketplace category mapping — so listings are checked against channel requirements while they are still drafts.
Build and own your own catalogue instead of renting a listing team to maintain it.
Onboard sellers at volume without the catalogue team scaling in step with them.
Bring large SKU counts online without a linear increase in headcount.
Produce catalogues for multiple clients with tenant isolation between every account.
Bring producer groups and cooperative sellers online under one governed catalogue.
HSN, GST and barcode fields are structured and validated before publishing, so classification errors surface as drafts rather than as listings.
Everything above this line is running today. Everything below is designed and in build. We keep the line visible because you will test it anyway.
360° spin video, product video, PDF catalogues and manufacturer datasheets as inputs alongside photographs.
Lifestyle backgrounds and marketplace-style infographic panels composed from the product image set.
Every extracted fact labelled as observed, read from a label, or inferred — with per-field confirmation before publishing.
A readiness score per marketplace with specific, actionable fixes rather than a bare number.
Direct publishing to Amazon, Flipkart, Meesho and ONDC. Today the workflow, compliance checks and category mapping are built and output is exportable to any channel.
Listen to your generated listing read aloud in your own language, and edit it by speaking. Built for sellers with limited text literacy, including self-help group members and first-time online sellers.
Not today. The publishing workflow, marketplace compliance checks and per-marketplace category mapping are built, and output is exportable to any channel. Direct publishing to live marketplace channels is on the roadmap, and we would rather tell you that now than at your technical evaluation.
Yes. Provider and model are selected per capability, per organisation. Credentials are encrypted at rest with AES-256-GCM, and your data goes to the provider you selected rather than wherever a vendor happens to hold a contract.
Per-organisation credit budgets, per-service spending caps and daily request quotas resolve across platform, organisation and user level. Budget checks run before the expensive calls, and bulk jobs return a cost estimate and wait for your confirmation before spending anything.
Every module runs standalone and every module talks to the kernel. These are the ones most often deployed alongside it.
One source of truth for product data, published to every channel.
Open page → Platform · STRA fast, search-friendly storefront with merchandising, checkout and payments end to end.
Open page → Platform · SUPGive suppliers their own workspace for listings, performance and settlements.
Open page →A working walkthrough with your catalogue, your order flow and your questions. No slideware.
