Ingestion
Multi-format input
Imports Excel with a column-mapping wizard and diff, or reads PDF and photo quotes with computer vision — the AI "sees" the scanned list and materializes the items.
Smart quoting with AI: from the client’s list to a finished quote, in minutes, not hours.
SolicitoWeb takes the list of items your client sends — a tender, a quote request, an order in PDF, Excel or photo — and builds the quote by matching each item against your internal catalog: it answers what you have to offer, at what price and what is missing. It is LNData’s AI-powered smart quoting platform, ideal for large item lists and public tenders.
Visit the site solicitoweb.com.br
Role · Proprietary product · architecture, full-stack engineering, applied AI and infrastructure
Status · Multi-tenant SaaS · in production · v1.26
In short, for decision-makers
The problem
Distributors, resellers and manufacturers receive orders and quotes every day in different formats and inconsistent naming — and a tender or a large order can have hundreds of lines. Matching each item against the catalog by hand is slow (hours per quote), depends on the senior salesperson who "knows the catalog by heart" and makes mistakes — the wrong gauge, the wrong connection gender, the accessory instead of the part.
What LNData did
We built SolicitoWeb, a SaaS platform that imports the client’s list, understands each item with AI and suggests the matching product from the catalog — with price, source priority and stock alerts — leaving the final call to a fast human curation step. The same engine serves electrical, plumbing, hardware and other sectors.
The result
Quotes that took hours come out in minutes, without depending on whoever knows the catalog by heart. The work becomes a reproducible pipeline that scales to catalogs of 5,000 to 50,000 items with daily updates — and a new sector comes on board without writing a single line of code.
SaaS
multi-tenant platform
Agnostic
of sector (data, not code)
Applied AI
semantic search + vision
In production
solicitoweb.com.br
minutes
to quote lists with hundreds of items
5k–50k
items per catalog, with daily sync
PDF · Excel · photo
the client’s list in any format
AI + curation
automatic suggestion, your decision
Architecture
The platform receives the external list through several paths, enriches and indexes each item in an asynchronous AI pipeline, and returns a finished quote — all orchestrated in containers, with the expensive tasks running in workers with failure recovery.
Input
- Excel · column wizard
- PDF / photo · computer vision
- Own catalog
- ERP integration · sync
AI engine
Semantic expansion (LLM)
Embeddings + Qdrant
Hybrid search + re-ranking
Async · Celery workers
Output
- Assisted curation
- Excel quote
- Branded PDF · discounts
What the platform does
An end-to-end flow that turns a raw list into a quote: import → enrich with AI → index → search → curate → export.
Enrichment
Semantic expansion
Each product and item gets an LLM-enriched technical description, with a prompt customizable per company, resolving abbreviations and standardizing vocabulary before indexing.
Match
Search with business rules
Semantic match with source priority (own product × supplier), preferred-brand filter, top-K suggestions per item and stock alerts.
Curation
Human decision at the end
The operator accepts, rejects or swaps each item↔product pair, with automatic acceptance by confidence threshold and manual pair addition.
Export
A quote with your brand
Export to Excel and to PDF with the client’s branding — logo, field selection, per-item discounts and total.
Integration
Catalog and prices from the ERP
An integrator installed at the client reads the ERP and runs a periodic full-sync of catalog and price tables over a REST API, with idempotency, auditing and developer documentation.
The intelligence that makes the difference
What sets SolicitoWeb apart from an ordinary search is the accuracy engine — the layer that stops the AI from confusing "1/2" with "1/4" or "female" with "male".
Hybrid search (dense + sparse)
Combines semantic similarity with lexical matching (RRF fusion) to beat the weakness of embeddings with numbers, fractions and negations — "3/4 valve" now suggests the right valve, not the lever.
Discriminating attributes per category
A schema per category (gauge, gender, thread; voltage, poles, current) is extracted by an LLM and used as a penalty in re-ranking — the candidate with the wrong gauge is demoted.
AI schema discovery
The AI itself samples the category and proposes the relevant attributes; a human curates. A new sector comes on board without writing a single line of code.
Beyond the core
Simple Mode, multi-tenancy and security
The power of the advanced engine wrapped in a one-step flow for the end salesperson — and the platform infrastructure that sustains many companies at once, with security taken seriously.
Simple Mode
A guided flow that orchestrates the whole pipeline — extraction, materialization, expansion, vectorization, attributes and match — from a single photo or spreadsheet, with a visual progress stepper.
True multi-tenancy
A user can belong to several companies with a different role in each; the server-side session controls the active company and lets you revoke remote access.
AI provider per company
Each company uses its own AI provider, with the key encrypted at rest, a daily request limit and pipeline blocking when no provider is configured.
Security taken seriously
Validated tenant isolation (with IDOR vulnerabilities fixed in code review), encryption at rest and revocable sessions across the whole platform.
Vue 3FlaskCeleryQdrant
Impact
The work that required a senior salesperson reading item by item became a reproducible pipeline that delivers in minutes.
Suggestions arrive already with the right product from your catalog — prioritizing your own products, preferred brands and what is in stock, so the salesperson just reviews and closes.
The same engine serves different sectors without writing code — a new sector comes on board with AI-assisted schema discovery, no deploy.
Multi-tenant in production, each company with its own AI provider, plans and usage limits.
Platform stack
PythonFlaskCeleryPostgreSQLQdrantOpenAIVue 3Docker
Want to see SolicitoWeb in action?
If your company matches order lists against a large catalog, SolicitoWeb solves it. Explore the platform or talk to LNData for a demo.