Catalog, dictionary, column-level lineage and multi-modal Knowledge Base — generated by generative AI from your databases, datalakes, pipelines and code. In 3 layers (technical, semantic and business), always up to date, queryable in natural language.
Engineers create tables, views, notebooks and pipelines fast. Cataloging metadata, lineage, business rules and data dictionary is manual, repetitive and historically abandoned. The cost: slow onboarding, expensive audits, risky refactors, and business decisions on data of unclear origin.
Each object and field gets a technical (DBA), semantic (analyst) and business (controller) description — independently generated by AI.
Parser covers 25 SQL dialects, Python/PySpark AST and notebooks. Reconstructs each column's origin back to the primary source — no client-declared metadata needed.
Upload PDF, DOCX, spreadsheets, video, audio, images, free-text notes — the AI absorbs, indexes and enriches the documentation with your vocabulary.
Ask in Portuguese: "where is the CPF?", "what's the difference between gross and net revenue?". Answers come with lineage and code reference.
Automatic detection of CPF, CNPJ, RG, email, card and phone across all sources — with lineage to the final derivation and responsible owners.
AI produces the first version; owners approve, edit or reject in a Kanban queue. The AI learns from each decision and adapts to internal vocabulary.
Tables, views, procedures, functions, triggers, packages, indexes, jobs, dashboards, ML models, dataflows — everything the source exposes.
Alerts when schema changes, null % rises, distribution shifts or new PII appears. Full versioning: see how table X looked 6 months ago.
Dedicated Kubernetes namespace per tenant. Read-only connection. Shared, dedicated or self-hosted LLM (Enterprise). LGPD / ISO 27001 compliant.
Terms, KPIs and rules with AI auto-linking: each concept is tied to the right fields by semantic similarity and feeds the semantic and business layers. Finance stops asking for technical explanations.
Publish the catalog as PDF, Excel, Word, HTML, CSV or Markdown/Confluence — with granular selection of what to include, scope by source or project, and scheduled export (weekly or monthly).
SSO/SAML and OIDC (Azure AD, Okta, Google), role-based access (admin, editor, owner, viewer, auditor), a 7-year immutable audit trail and a credential vault with BYOK. Read-only access to your source.
Each layer is editable independently. The engineer adjusts the technical layer without affecting the business one. Finance rewrites the business layer without touching the technical. Full history of who edited what, when and why.
Type, nullability, inferred keys, null %, cardinality, range, top values, distribution, implicit FKs, procedure dependencies.
For: data engineers, DBAs, architects.What the field represents in the domain: "product tax code", "invoice issue date", "unique B2B customer ID". Mapped to the central glossary by synonyms.
For: analysts, data scientists, new joiners.How the field is used in decisions: KPI it feeds, calculation rule, inferred owner, dashboards and reports that depend on it, consultation frequency.
For: business users, controllership, audit.Embedded in the portal, Catalog Chat lets any user — technical or business — talk to the catalog in natural language. Reuses the NL→SQL engine of WFS DataTalk AI, now pointed at the catalog graph. Supports questions on schema, lineage, business rules, change impact and governance.
Coverage per source goes beyond tables: views, materialized views, procedures, functions, triggers, sequences, packages, jobs, ML models, dashboards and dataflows. Additional connectors on demand in 2-4 weeks.
Generate the documentation in the format each team prefers — ready to print, edit, version or embed in the intranet. Granular selection of what goes in (objects, fields, docs, glossary, lineage, KPIs, SQL, quality), scope by tenant, project, source or schema, and automatic delivery.
Output formats
Delivery and scheduling
From 3 weeks studying the lakehouse to a few hours in the portal — full view of what exists, what it does and how everything connects.
"Where is the CPF?" answers in 1 click, with full list of fields, sources, lineage and derived tables. Defensible compliance.
Before changing a table, column-level lineage shows all pipelines, dashboards and reports depending on it — including the specific Power BI measure.
Business areas stop asking for technical explanations. Each KPI definition is visible in Catalog Chat with origin and rule derived from code.
In TOTVS→Databricks or on-prem→Cloud projects, the catalog serves as the source of truth of the current state and generates the future-state backlog with preserved lineage.
The PII Map continuously identifies where personal data is stored and replicated. Alerts when a new PII appears.
| WFS Smart Document AI | Atlan / Collibra / Alation | |
|---|---|---|
| Deployment time | Hours (plug-and-play) | Weeks to months |
| Automatic documentation | 3 AI-generated layers | 1 free-form field, manual |
| Column-level lineage | From code parser (SQL, Python, notebooks) | Depends on declared metadata |
| Catalog Chat | Native PT-BR, talks about lineage & rules | Keyword search |
| Knowledge Base | Multi-modal: PDF, DOCX, video, audio, images | Free-text field |
| BR ERP connectors | TOTVS Protheus, Datasul, Senior, Oracle EBS | Non-existent or via generic JDBC |
| Self-hosted LLM | Available in Enterprise | Rare or non-existent |
| Support | WFS in Brazil, in PT-BR, BR time zone | Global, rarely in PT |
Multi-tenant architecture with real isolation, read-only access and an immutable audit trail — designed to pass corporate security review.
Connections open in read-only mode (driver flag or BEGIN READ ONLY transaction). The platform reads metadata and code — it never writes to your source.
Credentials never live in the application database — only a reference to the secret in the vault (Azure Key Vault managed by WFS, or your own KMS on Enterprise). Automatic rotation.
Corporate login (Azure AD, Okta, Google Workspace) and roles by function: admin, editor, data owner, viewer and auditor. Magic link on first access, no password.
Every view, edit, approval and export is logged with user, IP and timestamp — in an immutable sink (Object Lock) for 7 years, for LGPD compliance.
Dedicated Kubernetes namespace and an isolated database per client, with row-level security — defense in depth against any cross-tenant leak.
For closed networks, a container runs in your environment and pulls jobs over outbound HTTPS with mTLS — no inbound ports, no VPN required.
LGPD · ISO 27001 · SOC 2 compliant
R$ 790/mês
or US$ 159/mo outside Brazil · 15-day free trial
3 connectors · 5 users · up to 25k objects · 1,300 WFS AI Credits/mo · Catalog Chat · column-level lineage. Ideal for initial validation.
Subscribe now — 15-day free trialR$ 2.190/mês
or US$ 449/mo outside Brazil · 15-day free trial
10 connectors · 20 users · up to 150k objects · 4,500 WFS AI Credits/mo · REST API · code lineage (views/procedures/ETL) · glossary and editorial queue. For consolidated data teams.
Subscribe now — 15-day free trialR$ 4.900/mês
or US$ 990/mo outside Brazil · 15-day free trial
Unlimited sources · 30 users · up to 500k objects · 13,500 WFS AI Credits/mo · multi-project governance · assisted onboarding and priority support. For enterprise groups.
Subscribe now — 15-day free trialR$ 9.800/mês
or US$ 1,980/mo outside Brazil · double the Scale plan
1 million objects · 60 users · 27,000 WFS AI Credits/mo · multi-company (up to 10) · self-hosted LLM · SSO/SAML · white-label on your domain · 99.9% SLA · 24×7 support. For large enterprise groups.
Talk to salesYou only pay for the AI you consume: every documentation, embedding and chat answer debits WFS AI Credits. Cost scales by documentation generated, not data volume — a database with millions of rows and 200 tables costs the same as 200 empty tables. One-off credit packs available to document the entire catalog at once. Technical setup billed separately as a dedicated deliverable.
45-min executive demo over a sample database, or 2-week Proof of Concept on a subset of your real data, at no cost.
Schedule demoWFS Smart Document IA is a platform for automatic catalog and documentation of databases, data lakes and pipelines. It connects to the customer's infrastructure, scans every object (tables, views, procedures, notebooks, dashboards) and produces documentation in 3 layers (technical, semantic and business) using generative AI, with column-level lineage rebuilt from real code.
After registering sources (DBs, data lakes, Git repos, BI tools) with read-only credentials, the product scans every object, computes statistics and data profiles, auto-detects PII, and processes each artifact with a corporate LLM (the WFS SD IA — collaborative across 5 of the best AIs on the market) producing descriptions in the 3 layers. Everything is browsable in a web portal with Catalog Chat in PT-BR.
Databases: Oracle, SQL Server, PostgreSQL, MySQL, DB2, Sybase. Lakehouses: Databricks, Microsoft Fabric, Snowflake, BigQuery, Synapse, Redshift. Storage: ADLS Gen2, S3, GCS, Delta Lake, Iceberg. Orchestrators: ADF, Airflow, dbt, Fabric Pipelines. NoSQL: MongoDB, Cassandra, CosmosDB. BI: Power BI, Tableau, Looker. Brazilian ERPs: TOTVS Protheus, Datasul, Senior, SAP. Additional connectors built on demand in 2-4 weeks.
Column-level lineage shows the exact origin of every column down to the primary source, traversing all transformations. WFS Smart Document IA rebuilds this by analyzing real code (SQL, Python, Databricks notebooks, dbt), without relying on manual declaration. Useful for safe refactoring (know what will break before touching), LGPD audit (where is the CPF?) and governance.
Four plans: Starter (R$790/mo, 3 connectors, 25k objects, 5 users, 1,300 credits), Professional (R$2,190/mo, 10 connectors, 150k objects, 20 users, 4,500 credits, REST API and code lineage), Scale (R$4,900/mo, unlimited sources, 500k objects, 30 users, 13,500 credits) and Enterprise (R$9,800/mo — double the Scale plan: 1 million objects, 27,000 credits, 60 users, multi-company, self-hosted LLM, SSO/SAML, white-label, 99.9% SLA, 24×7 support). Billed in BRL (or USD outside Brazil) per WFS AI Credits consumed, not per seat. Technical setup billed separately.
Yes, on the Enterprise plan. Self-hosted LLM models available via private AWS Bedrock, dedicated Azure AI or dedicated models on-prem. Sensitive data never leaves the customer environment — only metadata and code flow between the connector and WFS's dedicated tenant. Compliant with LGPD, ISO 27001 and SOC 2.