Guides

Your data is scattered across files and business systems, while you want AI that understands it. A clear path connects the two, and these guides follow that path.

Bring data in. Upload files or connect a business database: Data ingestion.

Incoming data needs an account. Where it is, what it is, and who uses it: without registration, growing data has no order. The Data Catalog registers it and enables later tracing. Any answer can be traced through the records to its source file.

Registered files are still unreadable to AI. PDF, Word, audio, and video are formats for people. Parsing, chunking, and vectorization turn them into retrievable data: Data processing, performed by workflows.

Processed data needs one more step before AI can use it. Define its scope and explain its usage so AI knows what and how to use: a knowledge base. Create, query, validate, and troubleshoot it on that page.

At this point AI can answer questions. To go further—perform tasks and write results to business systems—use agents. Agent reports and tables are new data and can enter this path again. The more you use it, the more data AI can use.

Q&A, processing, and agents run on models. The model service centrally integrates and routes language, embedding, reranking, and other models for use across the platform.

Stage

Your data

AI participation

Data ingestion

Files and database tables enter the platform.

You initiate ingestion.

Data Catalog

Registered, searchable, and traceable.

AI locates data through the Catalog before retrieval.

Data processing

Parsed, chunked, and vectorized.

Built-in AI organizes natural-language needs into processing flows.

Knowledge bases

Scoped and explained; retrieved and cited.

AI retrieves and answers within scope; answers are traceable.

Agents

Outputs become new data.

Reads data, runs tasks, and writes results.
Last updated on