Spark ERP

AI · Knowledge

Ask your documents, not a guess.

Add policies, manuals and web pages. Spark ERP cuts them into passages, embeds them with any embedding model, and searches by meaning and by exact words, so an assistant answers from what your organisation actually wrote.

Knowledge bases and embeddings in Spark ERP Your ERP dataModel providerPermissions Your ERP dataModel providerPermissions Audit trailAsk or trigger Human approvalProposes
Knowledge bases and embeddings and its guardrails
Your ERP dataModel providerPermissionsAudit trailHuman approvalKnowledge bases…
How Knowledge bases and embeddings works with you
Ask or trigger

A user request or event

1
Grounded

Uses your own records

2
Proposes

A draft, not a silent change

3
You approve

Within your permissions

4
Logged

Recorded for audit

5

What you can do

Add documents

Paste text, upload text files or add a web page.

Any embedding model

Vectors of any size work, and each search only compares vectors from the same model.

Meaning and words

Search by meaning, by exact words (useful for codes and names), or both together.

Tune it

Set passage size and overlap with a live preview, and the default search mode and minimum similarity.

Test a question

A retrieval test shows exactly what an agent or workflow would get, with scores.

Yours alone

Documents and vectors are stored in your own tenant database.

The real screens

A knowledge base with three documents and the number of passages each made.
Documents, each cut into passages.
A retrieval test where a question with no shared words finds the leave policy first.
A question with no shared words still finds the leave policy.

What to know today

Text, text files and web pages are supported. PDF and Word files, crawling a whole site and scheduled re-sync are not available yet, and access is set per knowledge base rather than per passage.

Try it on your own policies.

Bring a few documents and the questions people ask most.