CraftMan MarketPlace · AI Product
AI Prompt: Chat with Your Data
A Retrieval-Augmented Generation (RAG) chatbot platform: ask questions in plain language, get grounded answers cited from your documents and databases — not the open internet.
How it works
Step 1
Ingest
Load documents & tables
Step 2
Embed
Vectorise into semantic index
Step 3
Retrieve
Hybrid: vectors + keywords
Step 4
Ground
Only retrieved facts allowed
Step 5
Answer
Cited, plain-language reply
The same engine powers very different domains — the two flagship builds below.
Flagship Build 1 · Bible-v2 — Scripture Research Assistant
A study companion that answers questions about scripture with verse-cited responses. Hybrid retrieval (Elasticsearch semantic search + PostgreSQL canonical text) finds the right passages; the model may only answer from what was retrieved.
Where does the Bible talk about faith the size of a mustard seed?
Jesus uses the mustard-seed image for faith in two places: Matthew 17:20 — “if you have faith like a grain of mustard seed, you will say to this mountain, ‘Move…’” — and Luke 17:6, where even small faith can uproot a mulberry tree.
Sources: Matthew 17:20 · Luke 17:6 (retrieved passages)
- Semantic + keyword hybrid search over the full canon
- Every answer cites book · chapter · verse
- Cross-reference discovery between passages
- Runs fully on own infrastructure
Flagship Build 2 · ERP Sales-Data AI (Enterprise)
This build runs on a client's confidential ERP dataset. Everything shown below is synthetic demo data — fictional brands and figures that illustrate the experience only.
Business users ask sales questions in plain English. The bot translates them into safe, read-only SQL over the ERP's sales fact and dimension tables (products, brands, customers, divisions, time), then explains the result in words.
What were our top 3 brands by revenue in Q1?
Top 3 brands by Q1 revenue (demo data):
| # | Brand | Revenue | vs LY |
|---|---|---|---|
| 1 | AquaMax Pumps | ₹ 48.2 L | +12% |
| 2 | VoltEdge Motors | ₹ 41.7 L | +6% |
| 3 | HydroFlow Valves | ₹ 33.9 L | −3% |
Grounded on: sales fact table joined to brand & date dimensions · read-only query
Why did HydroFlow Valves dip this quarter?
HydroFlow's decline (demo data) is concentrated in the South division (−18%), while other divisions grew slightly. Two large accounts reduced reorder volume in February — worth a follow-up with the regional team.
Grounded on: division & customer breakdown of the same sales fact
- Plain-English → safe read-only SQL
- Answers cite the tables & filters used
- Division / brand / customer drill-downs
- Deployed inside the client network — data never leaves
Why teams trust it
Grounded answers
The model can only use retrieved context — if it isn't in your data, it says so.
Privacy-first
Runs against your own database and search index; confidential data stays in your stack.
Domain-tuned
Same core pipeline, tuned per domain — scripture study today, sales analytics tomorrow.
Want a private walkthrough?
Live demos are given one-on-one (the enterprise build involves confidential data). Bring your own documents or database — we'll show you what chatting with them feels like.
Contact for a Demo