Demo kits / Consumer Goods
Consumer Goods demo kit
The commercial world of Meridian Foods Group, a fictional manufacturer midway through a cocoa supply disruption, plus a real product corpus from USDA FoodData Central, packaged as a loadable knowledge graph with twenty-six competency questions and their expected answers. It loads into a free Fluree ledger in about two minutes.
No signup needed. After downloading, hand it to a coding agent, load it in a free workspace, or run it locally: the setup section covers all three, and takes about two minutes either way. Background reading: One Intelligence Layer, From Supplier Contract to Shelf.
Kit facts
- Download
- 42 KB zip, 9 files
- Model
- 35 classes and 145 properties
- Data
- 413 entities, 68 real USDA records
- Questions
- 26, each with an expected answer
- Standards used
- GS1 Web Vocabulary · GS1 EPCIS 2.0
- Data source
- CC0 1.0, USDA FoodData Central
- Runs on
- Free hosted workspace, or the Fluree CLI
- Updated
- August 2026
Meridian Foods, August 2026
Three layers in one graph: a fictional commercial world with a working storyline, a real licensed product corpus, and the upstream supply chain that connects farm plots to the promotions on shelf.
Meridian sells chocolate, biscuits, and spreads through four retail accounts. Its cocoa supplier declared force majeure in June; three live promotions are exposed through their bills of materials, retailer agreements constrain what can change, and the early warnings were sitting in documents months before the declaration. Every commercial fact is synthetic.
Beside the fiction sits a real corpus: 68 branded food records from USDA FoodData Central, public domain under CC0 1.0, typed with the GS1 Web Vocabulary and carrying per-record provenance. Fictional facts attach only to fictional products; the real records carry only what USDA publishes.
The questions run across all of it: trade, ERP, procurement, contracts, and documents in one traversal, which is the point.
What you can answer with it
- Promotion ROI which trade promotions made money, and what that verdict rests on
- Agreement limits what a retailer contract lets you change, before you negotiate
- Cost exposure trace a commodity move through recipes to the promotions it sinks
- Upstream trace farm plot to finished pallet, including where traceability stops
- Early warnings the signals sitting in documents months before a disruption
The documented trail
- Apr 14Buyer call note: Altena signals forward-hedging
- May 2Crop report: midcrop tracking 18% below average
- Jun 20Altena declares force majeure; allocations cut to 60%
- Aug 15Cocoa reaches $6,420 per tonne, up ~50% since April
Loading the kit
Three ways to run the kit, all free and all about two minutes. Every step is also written out in the README inside the download.
Paste this into Claude Code or Cursor
Any agent with terminal access can install the CLI, load the kit, validate it, and then answer the packaged questions and check itself against the expected answers that ship inside.
I downloaded cpg-demo-kit.zip from flur.ee. Please set it up and check that it works: 1. Unzip it and read README.md. 2. Install the Fluree CLI if I don't have it: brew install fluree/tap/fluree 3. Create a ledger and load the files in the order the README gives. 4. Run the SHACL validation and confirm the data conforms. 5. Then open competency-questions.md. Answer each question by querying the graph, and compare your answer to the expected answer written under it. Tell me which questions matched, which did not, and what the graph let you answer that separate systems could not.
Fluree AI, hosted
A workspace in the browser with the AI agent built in. Free tier, no credit card. This is the path for asking the questions in plain English.
The Fluree CLI
A single binary installed with Homebrew. This is the path for running the included SPARQL queries directly and for scripting.
brew install fluree/tap/fluree- 01
Create an empty ledger
In the workspace, add a new ledger; on the CLI, run fluree create. A ledger is a Fluree database, empty until you load it.
- 02
Import the six data files
Load model.jsonld first, then the five instances files. The model file defines the vocabulary the instance files use.
- 03
Run the questions
Ask the AI agent the competency questions as written and compare each response against the expected answer. CLI users can run the included SPARQL queries directly.
Questions with the answers attached
Twenty-six questions ship in the kit, each with the expected answer, so you can verify the responses are grounded rather than guessed. Here is one, exactly as packaged.
What early-warning signals preceded the June cocoa force majeure, and how much lead time did they offer?
A buyer call note from April 14 (Altena signaling forward-hedging, 67 days before declaration) and a crop report from May 2 (midcrop tracking 18% below average, 49 days before). The structured price series did not move decisively until June.
The ten commercial questions
Each is answerable only by combining more than one source domain. The note under each describes what the query touches.
- 01Which completed promotions at Hartland Grocers created profitable incremental margin, and which eroded it?Reads promotion evaluations for one retail account
- 02What baseline was the February TPR’s margin-eroding verdict measured against?Reads the baseline note stored on the evaluation record
- 03If we wanted to change the live Back-to-School TPR, what commitments and notice periods apply?Joins the promotion to its JBP lock window, notice period, and terms
- 04Which trade terms govern scan-backs at Hartland, and where did the monthly reconciliation come from?Traces a trade term to the amendment document that introduced it
- 05Which products and live promotions are exposed to cocoa through their bill of materials?Traverses commodity → BOM line → product → live promotion
- 06Cocoa is up roughly 50% since April. Which live promotions still clear their brand margin guardrail?Combines the price series, cost shares, discount depth, and guardrails
Sixteen further questions ship in the kit: three over the real USDA corpus, five over the upstream traceability layer in the core kit, four over the time module, two over the policy module, and two over the revisions module.
Extend the kit, one capability at a time
Each module is a separate download that adds one capability to the same scenario. Load the core kit first; each module’s README carries the full recipe with its outputs pasted verbatim.
Time travel
Every earlier state of the graph stays queryable, so you can ask what the data said on any past date.
Reconstruct the day the cocoa order was confirmed in full: two supply warnings on file, no force majeure declared yet, and a price curve that had not spiked.
Nine narrative windows inserted in order, one commit each. Add --at 6 to any query to answer as of June 4, the day the purchase order was confirmed in full.
Governed access
One question, asked by different people, returns different answers, enforced by the database rather than by prompts or separate copies.
Your analyst sees all ten promotions with unit costs. A retail buyer asking the same question sees only their own four, with cost columns blank.
One insert adds two identity records and six stored access policies. Pass the identity with --as; the query text never changes.
Decision memory
Changes supersede rather than overwrite, and the decision behind each change is stored as data you can query.
See the Northlight margin floor move from 22 to 24 percent, who raised it, the cocoa disruption they cited, and what the floor was before.
Two updates raise the Northlight margin floor and close a finished promotion; two decision records link each change to its evidence.
What is in each zip
The core kit is nine files. The three modules are separate zips, each self-contained with its own README. Nothing requires a signup.
cpg-demo-kit.zip
Download (42 KB)- model.jsonldThe vocabulary: GS1 typing, EPCIS 2.0 events, the cpg extension. Annotated throughout.
- instances-supply.jsonldSuppliers, contracts, commodity prices, the force majeure and its document trail.
- instances-products.jsonldThree brands, ten SKUs, bills of materials, margin guardrails.
- instances-commercial.jsonldRetailers, joint business plans, trade terms, promotions, POS, deductions, P&L.
- instances-usda-branded.jsonld68 real branded food records, per-record provenance.
- instances-traceability.jsonldGhana farm plots to export lots: EPCIS events plus the PO-to-invoice chains.
- shapes.jsonSHACL integrity rules, run in-database with fluree validate.
- competency-questions.mdAll 26 questions with expected answers; three as ready-made SPARQL.
- README.mdLoad instructions, standards references, data-source and license details.
Time travel
model.jsonld, nine window files, README with the commit map
Governed access
policy.jsonld, three queries, README with pasted outputs
Decision memory
two updates, the decision records, README with the sequence
The shipped data conforms to the included SHACL shapes; break it and fluree validate names the violations.
Questions about the kit
For a guided walkthrough of the kit, or to talk about running the same pattern on your own data, contact us.