Labs

Every lab runs against one shared stack, started once. The stack grows as you go: module 3 puts triples in Fuseki, module 8 adds Postgres and Ontop over it, module 10 adds a vector index. Labs are written so you can jump straight to any one of them, because each seeds whatever it needs.

Start the stack

git clone <this repo> semantic-layer-course
cd semantic-layer-course
make lab-up          # Postgres 5432, Fuseki 3030, Ontop 8080
make lab-status      # confirms all three answer

Needs Docker and about 3 GB of RAM. Stop it with make lab-down, wipe it with make lab-reset.

#	Lab	What you end up with	Time	Status
01	Reproduce the disagreement make lab lab01-two-answers	Run three defensible SQL queries against one loan book and get three different default counts.	15 min	todo
02	Watch consistency break make lab lab02-ace-coinflip	Run one natural-language loan question through a non-deterministic scorer five times and get five answers, then run the same question through a deterministic SPARQL rule five times and get one.	20 min	todo
03	Load Priya's mortgage into a triple store make lab lab03-first-triples	Write valid Turtle for customer C-1001 and loan L-1042, load it into Apache Jena Fuseki, and verify the triple count with a query.	20 min	todo
04	Answer the CRO's question three ways make lab lab04-sparql-drills	Run all three default definitions as SPARQL against Fuseki and reproduce the counts 1,847, 4,102 and 1,203 from one unchanged graph.	30 min	todo
05	Write the Meridian T-Box make lab lab05-build-ontology	Author meridian.ttl with the class hierarchy, properties with domain and range, disjointness and functional axioms; load it into Fuseki; run a SPARQL query that lists every class and its parent.	35 min	todo
06	Catch the bad loan data make lab lab06-shacl-gate	Write shapes.ttl, run pySHACL against the Meridian graph, get violations for Tobias's missing KYC document and an out-of-range credit score, then fix the data and get sh:conforms true.	25 min	todo
07	Derive the default set make lab lab07-run-reasoner	Run an OWL reasoner (owlrl, or ELK via Jena) over the Meridian ontology plus data, produce the inferred named graph, then query the three default classes and see L-2087 in Operational only, L-4590 in Regulatory and Operational, and L-5150 in all three.	35 min	todo
08	Query Postgres as a graph make lab lab08-ontop-vkg	Load Meridian's loan tables into Postgres, write the R2RML mapping, start Ontop as a SPARQL endpoint over it, then run the same SPARQL query from Module 4 and get the same answer with no data having moved.	40 min	todo
09	Govern a metric make lab lab09-metric-store	Define the non-performing loan ratio and total exposure as MetricDefinition triples, write a SHACL shape that validates metric definitions are well formed, then run a query that resolves a metric by name and returns the number plus its definition provenance.	25 min	todo
10	Make the credit policy queryable make lab lab10-graphrag	Chunk the Meridian Credit Policy PDF, extract concepts, build a small lexical graph in Fuseki, embed the chunks into a local vector index, then run a hybrid retrieval that answers the arrears question with a page citation and links the answer to the ontology class.	40 min	todo
11	Induce and review make lab lab11-induction	Auto-generate a candidate ontology plus R2RML mappings from Meridian's Postgres schema into a draft named graph, run ELK for consistency, diff it against your hand-written ontology, then promote the good parts and reject the bad.	35 min	todo
12	Wire the whole thing up make lab lab12-qa-agent	Run a local Q&A agent that classifies intent against the ontology, generates SPARQL, executes against Fuseki plus Ontop, runs the reasoner, retrieves a document citation, and returns the full answer plus evidence plus citations plus reasoning_steps JSON for the default question and for Daniel's application.	45 min	todo
13	Ship it on your tailnet make lab lab13-deploy	Run the full course stack with docker compose, verify each service is healthy, confirm the course app persists your progress across a container restart, and expose it over Tailscale.	30 min	todo