Make. Explore. Evolve.

A framework for analytical knowledge systems.
Three phases. Three stores. One continuous cycle.

Knowledge isn’t built once.
It’s made, explored, and evolved.

Every analytical system follows the same rhythm:
build a view of reality, look at what it reveals, refine it.

jinflow gives this rhythm a name — and a structure.

Three phases, endlessly repeated

Make
Compile artifacts.
Build the database.
Validate everything.
→
Explore
See the findings.
Understand patterns.
Ask new questions.
→
Evolve
Refine artifacts.
Author new probes.
Capture knowledge.

Evolve feeds back into Make. The cycle never ends.
Each turn produces a richer, more accurate analytical system.

One Phase, One Tool

Each phase is a standalone, deployable application.

jinflow make

The Builder
UI-less processing engine. Pure transformation.

Takes artifacts (CSVs, YAML definitions, contracts)
and produces a built DuckDB.

Validates. Compiles. Builds. Done.

jinflow make --artifact-dir ./content
Key Properties
No UI. No interactivity. No AI.

Workspace isolation. Never mutates source files.

Deterministic. Same artifacts in → same DuckDB out.

Deployable anywhere. Laptop. CI. Cloud function.

jinflow explore

The Explorer
Visual desktop application.

Opens the DuckDB that make built.
Browse dimensions, findings, verdicts, expert knowledge.

See what the data says. Understand patterns.
Form new questions.
Key Properties
Read-only. Never writes to the DuckDB.

No AI. Human exploration, human insight.

Desktop or web. macOS, Linux, Windows, browser.

Any DuckDB. File picker or env var. No path assumptions.

jinflow evolve

The Evolver
AI-powered REPL that closes the cycle.

Queries what make built.
Understands what explore revealed.
Authors the next generation of artifacts.

jinflow evolve
Key Properties
The only tool with AI. Claude as co-pilot.

Reads the DuckDB. Writes artifacts.

Suggest, validate, compile. Full authoring workflow.

Human confirms. AI proposes, human approves.

Evolve writes artifacts.
Make builds them. Explore reveals what changed.

Evolve
new probe YAML
→
Make
compile + build
→
Explore
see new findings
→
Evolve
refine thresholds
→
…

Every turn through the cycle refines understanding.
The system gets smarter because the artifacts get better.

AFS · KLS · SIS

Three named stores. Each with a clear owner.

Every piece of state has a home

AFS
Artifact Store
Git
All domain content.
CSVs, YAML definitions,
contracts, tenant configs.

Version-controlled.
Separate from engine code.
KLS
Knowledge Store
DuckDB
The analytical database.
Built product of make.

Medallion layers, findings,
verdicts, expert knowledge.
Read-only for explore & evolve.
SIS
System Information Store
SQLite → D1
Operational state.
Build journal, instrument
lifecycle, compilation logs.

Strictly optional.
Everything works without it.

Clear ownership. No ambiguity.

AFS
artifacts
→
make
reads AFS
writes KLS
→
KLS
database
→
explore
reads KLS
KLS
database
→
evolve
reads KLS
writes AFS
→
AFS
artifacts
→
make
rebuilds

AFS is the input to make. KLS is the output.
Evolve reads the output and writes the input. The cycle closes.

What jinflow Believes

Five commitments that shape every design decision.

  • Domain agnosticism. jinflow knows about artifacts, compilers, tenants, and layers — not about hospitals, materials, or billing. The domain is content, not framework.
  • Engine / content separation. The engine ships as a versioned package. Domain-specific content lives in a separate git repo — the AFS.
  • DuckDB as the universal contract. Make writes the KLS. Explore reads it. Evolve queries it. No other shared state between the three applications.
  • Workspace isolation. Make never mutates source directories. Temp workspace → compile → build → output.
  • Cloud readiness. Same Python library locally and in Lambda. Different callers, same engine.

nuMetrix on jinflow

Hospital material flow analytics — built on the cycle.

jinflow provides the how.
nuMetrix provides the what.

jinflow (framework)
Medallion pipeline (Bronze → Silver → Gold)
YAML → SQL compilers
Multi-tenant isolation
Taxonomy engine
Explorer shell
AI-powered REPL

The machinery. Domain-agnostic.
nuMetrix (domain)
Cases, procedures, materials, billing
OPALE, SAP, Navision mappings
47 probes, 11 hypotheses, 10 diagnoses
55 SMEbits, 15 BitBundles
Gold contract (7 entities)
3 source systems

The knowledge. All in the AFS.

A real example from nuMetrix

  • Make: Compile 47 probe YAMLs into SQL. Build DuckDB for hospital_zeta. 68K cases, 2.4M billing rows.
  • Explore: Open the Explorer. See that “Revenue Leakage from Unbilled Materials” is confirmed at CHF 2.8M.
  • Evolve: Ask Claude: “Which cost centers drive the leakage?” — AI queries KLS, identifies top 5 cost centers, suggests a new probe targeting OR documentation gaps.
  • Make: Rebuild with the new probe. Now the system detects OR-specific billing gaps.
  • Explore: See the new findings. The hypothesis evidence score shifts. Understanding deepens.

The jinflow vocabulary

M
Make
The builder phase. Artifacts in, DuckDB out. Validate, compile, build. No UI, no AI.
E
Explore
The discovery phase. Visual, read-only, human-driven. See what the data says.
V
Evolve
The refinement phase. AI-assisted. Query the KLS, author new artifacts into the AFS.
AFS
Artifact Store
Git repo with all domain content. CSVs, YAMLs, contracts. Input to make, output of evolve.
KLS
Knowledge Store
DuckDB. The built analytical database. Sole output of make. Read-only for explore and evolve.
SIS
System Information Store
SQLite. Operational state: build journal, instrument lifecycle. Optional. Cloud path: D1.

Make it.
Explore it.
Evolve it.

Three applications. Three stores.
One continuous cycle of understanding.

The knowledge system that gets better every time you use it.

jinflow.io