PCProvider ComplianceMinimum-repository architecture
Parent convergence · 126 repositories screened

Five repositories for the first proof—not fifteen platforms.

Each research lane found a strong winner. Combining all of them would create duplicate case databases, workflow engines, document stores and policy surfaces. This page shows the deliberate minimum and the evidence triggers for adding more.

Recommended proof build

One authority layer. Four document primitives.

The product meaning remains SISO-owned. External repositories provide infrastructure and file-format mechanics.

Never outsourced:case/backstory semantics · source and edition manifests · claim/evidence lineage · gap ownership · approvals/declarations · regulatory validators · business audit · fixture outcomes
Repository-count ladder

What each additional platform really buys.

Broad feature count is not the same as useful coverage. Integration and authority overlap are explicit costs.

1 repo · spike

WKS Platform

Closest out-of-box case/BPMN workflow. Still needs domain evidence, documents, tenancy proof and rules.

Benchmark, not default base
3 repos · broad benchmark

WKS + OpenQMS + OPA

Case workflow, QMS patterns and policy engine—but competing models and substantial integration.

Too broad for MVP
7 repos · AI pilot

Minimum + LangGraph + Promptfoo

Add resumable agent state and regression testing only after direct state transitions work.

Evidence-triggered
10 repos · high assurance

+ Temporal, Langfuse, OPA, OpenFGA

Durability, trace platform, rules and relationship auth with significant operating cost.

Later-stage scale
System mechanics

How the minimum system works.

A model may propose content; it never becomes the case authority or mutates an approved artifact directly.

  1. 1
    Receive and manifest

    Files enter versioned storage and receive tenant, case, hash, sensitivity, scope and source records before parsing.

  2. 2
    Structure the case

    Pydantic contracts validate backstory, facts, unknowns, claims, evidence, gaps, decisions and approvals stored in Postgres.

  3. 3
    Run a pinned document job

    A Python worker reads the exact master edition and performs DOCX, XLSX or PDF operations from a versioned job contract.

  4. 4
    Emit receipts

    Every transformation produces input/output census, delta log, validation results and an artifact manifest.

  5. 5
    Approve exact versions

    Consultant and client approvals bind to specific claims, decisions and artifact versions.

  6. 6
    Finalise mechanically

    Finalisation creates immutable output records; submission remains a separately authorised human event.

Deferred-by-default

Every optional repo has a trigger.

No “we may need it later” dependencies.

LangGraph

Add when measured branching, interrupts and resume behaviour exceed direct state transitions.

Temporal

Add when high-assurance recovery and long-running execution justify a workflow service.

OPA

Add when rule volume and effective-date versioning require a separate tested catalogue.

OpenFGA

Add when sharing relationships exceed tractable Supabase RLS policies.

Langfuse

Add when model-call volume requires dedicated trace and prompt comparison.

Docling

Add when mixed-format fixtures fail narrow local parsing probes.

Architecture proof

Test the minimum against two real cases.

One proven success and one incomplete case must pass tenant isolation, edition selection, document census, gap routing and approval history before the stack expands.

Open next evidence gate