# VDVC — Valsts dokumentu vadības klasifikators

**VARAM Dokumentu vadības klasifikācijas shēma** — machine-readable classification schema for the DVS Namejs document management system.

## Overview

This package transforms VARAM's Excel-based document classification schema (647 entries) into a structured, machine-readable format. It provides two variants:

| | V1 (as-is) | V2 (proposed) |
|---|---|---|
| **Approach** | 1:1 from Excel, normalised | Functional restructuring per MK Nr. 282 §33 |
| **Categories** | 493 leaf categories | 181 (132 + 45 + 4) |
| **I2 projects** | 33 groups × 4 sub-cats = 132 | 4 functional + 33 project tags |
| **I1 programmes** | 13 groups × ~17 sub-cats = 227 | 45 functional + 13 programme tags |
| **Correspondence** | 9 audience sub-categories | 1 category + 9 audience tags |
| **Retention** | 50+ free-text variants | 5 structured types + conditional rules |
| **Numbering** | P-1-13-4, I1-1-1.2, Il-9-2 | 1.1.13, 2.1.2, 3.1 (clean numeric) |
| **MCP served** | No | Yes |

## Files

```
├── vdvc-classification.xsd          XSD schema (both V1 and V2)
├── classification-v1-asis.xml       V1: 493 categories, original structure
├── classification-v2-proposed.xml   V2: 181 categories, functional
├── vocabularies.xml                 Controlled vocabularies (depts, programmes, projects, audiences)
├── migration-mapping.csv            V1→V2 lookup table (355 rows)
├── mcp-config.yaml                  MCP server configuration (V2 only)
├── index.html                       GUI editor (display V1, edit V2)
└── README.md                        This file
```

## Design Decisions

### Q1: Metadata Identifiers
- **Q1a**: Programme codes use internal IDs (`PRG-001`..`PRG-013`), not EU codes
- **Q1b**: Project codes use internal IDs (`PRJ-001`..`PRJ-033`), not EU project numbers
- **Q1c**: Department codes (e.g. `IPD`, `LN`) are in the vocabulary file, with cross-reference placeholder to VARAM Organizations Register

### Q2: V1 Normalisation
- **Q2a**: Mixed separators normalised (dots→hyphens: `P-1-18.1` → `P-1-18-1`). `Il` typo → `I1`.
- **Q2b**: 93 NBSP retention values treated as empty (element omitted)

### Q3: V2 Restructuring
- **Q3a**: P-1-13 correspondence collapsed to one category (`1.1.13`) with `retentionRule` elements per audience. Different retention terms preserved as conditional rules.
- **Q3b**: I1/I2 retention always on category+programme/project combination. Default retention on category, overrides via `retentionRule`.
- **Q3c**: Clean numeric domains: `1` (Pārvalde), `2` (Programmu vadība), `3` (Projektu ieviešana). Dot separators: `1.1.13`, `2.3.1`, `3.2`.

### Q4: GUI
- **Q4a**: Display V1 (read-only), edit V2
- **Q4b**: Download XML for manual commit (MVP)

### Q5: MCP
- **Q5a**: V2 only
- **Q5b**: Clean normalised structure in `generate_document`

## Retention Normalisation

50+ free-text variants → 5 structured XML types:

| Type | XML Element | Example Original |
|------|-------------|-----------------|
| Permanent | `<permanent/>` | "Pastāvīgi", "Patstāvīgi" |
| Duration | `<duration years="5"/>` | "5 gadi" |
| Triggered duration | `<duration years="5" trigger="project_closure"/>` | "5 gadi pēc projekta noslēguma pārskata apstiprināšanas" |
| Fixed date | `<fixedDate>2034-12-31</fixedDate>` | "31.12.2034.", "2031-12-31" |
| Fixed date (EU) | `<fixedDate euCommission="true">2032-12-31</fixedDate>` | "31.12.2032. EK" |

Additional special types: `<untilReplaced/>`, `<untilNoLongerNeeded/>`.

Trigger types: `creation`, `project_closure`, `programme_closure`, `contract_end`, `employment_end`, `replacement`, `assessment`, `last_debt_cleared`, `description_complete`, `procurement_execution`, `procurement_signing`, `disciplinary_decision`.

## V2 Conditional Retention (retentionRule)

When a category's retention depends on metadata (audience, programme, project), V2 uses `retentionRule` elements:

```xml
<vdvc:category code="3.1">
  <vdvc:name>Korespondence</vdvc:name>
  <!-- Default: 5 years after project closure -->
  <vdvc:retention>
    <vdvc:duration years="5" trigger="project_closure"/>
  </vdvc:retention>
  <!-- Override for PRJ-002: fixed date -->
  <vdvc:retentionRule label="Green LUPO">
    <vdvc:condition><vdvc:projectRef>PRJ-002</vdvc:projectRef></vdvc:condition>
    <vdvc:retention><vdvc:fixedDate>2034-04-30</vdvc:fixedDate></vdvc:retention>
  </vdvc:retentionRule>
  <vdvc:requiresTag type="project" required="true"/>
</vdvc:category>
```

Resolution order: specific project/programme/audience rule → default retention.

## Namespace

```
urn:vdvc:classification:2026
```

Prefix: `vdvc:`

## Legal Basis

- **MK noteikumi Nr. 282** (20.04.2012.) — "Noteikumi par dokumentu izstrādāšanu un noformēšanu" §33: classification schema must be based on document functions.
- **VARAM iekšējie noteikumi** — current classification schema approved by ministry order.

## ProcessGit Repository

```
https://processgit.org/VARAM/Document_classification_schema
```

MCP endpoint auto-generated from `mcp-config.yaml`.

## Migration from V1 to V2

Use `migration-mapping.csv` for automated document re-classification:

```csv
v1_code,v1_name,v2_code,v2_name,tag_type,tag_id,notes
I2-2-1,"Korespondence",3.1,Korespondence,project,PRJ-002,
I2-2-2,"Līgumi...",3.2,Līgumi un akti,project,PRJ-002,
P-1-13-1,"Sarakste ar republikas...",1.1.13,Sarakste,audience,republic-orgs,conditional retention
```

Migration steps:
1. Look up `v1_code` in CSV
2. Assign `v2_code` as new classification
3. Add `tag_type`:`tag_id` as document metadata
4. Resolve retention from V2 category + tag combination