You've already forked dokumenta-semantiska-analize
Import UAPF package
Rihards
9b3790c1fa
Per UAPF v2.5.0, tests move from sidecar files (tests/algorithms/<card-id>.test.yaml — removed in v2.5.0) into a top-level tests array on each algorithm card. Minimum two entries per card; the Algorithm Card viewer (UAPF chapter 13.16, ProcessGit Preview tab) consumes these as its primary interaction surface. This package's three cards now carry embedded tests: - algo.semantic_document_analysis.pii_redactor (deterministic redactor) — 3 cases: Latvian personas kods inline (positive — three entity types detected), plain administrative text (negative — no PII signals), financial figures with IBAN (mixed — financial yes, personas_kods no). - algo.semantic_document_analysis.vdvc_semantic_extractor (stochastic LLM extractor, EU AI Act high-risk + mandatory oversight) — 2 cases: regulatory construction-permit appeal (in-domain, expected topic + applicable_regulations), non-regulatory thank-you note (out-of-domain, low confidence). Both carry ai_confidence_score tolerance bands appropriate for a stochastic output. - algo.semantic_document_analysis.completion_event_emitter (deterministic CloudEvents emitter) — 2 cases: successful completion event, failure completion event. The emitter does not gate on payload contents, so both succeed. Other changes: - uapf.yaml + manifest.json: version 3.1.0 -> 3.2.0 - README.md: v3.2.0 section added describing embedded tests and the removed sidecar location BPMN file unchanged from v3.1.0 — uapf:algorithmCardRef on each service task per UAPF v2.4.0 + ioSpecification synthesis. Mappings unchanged. DMN tables unchanged. uapf-cli validate against v2.5.0 schemas passes cleanly.
120 lines
4.1 KiB
YAML
120 lines
4.1 KiB
YAML
kind: uapf.algorithm.card
|
|
id: algo.semantic_document_analysis.vdvc_semantic_extractor
|
|
version: 1.0.0
|
|
name: VDVC semantic metadata extractor
|
|
intent: |
|
|
Extracts a VDVC v1.1-conformant structured semantic summary from the redacted document text — primary topic, keywords, classification, summary, sensitivity signals. Output validates against resources/schemas/vdvc-semantic-summary.schema.json. This is the sole model-inference step in the process; everything else in the package is deterministic.
|
|
algorithm_kind: extractor
|
|
io:
|
|
inputs:
|
|
- id: redacted_content
|
|
type: string
|
|
cardinality: single
|
|
constraints:
|
|
maxLength: 200000
|
|
documentation: Output of the upstream PII redactor.
|
|
- id: schema_ref
|
|
type: string
|
|
documentation: Path to the JSON Schema the output must validate against.
|
|
outputs:
|
|
- id: semantic_summary
|
|
type: object
|
|
schema: ../resources/schemas/vdvc-semantic-summary.schema.json
|
|
- id: sensitivity_control
|
|
type: object
|
|
- id: ai_confidence_score
|
|
type: probability
|
|
- id: output_pii_error_count
|
|
type: integer
|
|
constraints:
|
|
minimum: 0
|
|
implementation:
|
|
type: external
|
|
medium: llm_prompt
|
|
uri: uapf-ip://capability/ai.extract@1
|
|
hash: sha256:0000000000000000000000000000000000000000000000000000000000000000
|
|
runtime:
|
|
capability: ai.extract@1
|
|
note: Host-fulfilled UAPF-IP capability. Specific model identity and prompt hash
|
|
are runtime concerns of the host; the Card declares the contract, not the implementation
|
|
choice.
|
|
determinism: stochastic
|
|
side_effects: external_call
|
|
confidence:
|
|
type: probability
|
|
threshold: 0.7
|
|
below_threshold: route-to:human.legal_reviewer (enforced by DMN human-validation-gate)
|
|
complexity:
|
|
typical_latency_ms: 8000
|
|
max_latency_ms: 60000
|
|
failure_mode: default:null + flag — DMN human-validation-gate routes low-confidence
|
|
outputs to PENDING_REVIEW.
|
|
limitations:
|
|
- Garie dokumenti (>50 000 znaki) tiek apgriezti — pirmie 50K + pēdējie 5K
|
|
- Nav juridisks vērtējums — tikai semantiska klasifikācija
|
|
- Latviešu valodas juridiskā retorika var samazināt recall
|
|
reference:
|
|
legal: EU AI Act 2024/1689, Pielikums III (augstā riska MI sistēmas), 13. pants
|
|
(caurspīdība).
|
|
url: https://eur-lex.europa.eu/eli/reg/2024/1689/oj
|
|
owners:
|
|
- type: team
|
|
id: uapf-stewards
|
|
contact: stewards@uapf.dev
|
|
lifecycle:
|
|
status: draft
|
|
since: '2026-05-20'
|
|
audit:
|
|
log_inputs: redacted
|
|
log_outputs: full
|
|
retention: 7y
|
|
risk:
|
|
aiActRiskClass: high
|
|
humanOversight: mandatory
|
|
transparencyTier: tier-3-full
|
|
tests:
|
|
- name: Regulatory iesniegums about administrative decision
|
|
description: Typical Latvian administrative complaint with redacted PII. The extractor
|
|
should identify topic + risk + applicable regulation.
|
|
inputs:
|
|
redacted_content: Iesniedzējs [NAME] iesniedza sūdzību par būvvaldes lēmumu Nr.
|
|
12345 atteikt būvatļauju adresē [ADDRESS]. Tiek lūgts pārskatīt lēmumu.
|
|
schema_ref: schemas/iesniegums/v1
|
|
expected_outputs:
|
|
semantic_summary:
|
|
topic: construction-permit-appeal
|
|
subject_area: administrative-law
|
|
applicable_regulations:
|
|
- BL
|
|
- APL
|
|
language: lv
|
|
sensitivity_control:
|
|
contains_decision_reference: true
|
|
external_communication_recommended: false
|
|
ai_confidence_score: 0.87
|
|
output_pii_error_count: 0
|
|
tolerance:
|
|
ai_confidence_score: 0.1
|
|
output_pii_error_count: 0
|
|
- name: Non-regulatory thank-you note
|
|
description: Out-of-domain input. Extractor should yield low-confidence summary
|
|
and a sensitivity flag that no decision is referenced.
|
|
inputs:
|
|
redacted_content: Paldies par jūsu pakalpojumu! Bija ļoti patīkami sadarboties
|
|
ar [NAME] no jūsu komandas.
|
|
schema_ref: schemas/iesniegums/v1
|
|
expected_outputs:
|
|
semantic_summary:
|
|
topic: non-actionable-correspondence
|
|
subject_area: feedback
|
|
applicable_regulations: []
|
|
language: lv
|
|
sensitivity_control:
|
|
contains_decision_reference: false
|
|
external_communication_recommended: false
|
|
ai_confidence_score: 0.62
|
|
output_pii_error_count: 0
|
|
tolerance:
|
|
ai_confidence_score: 0.15
|
|
output_pii_error_count: 0
|