Centurio System- & Moduldokumentation (CH)
🔍
🏛️ Plattform-Architektur
🇩🇪 Deutsch 🇬🇧 English 🇫🇷 Français 🇪🇸 Español 🇮🇹 Italiano 🇳🇱 Nederlands 🇵🇱 Polski 🇵🇹 Português 🇨🇿 Čeština 🇸🇰 Slovenčina 🇭🇺 Magyar 🇷🇴 Română 🇧🇬 Български 🇬🇷 Ελληνικά 🇩🇰 Dansk 🇫🇮 Suomi 🇸🇪 Svenska 🇱🇹 Lietuvių 🇱🇻 Latviešu 🇪🇪 Eesti 🇭🇷 Hrvatski 🇸🇮 Slovenščina 🇲🇹 Malti 🇮🇪 Gaeilge 🇷🇺 Русский 🇺🇦 Українська 🇹🇷 Türkçe 🇳🇴 Norsk bokmål 🇷🇸 Српски 🇮🇸 Íslenska 🇨🇭 Schwiiz/Suisse
🔑 Login
🏠 Alli Modul / core_platform / KI-Helfer & Automatisierung
🤖

KI-Helfer & Automatisierung

agents core_platform 🛡️ RLS Protected

Autonome KI-Assistenten, Belegextraktion, Confidence-Scoring und manuelle Freigabe-Workflows.

Fachliche Bestimmung: Orchestrierte Agenten-Infrastruktur für automatisierte Belegverarbeitung (OCR, Quittungserkennung, Rechnungsprüfung). Befolgt strikt die Regel 'AI is assistance, not authority': KI-Entscheidungen mit geringer Konfidenz oder kritische Finanzmutationen erzeugen automatisch einen Prüfauftrag (Review Task) und dürfen niemals unbestätigt gebucht werden.
⚡ Workflows & Läbenszykle (4) 🗄️ Datemodell & Fälder (5) 🔌 Iistigspünkt (UI & APIs) (97) 🛡️ Regulatorie & Compliance

⚡ Workflows & Läbenszykle

KI-Extraktion mit Human-in-the-Loop Freigabe agent_review_cycle

Belegeingang, KI-Extraktion von Positionen, Konfidenzprüfung und bedarfsweise manuelle Überprüfung.

flowchart TD A[Dokumenteneingang] --> B[OCR & KI-Extraktion] B --> C{Confidence >= 85%?} C -- Ja --> D[Automatische Vorbelegung] C -- Nein --> E[Review-Aufgabe für Mitarbeiter] E --> F[Mitarbeiter prüft & korrigiert] F --> D D --> G[Auditierter Geschäftsbeleg]
1. INGEST
Belegeingang
Empfang von Bild oder PDF in der universellen Inbox.
Aktionen: OCR-Prozess starten
2. EXTRACT
Extraktion
KI-Agent ermittelt Lieferant, Beträge, Steuersätze und Positionen.
Aktionen: Confidence-Score berechnen
3. GATE
Konfidenzprüfung
Liegt die Konfidenz unter dem Schwellenwert (z.B. 85%), Weiterleitung in Reviews.
Aktionen: Review-Task anlegen
4. CONFIRM
Freigabe
Benutzer bestätigt oder korrigiert die extrahierten Daten im Review-Screen.
Aktionen: Beleg buchen, Regel anlernen

Eigenen Prozess starten (Custom Workflow) cascade_target:business.process.custom

Kaskade (review_gated).

1. cascade
Eigenen Prozess starten (Custom Workflow)

Benutzerbefehl: general agent plan:command_general_agent

Geprüfter Durable-Plan (durable_plan), Start: manual_command.

1. general_agent
general_agent
{"key": "execute_command", "label": "Befehl prüfen, ausführen oder zur Freigabe vorlegen", "adapter": "agent.execute", "agent": "general_agent", "condition_fact": null}
Aktionen: general_agent

Benutzerbefehl: workflow dictation agent plan:command_workflow_dictation_agent

Geprüfter Durable-Plan (durable_plan), Start: manual_command.

1. workflow_dictation_agent
workflow_dictation_agent
{"key": "execute_command", "label": "Befehl prüfen, ausführen oder zur Freigabe vorlegen", "adapter": "agent.execute", "agent": "workflow_dictation_agent", "condition_fact": null}
Aktionen: workflow_dictation_agent

🗄️ Datemodell & Fälder

AgentChatThreads agent_chat_threads RLS Aktiv
12 Fälder / Spalte

Tabelle agent_chat_threads

Feldname Datetyp Pflicht Schlüssel Fachlichi Bedütig
user_id UUID Ja * FK ➔ users.id Feld user_id
title VARCHAR(255) Nein - Feld title
messages JSONB Ja * - Feld messages
id UUID Nein PK Feld id
tenant_id UUID Ja * FK ➔ organizations.id Feld tenant_id
created_at DATETIME Ja * - Feld created_at
updated_at DATETIME Ja * - Feld updated_at
created_by UUID Nein - Feld created_by
updated_by UUID Nein - Feld updated_by
source_type VARCHAR(32) Nein - Feld source_type
source_id UUID Nein - Feld source_id
metadata JSONB Ja * - Feld metadata
AgentDefinitions agent_definitions
18 Fälder / Spalte

Tabelle agent_definitions

Feldname Datetyp Pflicht Schlüssel Fachlichi Bedütig
name VARCHAR(64) Ja * - Feld name
version INTEGER Ja * - Feld version
description TEXT Nein - Feld description
enabled BOOLEAN Ja * - Feld enabled
accepted_inputs ARRAY Ja * - Feld accepted_inputs
trigger_events ARRAY Ja * - Feld trigger_events
required_tools ARRAY Ja * - Feld required_tools
output_schema JSONB Ja * - Feld output_schema
risk_level VARCHAR(48) Ja * - Feld risk_level
approval_policy VARCHAR(48) Ja * - Feld approval_policy
confidence_threshold NUMERIC(4, 3) Ja * - Feld confidence_threshold
review_threshold NUMERIC(4, 3) Ja * - Feld review_threshold
system_instruction TEXT Nein - Feld system_instruction
prompt_version VARCHAR(32) Ja * - Feld prompt_version
id UUID Nein PK Feld id
created_at DATETIME Ja * - Feld created_at
updated_at DATETIME Ja * - Feld updated_at
metadata JSONB Ja * - Feld metadata
AgentMandates agent_mandates RLS Aktiv
17 Fälder / Spalte

Tabelle agent_mandates

Feldname Datetyp Pflicht Schlüssel Fachlichi Bedütig
delegator_user_id UUID Ja * FK ➔ users.id Feld delegator_user_id
label VARCHAR(255) Ja * - Feld label
intent_keys JSONB Ja * - Feld intent_keys
agent_names JSONB Ja * - Feld agent_names
max_risk_level INTEGER Ja * - Feld max_risk_level
may_mutate BOOLEAN Ja * - Feld may_mutate
expires_at DATETIME Ja * - Feld expires_at
revoked_at DATETIME Nein - Feld revoked_at
id UUID Nein PK Feld id
tenant_id UUID Ja * FK ➔ organizations.id Feld tenant_id
created_at DATETIME Ja * - Feld created_at
updated_at DATETIME Ja * - Feld updated_at
created_by UUID Nein - Feld created_by
updated_by UUID Nein - Feld updated_by
source_type VARCHAR(32) Nein - Feld source_type
source_id UUID Nein - Feld source_id
metadata JSONB Ja * - Feld metadata
AgentRuns agent_runs RLS Aktiv
35 Fälder / Spalte

Tabelle agent_runs

Feldname Datetyp Pflicht Schlüssel Fachlichi Bedütig
agent_name VARCHAR(64) Ja * - Feld agent_name
agent_version INTEGER Ja * - Feld agent_version
input_type VARCHAR(32) Ja * - Feld input_type
input_reference UUID Nein - Feld input_reference
status VARCHAR(48) Ja * - Feld status
started_at DATETIME Nein - Feld started_at
finished_at DATETIME Nein - Feld finished_at
model_provider VARCHAR(32) Nein - Feld model_provider
model_name VARCHAR(64) Nein - Feld model_name
prompt_version VARCHAR(32) Nein - Feld prompt_version
input_tokens INTEGER Ja * - Feld input_tokens
output_tokens INTEGER Ja * - Feld output_tokens
cost NUMERIC(10, 6) Ja * - Feld cost
confidence NUMERIC(4, 3) Nein - Feld confidence
result JSONB Ja * - Feld result
error TEXT Nein - Feld error
attempt INTEGER Ja * - Feld attempt
requires_review BOOLEAN Ja * - Feld requires_review
reviewed_by UUID Nein FK ➔ users.id Feld reviewed_by
reviewed_at DATETIME Nein - Feld reviewed_at
correlation_id UUID Nein - Feld correlation_id
causation_id UUID Nein - Feld causation_id
source_event_id UUID Nein - Feld source_event_id
source_agent_run_id UUID Nein - Feld source_agent_run_id
workflow_run_id UUID Nein - Feld workflow_run_id
workflow_depth INTEGER Ja * - Feld workflow_depth
id UUID Nein PK Feld id
tenant_id UUID Ja * FK ➔ organizations.id Feld tenant_id
created_at DATETIME Ja * - Feld created_at
updated_at DATETIME Ja * - Feld updated_at
created_by UUID Nein - Feld created_by
updated_by UUID Nein - Feld updated_by
source_type VARCHAR(32) Nein - Feld source_type
source_id UUID Nein - Feld source_id
metadata JSONB Ja * - Feld metadata
AgentToolCalls agent_tool_calls RLS Aktiv
14 Fälder / Spalte

Tabelle agent_tool_calls

Feldname Datetyp Pflicht Schlüssel Fachlichi Bedütig
agent_run_id UUID Ja * FK ➔ agent_runs.id Feld agent_run_id
workflow_run_id UUID Nein - Feld workflow_run_id
sequence INTEGER Ja * - Feld sequence
tool_name VARCHAR(64) Ja * - Feld tool_name
input JSONB Ja * - Feld input
output JSONB Ja * - Feld output
status VARCHAR(48) Ja * - Feld status
started_at DATETIME Nein - Feld started_at
finished_at DATETIME Nein - Feld finished_at
error TEXT Nein - Feld error
id UUID Nein PK Feld id
tenant_id UUID Ja * FK ➔ organizations.id Feld tenant_id
created_at DATETIME Ja * - Feld created_at
updated_at DATETIME Ja * - Feld updated_at

🔌 Iistigspünkt (UI & APIs)

Art Method Endpunkt / Pfad Titel & Zweck Berächtigig
⚡ REST API DELETE /api/v1/agents/chat/thread
clear_chat_thread
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API GET /agents
agents_page
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/agents
list_agents
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/agents/automation-catalog
list_automation_catalog
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/agents/chat/thread
get_chat_thread
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/agents/core-workflows
list_core_workflows
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/agents/coverage-audit
list_coverage_audit
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/agents/integration-status
integration_status
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/agents/mandates
list_agent_mandates
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/agents/workflows
list_workflows
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/agents/{agent_name}/runs
list_agent_runs
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/capabilities
list_capability_catalog
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/capabilities/coverage
capability_coverage
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/capabilities/{capability_key}
get_capability_descriptor
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/agent-roles
list_agent_roles
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/customizations/attribute-rules/active
list_attribute_rules
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/customizations/attributes/active
list_custom_attributes
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/customizations/order-release/active
active_order_release_policies
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/extension-forms/{object_key}/draft
get_extension_draft_form
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/extensions/{object_key}/{entity_id}
get_extensions
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/extensions/{object_key}/{entity_id}/form
get_extension_form
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/order-release/{quote_id}
quote_release_decision
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/process-catalog
process_catalog
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/process-library
process_library
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/process-proposals
list_process_proposals
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/process-runs/{run_id}
get_process_run
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/reflection
inspect_system_reflection
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/simulation-targets/{object_key}
simulation_targets
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/status-machines
list_status_machines
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/worker-bindings
list_worker_bindings
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/worker-supervisors
list_worker_supervisors
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /api/v1/semantic/workflow-setup/options
workflow_setup_options
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /automatisierung
automation_alias
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /portal/{slug}/assistent
assistant_page
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /semantic
semantic_page
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /semantic/advanced
semantic_advanced_page
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /semantic/changes
semantic_changes_page
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /semantic/customizing
semantic_customizing_page
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /semantic/designer
semantic_designer_page
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /semantic/engine
semantic_engine_page
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /semantic/engine/{run_id}
semantic_engine_instance_page
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /semantic/integration
semantic_integration_page
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /semantic/modelle
semantic_models_redirect
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /semantic/processes
semantic_process_page
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /semantic/records/{object_key}/{entity_id}
semantic_record_page
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /semantic/setup
semantic_setup_redirect
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /semantik
semantik_home
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API GET /semantik/objekte
semantik_objects
Semantic binding: not_evaluated; mutation=False
 agents
⚡ REST API POST /api/v1/agents/command
general_command
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/agents/mandates
create_agent_mandate
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/agents/mandates/{mandate_id}/revoke
revoke_agent_mandate
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/agents/transcribe
transcribe_command_audio
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/agents/{agent_name}/trigger
trigger_agent
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/capabilities/invoke
invoke_capability
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/agent-context/preview
agent_context_preview
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/context
resolve_context
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/context/entity
resolve_entity_context
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/customizations/attribute-rules/proposals
propose_attribute_rule
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/customizations/attribute-rules/retirement-proposals
propose_attribute_rule_retirement
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/customizations/attributes/proposals
propose_custom_attribute
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/customizations/attributes/retirement-proposals
propose_attribute_retirement
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/customizations/attributes/suggest-from-text
suggest_custom_attribute_from_text
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/customizations/context/proposals
propose_context_meaning
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/customizations/entry-points/adjustments
propose_entry_adjustment
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/customizations/entry-points/proposals
propose_entry_point
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/customizations/forms/proposals
propose_form
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/customizations/main-processes/proposals
propose_main_process
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/customizations/meaning/proposals
propose_meaning
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/customizations/order-release/deactivation-proposals
propose_order_release_deactivation
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/customizations/order-release/proposals
propose_order_release_policy
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/customizations/subprocesses/proposals
propose_subprocess
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/customizations/workflows/presentation-adjustments
propose_workflow_presentation_adjustment
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/designer/proposals
propose_canvas
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/entries/simulate
simulate_active_entry
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/entry-context
entry_context
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/order-release/{quote_id}/approve
approve_quote_release
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/process-runs
start_custom_process
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/process-runs/{run_id}/decision
decide_custom_process
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/process-runs/{run_id}/form
submit_custom_process_form
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/process-runs/{run_id}/resume
resume_custom_process
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/proposals
create_proposal
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/proposals/{release_id}/activate
activate
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/proposals/{release_id}/approve
approve
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/proposals/{release_id}/simulate
simulate
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/worker-bindings
create_worker_binding
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/worker-bindings/{binding_id}/activate
activate_worker_binding
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/worker-bindings/{binding_id}/approve
approve_worker_binding
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/worker-bindings/{binding_id}/simulate
simulate_worker_binding
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /api/v1/semantic/workflow-setup
workflow_setup
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /portal/{slug}/assistent
assistant_ask
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /semantic/engine/{run_id}/decision
semantic_engine_decision
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /semantic/engine/{run_id}/form
semantic_engine_form_submit
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API POST /semantic/engine/{run_id}/retry
semantic_engine_retry
Semantic binding: not_evaluated; mutation=True
 agents
⚡ REST API PUT /api/v1/semantic/extensions/{object_key}/{entity_id}/{field_key}
set_extension
Semantic binding: not_evaluated; mutation=True
 agents
🖥️ WEB-UI GET /reviews
Review-Center
Mitarbeiter-Freigabe für unsichere Belege und Daten
 reviews:read
⚡ REST API POST /api/v1/agents/run
Agent ausführen
Startet eine KI-Aufgabe asynchron
 agents:write
⚡ REST API GET /api/v1/audit
Audit-Logs abrufen
Revisionssichere Protokolleinträge durchsuchen
 audit:read

🛡️ Regulatorie & Compliance

Dieses Modul unterliegt folgenden gesetzlichen, regulatorischen und Revisionsstandards:

  • EU AI Act (Transparenz und menschliche Aufsicht)
  • GoBD Nachvollziehbarkeit
Verknüpfti Modul:
➔ inbox ➔ documents ➔ reviews ➔ matching