A public civic record and verification system for structured institutional analysis.
Observation sometimes becomes clearer when structure is applied.
Not designed for attention.
Designed for understanding.
Current release: v12
Release documentation:
https://civic-decision-engine-production.up.railway.app/
The Civic Decision Engine is a diagnostic system for analysing institutional behaviour through structured civic records.
It does not argue a case.
It classifies and interprets institutional behaviour based on structured inputs.
The system is designed to make institutional progression visible through:
- conditions
- trajectories
- transitions
- structural continuation
Version 12 introduces additive attachment infrastructure for referenced evidence artifacts while preserving canonical record verification hashes and canonical serialization.
New capabilities include:
- Additive attachment infrastructure for referenced evidence artifacts
- Independent SHA-256 hashing for attachment content
- Attachment metadata projection through public record manifests
- Optional source-document date metadata
- Privacy filtering for private, withheld, and deleted attachments
- Admin-only upload infrastructure protected by
CDE_ADMIN_TOKEN - Local read-only attachment inspection tooling
- Controlled production verification using synthetic test artifacts
- Preservation of canonical record verification hashes and canonical serialization
Attachments are additive and non-canonical. They do not replace canonical records, alter verification hashes, or change public record creation behavior.
v12 does not include public attachment downloads, public attachment serving, attachment search, OCR, PDF text extraction, semantic indexing of attachment content, or a public upload UI.
Given one or more civic cases, the system:
- structures the case into a standard form
- detects behavioural conditions
- identifies progression across cases
- classifies structural trajectories
- interprets the resulting system state
Outputs include:
- Timeline analysis → what is happening
- Pattern analysis → what it means
- Conditions analysis → what structure is present
- Public record generation → what is preserved
Provides a deterministic report layer that assembles and describes visible record data and existing framework outputs, including conditions, trajectories, administrative outputs, record evolution outputs, determination traces, rule citations, and evidence attribution outputs.
The report is descriptive only and does not perform new analysis or create new findings.
Identifies visible support boundaries for existing framework outputs by classifying whether outputs are supported, partially supported, or unsupported within the Evidence Attribution Matrix.
The layer evaluates visible support inside the framework only and does not determine truth, legal sufficiency, evidential sufficiency in the real world, liability, intent, blame, or factual correctness.
Identifies which framework layers and evidence categories are visibly present or not represented in the current record-derived outputs.
The layer is descriptive only. It does not generate hypothetical scenarios, assign meaning to absence, infer intent, predict outcomes, validate evidence, determine truth, or modify the record.
Provides an internal explainability certification layer that evaluates whether required Stage 19 explainability components are present and inspectable for a record.
The certification is internal to the framework only and does not certify truth, legality, liability, factual correctness, real-world evidential sufficiency, intent, blame, or wrongdoing.
Provides deterministic self-description of CREF identity, purpose, scope, implemented guarantees, declared constraints, reasoning architecture, and reflexive methodology.
Stage 20 describes only implemented framework definitions and outputs. It does not create methodology, introduce reasoning, infer facts, validate evidence, determine truth or liability, or modify records.
Organises the Admin Record Evidence view into Executive Report, Review Report, and Full Inspection Report modes. Executive mode presents concise current-state and progression summaries; Review mode adds trace, citation, attribution, and boundary summaries; Full Inspection mode preserves every existing report section and supporting-evidence detail.
Stage 21 changes presentation and report composition only. It does not change framework reasoning, evidence relationships, thresholds, classifications, hashes, or record data. Full Inspection Report remains the default mode.
The Civic Decision Engine supports publicly verifiable civic records.
Each record includes:
- stable reference identifier
- verification URL
- SHA-256 integrity hash
- canonical citation exports
- version lineage
- export timestamp
- structured condition classification
Records are:
- independently verifiable
- versioned
- hash-preserved
- publicly accessible
The archive supports:
- filtering
- pagination
- server-side search
- multilingual verification views
This version introduces the structured CONTINUATION_MAP.
It does not predict outcomes.
It reflects what structure tends to produce when conditions remain unchanged.
The map is derived from state, not interpretation.
This marks the transition from describing cases to identifying structural continuation.
The Conditions Layer is the core diagnostic layer of the system.
Conditions are reusable structural classifications describing institutional behaviour patterns.
Examples include:
TRANSFER_OF_BURDENESCALATION_WITHOUT_RESPONSESTABILITY_WITHOUT_CONFIRMATION
Conditions are:
- diagnostic
- reusable
- structurally defined
- independent of originating cases
This allows repeated institutional behaviour to be recognised across separate records and environments.
These documents define the formal structure and meaning of the system outputs.
The schema guarantees structure.
The specification defines interpretation.
https://civic-decision-engine-production.up.railway.app/
No setup required.
Users can:
- submit cases
- compare sequences
- browse records
- verify public records
- search archive entries
- explore condition patterns
- view structural trajectories
The /records archive provides:
- public record indexing
- institution filtering
- trajectory filtering
- condition visibility
- server-side search
- scalable pagination
- verification routing
Each archive entry resolves to a canonical verification page.
The system supports cross-case comparison to identify repeated structural behaviour across submissions.
Cross-case comparison operates independently and does not persist results.
This mode reveals patterns.
It does not create records.
This example shows a transition from delay to escalation.
curl -X POST "http://127.0.0.1:8000/analysis/timeline" \
-H "Content-Type: application/json" \
-d '{
"cases": [
{
"strike_reference": "Strike-Example-001",
"case_title": "Administrative complaint awaiting substantive response",
"civic_domain": "local_government",
"decision_trigger": "Acknowledged but no substantive response",
"urgency": "medium",
"institutions": ["Local Authority"],
"case_lifecycle": {
"current_stage": "awaiting_response",
"status": "active",
"stalled": false,
"days_open": 20
}
},
{
"strike_reference": "Strike-Example-002",
"case_title": "Unresolved complaint with missed deadline",
"civic_domain": "local_government",
"decision_trigger": "No response beyond deadline",
"urgency": "high",
"institutions": ["Local Authority"],
"case_lifecycle": {
"current_stage": "awaiting_response",
"status": "active",
"stalled": true,
"days_open": 90
}
}
]
}'
The system detects a transition from early delay to escalation:
{
"results": [
{
"conditions": [
"TRANSFER_OF_BURDEN",
"ESCALATION_WITHOUT_RESPONSE"
],
"trajectory": "Deteriorating",
"moment_of_change": {
"from": "TRANSFER_OF_BURDEN",
"to": "ESCALATION_WITHOUT_RESPONSE"
}
}
]
}
The sequence shows deterioration:
From delayed procedural burden
to escalation without response.
{
"pattern_summary": "Transition from TRANSFER_OF_BURDEN to ESCALATION_WITHOUT_RESPONSE detected within submitted sequence.",
"pattern_interpretation": "The submitted sequence shows deterioration from delayed procedural burden to escalation without response.",
"system_state": "TRANSITION_TO_ESCALATION",
"signals": [
"TRANSITION_DETECTED",
"ESCALATION_WITHOUT_RESPONSE_PRESENT"
]
}
Verification pages expose:
- canonical citations
- integrity hashes
- version lineage
- export timestamps
- permalink continuity
- multilingual rendering
- verification manifests
This allows records to be cited in:
- research
- journalism
- civic submissions
- administrative complaints
- public-interest documentation
The system does not determine outcomes.
It makes visible:
- when delay becomes pattern
- when pattern becomes structure
- when structure becomes persistent
- when continuation no longer requires explanation
uvicorn api.main:app --reload
Then open:
http://127.0.0.1:8000/docs
Core components include:
- Conditions Layer
- Timeline Analysis
- Pattern Interpretation
- Public Record Verification
- Archive Infrastructure
- Verification Hashing
- Citation Infrastructure
- Continuation Mapping
Active development (v12)
Current operational layers:
- Conditions Layer integrated
- Timeline analysis operational
- Pattern analysis operational
- Public verification infrastructure operational
- Searchable archive infrastructure operational
- Citation infrastructure operational
- Version lineage operational
The record does not argue.
It becomes clear enough
to be returned to.