TCRIA is a legal evidence chain-of-custody and governance system.

It ingests heterogeneous collections of documents and produces an auditable bundle that records classification, traceability signals, governance gates, and accountability metadata.

Its role is not to write legal pleadings or decide legal theses.
Instead, it preserves a controlled documentary chain of custody before human legal decision-making.

Legal cases often begin with large collections of mixed documents, such as:

• PDFs
• DOCX files
• notes
• reports
• emails
• drafts
• contextual material

Most tools only extract text or rank files.

TCRIA introduces a governance layer that ensures:

• accusatory narratives are not promoted without accountability
• documentary evidence remains traceable
• responsibility for narrative promotion is explicit
• heterogeneous archives can be processed without producing uncontrolled conclusions

The system therefore acts as a custody and governance layer for legal documentation.

Most document analyzers simply extract text and rank files.

TCRIA adds governance gates so that accusatory content does not "pass" unless it carries explicit accountability metadata (DecisionRecord) and avoids prescriptive or condemnatory language.

This protects the chain of custody of narrative responsibility.

TCRIA treats document processing as a custody workflow, not merely analysis.

The system records how documents move through the pipeline:

document ingestion ↓ classification ↓ traceability signals ↓ governance gates ↓ audit bundle

This produces a controlled evidentiary trail before human interpretation.

Scan folders of mixed files:

• PDF
• DOCX
• TXT
• MD

Artifacts are classified as:

• neutral / context
• supporting evidence
• relevant evidence
• accusatory candidates

Three governance checks protect the custody chain:

prescriptiveGate

Blocks condemnatory or prescriptive language.

complianceGate

Requires explicit accountability metadata in strict mode.

traceabilityCheck

Looks for signals such as:

• dates
• references
• evidentiary markers
• currency indicators

Each run produces:

• JSON audit bundle
• Markdown report
• PDF audit report

These outputs document how the system interpreted the collection.

Blocked artifacts are not reprocessed by the engine.

Instead they generate a diagnostic report analyzing potential evidentiary relevance without promoting them to the official accusation bundle.

To promote accusatory content through the compliance gate, the document must include a DecisionRecord header.

Example:

[TCR-IA DECISION RECORD]
responsibleHuman: Rodrigo Baptista da Silva
declaredPurpose: Auditoria documental e organização de evidências para fins jurídicos
approved: YES
approvedAt: 2026-03-05
[/TCR-IA DECISION RECORD]

This ensures that responsibility for narrative promotion remains human-declared.

TCRIA intentionally does not:

• generate legal pleadings
• write accusations
• construct legal theses
• produce petitions automatically

Those activities require human legal judgment and responsibility.

TCRIA instead provides a governed evidentiary foundation that can later support human legal workflows.

From the repository root:

python -m venv .venv
source .venv/bin/activate
pip install -U pip
pip install -e .

Run from the repository root:

tcria scan ~/Downloads -o output/audit --strict

If your audit script is located elsewhere:

tcria scan ~/Downloads --script path/to/audit_accusation_bundle_with_tcr_gateway.py

A simple web interface is available via Streamlit:

streamlit run app.py

Each run produces an audit bundle:

output/audit/
    audit.json
    audit.md
    audit_report.pdf

These artifacts document:

• classification results
• governance gate outcomes
• traceability signals
• compliance diagnostics