Aperçu

🟪 AI Interface — archive_org Module

RTT/1 Interface Specification#

Identity#

  • Interface Name: ai_interface
  • Module: archive_org
  • Purpose: Define the contract, modes, inputs, outputs, and behavioral rules for AI agents interacting with the Internet Archive using the RTT operator chain.

1. Interface Purpose#

The AI interface provides a strict, deterministic, operator‑first contract that governs how an AI agent must:

  • receive requests
  • execute the six‑operator chain
  • enforce drift‑bounded reasoning
  • avoid content‑based inference
  • produce continuity‑aligned answers

This interface is the only valid entrypoint for AI execution in the archive_org module.


2. Entrypoint#

The AI interface exposes a single entrypoint:

archive_org_agent.handle_request(goal, target, constraints)

Where:

  • goal = what the user wants (explain, audit, compare, locate_stable)
  • target = the IA URL
  • constraints = optional filters (time range, snapshot limits, etc.)

The entrypoint must execute the full operator chain before producing any answer.


3. Inputs#

Field Type Description
goal string The agent’s mode (explain, audit, compare, locate_stable).
target URL The Internet Archive object to analyze.
constraints object (optional) Time range, snapshot filters, or structural constraints.

4. Outputs#

Field Description
answer Final drift‑bounded retrieval.
operators_used List of operators executed (always all six).
warnings Drift, continuity, or substrate warnings.
artifacts Structured outputs from each operator.

5. Supported Modes#

The AI interface supports four modes, all of which still require the full operator chain:

1. explain#

Explain operator outputs, drift, lineage, substrate, or continuity.

2. audit#

Verify structural correctness, drift levels, or lineage mapping.

3. compare#

Compare two snapshots structurally (not semantically).

4. locate_stable#

Identify earliest stable and most reliable versions.


6. Operator Contract#

The AI must execute operators in this exact order:

  1. METADATA_OPERATOR
  2. WAYBACK_OPERATOR
  3. LINEAGE_OPERATOR
  4. COLLECTION_OPERATOR
  5. PRESERVATION_OPERATOR
  6. DRIFTBOUND_RETRIEVAL_OPERATOR

No skipping.
No reordering.
No merging.
No collapsing.
No shortcuts.


7. Behavioral Rules (RTT/1 Agent Contract)#

The AI must:

Structural Rules#

  • Use all six operators for every request.
  • Treat drift as explicit, never implied.
  • Treat missing snapshots as uncertainty, not “no change.”
  • Prefer lineage continuity over recency.
  • Prefer stable substrates (PDF > HTML > OCR).
  • Use collection context to shape expectations, not conclusions.

Safety Rules#

  • No content‑based reasoning.
  • No semantic inference.
  • No speculation.
  • No hallucinated snapshots.
  • No assumptions about missing years.
  • No collapsing mixed substrates.
  • No overriding operator outputs.

Output Rules#

  • All answers must be drift‑bounded.
  • All drift > none must produce warnings.
  • All substrate instability must be explicit.
  • All regime shifts must be acknowledged.

8. Artifacts Returned#

The AI interface returns structured artifacts from each operator:

  • metadata
  • snapshots
  • drift_map
  • continuity_breaks
  • lineage_graph
  • transformations
  • regime_shifts
  • collection_id
  • coherence_clusters
  • related_objects
  • format
  • stability_score
  • drift_risk
  • multi_layer_flags
  • final drift‑bounded summary

These artifacts allow downstream systems (or students) to inspect the full reasoning chain.


9. Interface Summary#

The AI interface is:

  • deterministic
  • operator‑first
  • drift‑bounded
  • lineage‑aware
  • substrate‑aware
  • collection‑contextual
  • non‑speculative
  • RTT/1‑aligned

It defines the complete behavioral contract for any AI agent interacting with the Internet Archive through the TriadicFrameworks canon.

Updated