Skip to content

SPEC.md

Latest commit

 

History

History
95 lines (73 loc) · 3.29 KB

SPEC.md

File metadata and controls

95 lines (73 loc) · 3.29 KB
Raw

Hyperlex Technical Specification v0.3

Hermes skill (Python package repo). No hard dependency on Abraxas. Relevant Abraxas wire capabilities are Hyperlex modules under hyperlex.compat.abraxas.

API freeze: docs/api-v1.md · hyperlex.API_V1 (unchanged in 0.3.x)
Extended: hyperlex.API_EXTENDED (additive; includes Phase 5 simulation)

Runtime API

Ingest & analysis

  • ingest_signal(query: str, source: str = "mock") -> str
  • fetch_ingest(query, source="mock", structured=True, max_terms=8) -> dict
  • detect_memetic_patterns(query=..., ingest_source="mock", ...) -> dict
  • match_lineage(text, terms=None, min_confidence=0.42, registry=None) -> dict | None
  • compute_lineage_confidence(hits, family_terms, corpus) -> (float, dict)

Receipts

  • emit_receipt(result, out_dir=None, validate=True, append_ledger=True, ledger_path=None) -> Path
  • verify_receipt(payload) -> (bool, str)

Calibration

  • extract_forecasts(result, receipt_ref=None) -> list[dict]
  • settle / score_pair / score_series / settle_and_log / recompute_series
  • NOT_COMPUTABLE

Relay

  • relay_from_result / relay_forecasts / relay_series / list_runes

Lineage backfill (0.2.12+)

  • apply_backfill / inventory_backfill / list_backfill_packs / backpropagate_lineage

Phase 5 simulation (0.3.0+)

  • simulate_cultural_transmission(seed_term, ...) -> dict — multi-community cascade
  • run_multi_agent_memetics(seed_term, ...) -> dict — role lattice adoption
  • forecast_hyperstition_risk(...) / risk_from_analysis(result, ...)
  • build_family_phylogeny(family_id) / list_phylogeny_families()
  • run_phase5_scenario(seed_term, ..., analysis_result=None) -> dict

All Phase 5 packets set brier: null and provenance: SPECULATIVE (phylogeny: INFERRED structure).

Compat (optional host import)

from hyperlex.compat.abraxas import (
    to_brier_ledger_entry,
    to_brier_score_packet,
    to_operator_brier_review,
    list_hlx_runes,
    envelopes_from_result,
    CLAIM_LABELS,
)

Sources

mock, real, glossary, glossary_expanded, web, reddit, urban, wikipedia, x_search, firecrawl, crawl4ai, combined

Command surface

Hermes skill CLI (scripts/hyperlex.py)

check | doctor | sources | ingest | analyze | scan | relay
extract-forecasts | settle | score-series | verify-score-log
emit-receipt | list-receipts | ledger-stats | ledger-diff | archive-export
lineage-backfill | lineage-backprop
simulate          # Phase 5
diagram | signal | feedback | validate | verify-receipt | smoke

Package CLI

python -m hyperlex check|analyze|scan|relay|settle|score-series|simulate|version

Output contracts

Analysis result

  • observed, inferred, speculative
  • provenance.brier always null on open analysis
  • analysis: neologisms, virality, memetics, hyperstition, optional lineage

Phase 5 scenario

  • schema hyperlex.phase5_scenario.v1
  • nested: transmission, multi_agent, hyperstition_risk, optional phylogeny
  • brier: null at every level

Calibration

  • Forecasts never attach Brier; settlements + series after operator decision

Error handling

  • Ingest failures degrade; do not crash
  • Missing settlements → NOT_COMPUTABLE
  • Simulation is deterministic for fixed params (no RNG in 5.0)