Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Roadmap

The project converges two coupled artifacts that share substrate and are cited together: a byte-level sequence model for relational metadata understanding, and a BFO/CCO-grounded ontology + synthetic corpus that grounds and supplies its pretraining data. They are coupled by design — the ontology is the domain-adaptive surface, the H-Net+RWKV model is the ultimate fitness measure, and the reasoner (HermiT) is the model-independent oracle between them. A better ontology yields a better corpus yields a better model; the model’s behaviour, in turn, is the gate that certifies the ontology machinery.

The current go-forward programme — the milestone ladder, its gates, and the scaling discipline — is the Signals Programme, of which this page is the landing view. The end-to-end pipeline is documented stage-by-stage in The Relational Data Generation Pipeline; the agent-mediated machinery that runs it is specified in End-to-end + Meta-Harness + reasoner. This page summarises what is delivered, what is in flight, and what is gated, and points to the authoritative reference for each.

Two earlier framings have been reorganized under the Signals Programme — not retired. (1) The four-phase K2.5 PARL roadmap (supervised → reward → PARL → swarm RL) folded into the Signals milestone ladder; the swarm modules persist as infrastructure and the RL approach evolves through the GEPA / Agent-Lightning layers (see Agent Swarm). (2) The RLVR-for-ontology-generation line is long-horizon, not superseded: its four-component verifier R(O, I) is now realized as the deterministic membrane stack (HermiT/CCO, OntoClean, OQuaRE) built this cycle, and an SAE-instrumented-Qwen generator fine-tuned by GRPO against that reward is the Signals M4 apparatus for autonomous, local ontology extension. The agent-mediated propose/dispose loop is building and proving that reward now — current work in direct service of M4. The Concept brief, Semantic engine reference, and RLVR chapter carry the M4 research design; the reward-modeling / PARL / swarm-RL design notes remain in the tree.

Delivered substrate

The model track produced a working, air-gap-deployable training and evaluation stack, and a first real backbone.

  • End-to-end training pipeline. BDD-backed training on the gt-signals-dbpedia benchmark (120 DBpedia labels, 814 tables), with per-epoch boundary diagnostics and checkpoint discipline (outputs/runs/{run_id}/… with sidecars and pre-rendered Bokeh plots).
  • Gateway + leaderboard. A FastAPI gateway (port 8091) + React UI + the ABI-patched flash-attn / mamba-ssm build path. The leaderboard reads outputs/runs/ directly — no tracking daemon, no W&B, no MLflow. Deployment targets: devenv (local), CAI (PGlite), Zarf air-gap K8s.
  • v2 mixed-corpus pretrain (2026-04-27). The project’s first real byte-level backbone — 122k training steps on a mixed corpus (FineWeb-Edu + SQaLe + SchemaPile + FinePDFs-lab), single GPU. Stratified held-out evaluation shows non-degenerate representations across the trained-time slices, with bits-per-byte drops on domain-targeted data and general prose held flat. See Training Regime §10 for the full table.
  • Synthetic-corpus byte pretraining imparts transferable column skill. On a cells-only GitTables-DBpedia CTA frozen-backbone probe, a corpus-pretrained backbone reaches 0.66–0.70 accuracy vs. ≈0.12 at random init, with non-overlapping bootstrap CIs and rising Hewitt-Liang selectivity. Caveat: per-column CTA did not separate the full / no-ontology / no-schema ablation arms — flat CTA is surface-solvable — so the ontology’s load-bearing claim moved to the relational axis (the M2 lift, below). See the repo-root EVIDENCE.md (E1–E3, “already-supported claims”).

These artifacts are the read surface and the warm-start for everything below.

The ontology + corpus, and its rigor program

The ontology is the annotation vocabulary for Column Type / Column Property Annotation (CTA/CPA) over wide relational tables. It is content-derived from FinePDFs (the qdrant/ColBERT aperture, src/aegir/ontology/domain_index.py), then realized to a HermiT-validated OWL artifact at corpora/ontology/sdg-ontology.{omn,owl} (with HERMIT_CERTIFICATE.md). Its classes are intermediate-depth subsumers — the property-bearing classes a heterogeneous-but-coherent column belongs to — not leaf terms.

How it is authored, and every metric, gate, and membrane that governs it, is the subject of the canonical Ontology Authors Guide. The standing discipline is propose / dispose: an agent (or a human) proposes axioms; a stack of deterministic membranes disposes — a parse membrane (OWLAPI well-formedness), a reasoning-authority membrane (CCO imported, HermiT validates grounding against CCO’s disjointness axioms in the loop), and an OntoClean meta-property membrane (anti-rigid-cannot-subsume-rigid). The two strongest are un-fakeable; a failure returns its reason, and the author refines (the agent-mediated feedback loop).

The rigor program (delivered, gate GREEN). Benchmarked against the IOF/BFO signature, the realized ontology is metered by scripts/ontology_metrology.py (IOF rigor dimensions — definitional_completeness, bfo_grounded, realizable_machinery, def_annotation_coverage; OntoQA/OQuaRE structural metrics; OntoClean taxonomic-correctness proxies) and gated by the OQuaRE quality gate (scripts/ontology_oquare.py — six SQuaRE characteristics, IOF-anchored [1,5] bands, floors aggregate ≥ 3.5 and FunctionalAdequacy ≥ 3.0 plus HermiT-consistent, aim 3.9, wired HARD into aegir.lineup.sync._gate). Two pre-registered objectives, OQ-Structure and OQ-Rigor, are both MET: current definitional_completeness 0.554, bfo_grounded 0.896, realizable_machinery 10, 0 unsatisfiable classes, OQuaRE 4.24 — GREEN. The standing rule: no sync --push of the ontology Data Product below GREEN. See the repo-root EVIDENCE.md (OQ-Rigor / OQ-Structure).

The corpus pipeline + DDL spine (delivered; phase gate PASSED 2026-06-07). just metaflow is the entire corpus pipeline, idempotent per input window (src/aegir/flows/sdg_corpora_flow.py): harvest (FinePDFs, cursor-windowed) → the aperture (qdrant ColBERT MaxSim, src/aegir/ontology/domain_index.py) → derive (the gRPC engine over the axiom-pattern library, src/aegir/ontology/patterns.py) → membrane-gated promotion into the live catalog (src/aegir/ontology/catalog/catalog.json) → realization (HermiT-certified OWL) → the DDL spine — referential-integrity-true tables / views / FKs, polyglot-validated (Trino ∩ Spark), projected into Atlas (Phase Gate — Governance & DDL Spine) → dual-register prose chapters with retained thinking traces → verification and measurement (congruence — input-window concepts ↔ chapter concepts over the same ColBERT substrate, src/aegir/ontology/congruence.py — plus the sensitive scan, naturalness norms, and shape EMD vs SchemaPile) → one immutable, prev-linked run-zettel per completed run (src/aegir/lineup/zettel.py, citing the run’s strategy_id) and a lineup re-projection. The corpus is the byte-pretraining data and an independent publishable deliverable (corpora/, the sdg-corpora submodule). Path A (src/aegir/flows/path_a_training_flow.py) is the continued-pretraining augmentation on RWKV World v3 — the locked data-value harness whose calibrated α Path B (the Aegir architecture thesis) inherits.

The inverted topic layer (R1 — delivered; phase gate PASSED 2026-07-07). The live topic instrument (src/aegir/ontology/topic_layer.py): topics ≡ ontology-grounded concept anchors in qdrant — 462 anchors (the 29 SKOS domains + all 433 live catalog terms), content-addressed by collection sha — and items are anchor-proportional passage windows that align to one topic or none via hierarchical-margin MaxSim under the pre-registered, null-calibrated gate (τ* = 0.1065, the shuffled-window-null p95, report-not-tune). Ambiguous mass is the error signal and the lexicon is the parameter: anchor surfaces iterate through membranes M1–M8 that return their reasons (M8 — positive voice only — is the durable methodological rule). Every association is pinned to the collection state that adjudicated it and walkable in the lineup. This is R1’s topic layer: the corpus census over ontology-grounded topics re-runs per input window with no topic-model re-fit; the BERTopic-era instruments (topic_alignment.py, build_topic_model.py, T_I.pkl) are deprecated to v0.3 reproducibility. Full evidence table and methods: Phase Gate — The Inverted Topic Layer.

Go-forward — the Signals Programme milestone ladder

The programme is a high-dimensional, multi-objective optimization decomposed into a sequence of factored, low-DOF gates over a scaling ladder; the standing EVIDENCE rule is no scaled spend without a green gate. Full charter, the α×β interaction design, and the final-gate text are on the Signals Programme page; the pre-registered hypotheses, instruments, and decision rules are in EVIDENCE.md (repo root).

  • M0 — substrate-evolution machinery. DELIVERED (SUPPORTED). The reasoner gates and computes; the harness evolves: the HermiT consistency gate (inc-2a), the single-file harness H₀ (inc-2b), the Meta-Harness outer loop + a discovered harness H₁ (inc-2c), and the realization-as-CPA beachhead (inc-2d). See End-to-end + Meta-Harness and the §Meta-harness entries in EVIDENCE.md.
  • M1 — architecture baseline (the H-Net isolation gate). UNTESTED — the next gated milestone. Train H-Net+RWKV on RWKV-7’s open corpus, swapping only the tokenizer for byte-level dynamic chunking, and establish parity up the scaling ladder to RWKV-7-matched params. DOF = 1. Gate: H-Net+RWKV ≥ RWKV-7 at matched scale on standard evals.
  • M2 — instrument validity (the decisive corpus gate + proxy calibration). UNTESTED. A same-architecture, matched-budget matrix at ≥2 ladder rungs — arms {grounded mix / no-ontology ablation / standard-only} — with a 2-factor α×β cell (α = ontology-corpus fraction, β = SQL/DDL fraction) replicated across rungs. Eval: a FLOOR (grounded ≈ standard on general LM evals → non-degeneracy) AND a LIFT (grounded > ablation on relational + de-novo Data-Element elucidation, cells-only / control-task / PR-metric, bootstrap CI). It also calibrates the cheap proxies (R1 / coverage-close / corpus-quality) against the pretrain signal. Preconditions, both MET: the corpus’s max non-repetitive token yield is measured (Y_eff ≈ 7.93M byte-tokens, which caps α at scale), and the discriminating relational eval is realization-as-CPA — per chapter, cited templates → a Domain/Range TBox, corpus tables → an ABox, HermiT realizes each column’s type (CPA by a sound-and-complete oracle), with a Domain/Range-ablated matched-token control (pilot selectivity 0.79, 95% BCa CI [0.63, 0.88], permutation p = 1e-4 → CI-clean).
  • M3 — scale + the FINAL PHASE GATE. UNTESTED. Conditioned on M2: climb to RWKV-7-matched params, extrapolate α*(N)/β*(N) to target scale, and confirm the lift persists.
  • M4 — the forward door (unlocked by the final gate). Iterate the machinery: close the generator loop (SAE-instrumented Qwen with process-reward fine-tuning — the anti-Goodhart reward on the ontological-reasoning circuit, not the verdict), and apply the pipeline to a novel domain with minimal re-tuning (RASE generalization).

The final phase gate

At RWKV-7-matched scale, the ontology-grounded data mix yields an H-Net+RWKV model that (a) matches RWKV-7 on general/standard evals — the non-degeneracy floor — AND (b) exceeds the no-ontology-ablation control on relational understanding + de-novo Data-Element elucidation, CI-clean, with the α×β interaction and the mix-optimum scale-drift characterized.

Passing certifies the ontology machinery as a valid instrument for relational domain adaptation and authorizes the M4 forward door. Failing localizes the break to a named arrow — ontology→corpus degeneracy (M2 floor), corpus→model transfer (M2 lift), or scale-drift (M3) — each with its own remediation. This gate supersedes the proxy-only corpus-as-deliverable gate: the proxies are calibrated by it, never trusted ahead of it.

Next moves (July 2026)

With both corpus-side phase gates passed, the near-term ladder is:

  • Path A training runs (src/aegir/flows/path_a_training_flow.py) — the World-v3 continued-pretraining harness that isolates the data value of the generated corpus holding architecture fixed.
  • M1 — the H-Net isolation gate (above) — the model track’s next green light; blocks M2.
  • The latent-lens ladder. Supervision for it now accumulates by operation: every topic-layer association record is a training pair for the anchor-projection head (L3), and lens identity fields are already stamped on every record and collection state (L1 — done). L2, the latent-prediction auxiliary for Path-B pretraining, is the next major move on this ladder.

Shared infrastructure

Both artifacts depend on shared substrate beyond the ontology:

  • Gateway, leaderboard, and lineup. The FastAPI gateway + React UI (above), plus the lineup / KB — the LINEUP navigation primitive (Ward Cunningham, credited; not a wiki — no editing/forking) over the KB, which is a build projection of the three Data Products (ontology / relational / content) at build/dev/{current,scratch,archive} (src/aegir/lineup/, just kb-build). Roots are refs: current = the latest sdg-corpora release as a complete kasten, scratch = trunk (the full live projection), archive = past releases and frozen snapshots; promotion snapshot-freezes current into archive while trunk rolls on. Topic-layer associations project as walkable item notes (Scratch → Content → Items × Topics → item → term). The read surface for run sidecars, ontology-rigor metrics, and corpus-quality surfaces alike.
  • The strategy pillar. The sdg-strategy submodule (strategy/) is the “how it was made” Data Product — four pillars (lens / voices / knobs / targets) under a Merkle strategy_id (src/aegir/strategy/manifest.py), shadows-as-branches, promotion by cherry-pick. Every run-zettel cites its strategy_id; strategy drift at flow start is a hard failure.
  • Lineage substrate. The Atlas-on-AGE provenance graph, with OpenLineage / Marquez compatibility and Atlas deep integration — implemented in full. The discipline that keeps it non-dependent: the ontology is the single source of truth, so Atlas and every projection stay rebuildable-from-the-ontology-or-it’s-a-bug, never a master (Atlas edits suggest, never commit). See the Signals Programme’s source-of-truth diagram.
  • The engine and meta-harness. A gRPC engine serving Qwen3.6-35B-A3B-FP8 via vLLM under strict layering (engine→vLLM, workloads→gRPC), and the agent-mediated RETE/FSM control spine (src/aegir/meta_harness/) that orchestrates membrane-gated proposal.
  • Worktree-aware development tooling. git worktree-based cross-checkout dev with shared .git and per-worktree service gating; the cross-worktree SAE-feature streaming pipe lives here. See Worktree Aware Development.

Design principles

  1. Each milestone produces a usable artifact. M0 produced the substrate machinery (reasoner gates, evolving harness); the model track produced a training loop with leaderboard, an air-gap-deployable gateway, and a pretrained checkpoint; the ontology is itself a HermiT-validated, OQuaRE-GREEN, citable deliverable independent of any downstream model result.

  2. Empirical gates are real gates, not aspirations. No scaled spend proceeds without a green gate. The OQuaRE publish gate refuses sync --push of the ontology below GREEN; the M1 isolation gate blocks M2; the M2 floor-and-lift gates M3; the M3 final gate authorizes the forward door. A gate that honestly held RED (the ontology rigor gate, before the rigor-evolution loop closed it) is the evidence that the gates bind.

  3. Locked artifacts are hash-tracked end-to-end. Every run records its catalog version, locked-weights/null-statistics hashes, and run id in sidecar metadata; a strict-resume policy refuses to resume any run whose locked artifacts have drifted. Pipeline runs additionally pin the Merkle strategy_id and the topic-registry collection sha — every topic association carries the collection state that adjudicated it.

  4. Outward contracts stay narrow. The project publishes the sdg-corpora SHARE tier (ontology + SKOS vocabulary + DDL spine + corpus), trained checkpoints, and (when stable) the SAE feature dictionary. Consumers’ internal architectures (DST fusion, FSM session state, governance pipelines) are not the project’s concern; this decoupling is what makes each artifact shippable in isolation.

  5. Complexity is bounded. Each milestone adds exactly one new dimension of complexity (1-DOF where separable; a 2-factor cell only where the interaction is the hypothesis). Failure modes that respect this discipline are easy to diagnose; an honest revision pass on this document is the only protection against drift.

Long-horizon work

The agent swarm modules in src/aegir/swarm/ are scaffolding for the long-horizon multi-agent training task; no operational training uses them yet. The four-phase K2.5 PARL roadmap (supervised → reward → PARL → swarm RL) folded into the Signals milestone ladder, and the RL approach evolves through the GEPA / Agent-Lightning layers; the reward-modeling, PARL-training, and swarm-RL design notes remain in the tree as that design record.