feat(W-007): wire DerivedRecognizer registry into CognitiveTurnPipeline (ADR-0149) (#274)

- RecognizerRegistry.first_admitted() — deterministic first-registered selection
- CognitiveTurnPipeline consults runtime registry when no recognizer explicitly passed
- ChatRuntime gains _pending_recognizer_examples + record_recognition_example()
- checkpoint_engine_state() derives and registers recognizer from accumulated examples
- RuntimeConfig.recognition_grounded_graph gate (already existed) controls wiring
- ADR-0149 decision record

chat/runtime.py

@@ -51,6 +51,8 @@ from teaching.discovery import (
 )
 from teaching.discovery_sink import DiscoveryCandidateSink
 from engine_state import EngineStateStore
+from recognition.anti_unifier import derive_recognizer
+from recognition.outcome import FeatureBundle
 from recognition.registry import RecognizerRegistry
 from core.config import DEFAULT_CONFIG, DEFAULT_IDENTITY_PACK, RuntimeConfig
 from core.physics.drive import DriveGradientMap, GradientField
@@ -634,6 +636,9 @@ class ChatRuntime:
         self._recognizer_registry: RecognizerRegistry = RecognizerRegistry()
         self._turn_count: int = 0
         self._pending_candidates: list[DiscoveryCandidate] = []
+        self._pending_recognizer_examples: list[
+            tuple[tuple[str, ...], FeatureBundle]
+        ] = []
         if self._engine_state_store is not None and self._engine_state_store.exists():
             self._load_engine_state()
 
@@ -652,6 +657,13 @@ class ChatRuntime:
         store = self._engine_state_store
         if store is None:
             return
+        if (
+            self.config.recognition_grounded_graph
+            and self._pending_recognizer_examples
+        ):
+            recognizer = derive_recognizer(tuple(self._pending_recognizer_examples))
+            self._recognizer_registry.register(recognizer)
+            self._pending_recognizer_examples.clear()
         store.save_recognizers(self._recognizer_registry.all())
         candidates_to_save = self._pending_candidates
         if self.config.auto_contemplate and candidates_to_save:
@@ -664,6 +676,18 @@ class ChatRuntime:
         store.save_discovery_candidates(candidates_to_save)
         store.save_manifest(self._turn_count)
 
+    def record_recognition_example(
+        self,
+        tokens: tuple[str, ...],
+        bundle: FeatureBundle,
+    ) -> None:
+        self._pending_recognizer_examples.append((tuple(tokens), bundle))
+
+    def first_admitted_recognizer(self):
+        if not self.config.recognition_grounded_graph:
+            return None
+        return self._recognizer_registry.first_admitted()
+
     def _checkpointed_response(self, response: ChatResponse) -> ChatResponse:
         self._turn_count += 1
         self.checkpoint_engine_state()

core/cognition/pipeline.py

@@ -123,7 +123,12 @@ class CognitiveTurnPipeline:
         self.runtime = runtime
         self._last_node_id: str | None = None
         self.teaching_store = teaching_store or TeachingStore()
-        self._recognizer = recognizer
+        if recognizer is not None:
+            self._recognizer = recognizer
+        elif hasattr(runtime, "first_admitted_recognizer"):
+            self._recognizer = runtime.first_admitted_recognizer()
+        else:
+            self._recognizer = None
         self._prior_surface: str | None = None
         self._turn_number: int = 0
         # ADR-0021 §Articulation: subjects of prior SPECULATIVE teaching

docs/decisions/ADR-0149-derived-recognizer-pipeline-wiring.md

@@ -0,0 +1,65 @@
+# ADR-0149 — Integrate DerivedRecognizer into CognitiveTurnPipeline
+
+**Status:** Accepted
+**Date:** 2026-05-25
+**Work item:** W-007
+
+---
+
+## Context
+
+ADR-0143 introduced deterministic `DerivedRecognizer` derivation and matching.
+ADR-0144 gave `CognitiveTurnPipeline` an epistemic graph carrier and an optional
+`recognizer` constructor slot. ADR-0146 added persisted engine state, including
+`RecognizerRegistry`. ADR-0148 / W-003 wired vault promotion so `COHERENT`
+entries can eventually become recognition evidence.
+
+The missing edge was live admission: the runtime had a registry, but the
+pipeline never consulted it. A populated registry was therefore inert unless a
+test or caller manually passed a recognizer to `CognitiveTurnPipeline`.
+
+---
+
+## Decision
+
+`RecognizerRegistry.first_admitted()` returns the first registered recognizer in
+deterministic insertion order, or `None` when empty.
+
+`ChatRuntime` now exposes `first_admitted_recognizer()`, gated by
+`RuntimeConfig.recognition_grounded_graph`. When the flag is false, the method
+returns `None` and the turn path is byte-behavior compatible with the previous
+runtime.
+
+`CognitiveTurnPipeline` uses an explicitly supplied recognizer first. When no
+recognizer is supplied, it asks the runtime for the first admitted recognizer.
+If the registry is empty, recognition remains absent and the existing
+intent-derived graph path is used.
+
+`ChatRuntime.record_recognition_example(tokens, bundle)` records deterministic
+training pairs for test harnesses and future automated collection. At
+`checkpoint_engine_state()`, after the vault promotion boundary, a non-empty
+pending example set is passed to `derive_recognizer()` and the result is
+registered before engine-state persistence.
+
+---
+
+## Null-Drop
+
+`recognition_grounded_graph=False` means the registry is ignored, no recognizer
+is passed to the pipeline, and pending examples are not derived at checkpoint.
+The default therefore preserves the previous turn behavior.
+
+---
+
+## Follow-Up
+
+Automated example collection is intentionally out of scope. `derive_recognizer()`
+requires `(TokenSequence, FeatureBundle)` pairs with span-level evidence, not
+teaching corpus templates. The follow-up path is:
+
+```text
+finalize_turn -> FeatureBundle with evidence spans -> record_recognition_example -> checkpoint derivation
+```
+
+That follow-up makes the registry self-populating; this ADR makes the registry
+live and replay-persistent.

recognition/registry.py

@@ -23,6 +23,12 @@ class RecognizerRegistry:
     def all(self) -> list[DerivedRecognizer]:
         return list(self._registry.values())
 
+    def first_admitted(self) -> DerivedRecognizer | None:
+        """Return the first registered recognizer, or None if registry is empty."""
+        if not self._registry:
+            return None
+        return next(iter(self._registry.values()))
+
     def __len__(self) -> int:
         return len(self._registry)
 

tests/test_adr_0149_recognizer_pipeline_wiring.py

@@ -0,0 +1,146 @@
+from __future__ import annotations
+
+from dataclasses import replace
+
+from chat.runtime import ChatRuntime
+from core.cognition import CognitiveTurnPipeline
+from core.config import DEFAULT_CONFIG
+from engine_state import EngineStateStore
+from recognition.anti_unifier import derive_recognizer
+from recognition.outcome import EvidenceSpan, FeatureBundle, NegativeEvidence
+from recognition.registry import RecognizerRegistry
+
+
+def _config(*, recognition_grounded_graph: bool):
+    return replace(
+        DEFAULT_CONFIG,
+        recognition_grounded_graph=recognition_grounded_graph,
+    )
+
+
+def _span(tokens: tuple[str, ...], start: int, end: int) -> EvidenceSpan:
+    return EvidenceSpan(start=start, end=end, text=" ".join(tokens[start:end]))
+
+
+def _bundle(
+    tokens: tuple[str, ...],
+    agent_span: tuple[int, int],
+    count_span: tuple[int, int],
+    unit_span: tuple[int, int],
+    agent: str,
+    count: int,
+    unit: str,
+) -> FeatureBundle:
+    return FeatureBundle.from_mapping(
+        {
+            "agent": (agent, _span(tokens, *agent_span)),
+            "count": (count, _span(tokens, *count_span)),
+            "intentionality": (
+                "possession",
+                _span(
+                    tokens,
+                    1 if tokens[0] in {"A", "The"} else 0,
+                    3 if tokens[0] in {"A", "The"} else 2,
+                ),
+            ),
+            "modality": (
+                "actual",
+                NegativeEvidence(0, len(tokens), "no modal counter-marker present"),
+            ),
+            "polarity": ("+", NegativeEvidence(0, len(tokens), "no negator present")),
+            "relation": ("has", _span(tokens, count_span[0] - 1, count_span[0])),
+            "tense": ("present", _span(tokens, count_span[0] - 1, count_span[0])),
+            "unit": (unit, _span(tokens, *unit_span)),
+        }
+    )
+
+
+def _examples() -> list[tuple[tuple[str, ...], FeatureBundle]]:
+    john = ("John", "has", "5", "apples")
+    mary = ("Mary", "has", "3", "books")
+    school = ("A", "school", "has", "100", "students")
+    library = ("The", "library", "has", "12", "chairs")
+    return [
+        (john, _bundle(john, (0, 1), (2, 3), (3, 4), "John", 5, "apple")),
+        (mary, _bundle(mary, (0, 1), (2, 3), (3, 4), "Mary", 3, "book")),
+        (
+            school,
+            _bundle(school, (1, 2), (3, 4), (4, 5), "school", 100, "student"),
+        ),
+        (
+            library,
+            _bundle(library, (1, 2), (3, 4), (4, 5), "library", 12, "chair"),
+        ),
+    ]
+
+
+def _recognizer():
+    return derive_recognizer(_examples())
+
+
+def test_registry_empty_no_recognizer_passed(tmp_path) -> None:
+    runtime = ChatRuntime(
+        config=_config(recognition_grounded_graph=True),
+        engine_state_path=tmp_path,
+    )
+
+    result = CognitiveTurnPipeline(runtime).run("A baker has 24 loaves", max_tokens=4)
+
+    assert result.epistemic_graph is None
+
+
+def test_registry_with_recognizer_wires_into_pipeline(tmp_path) -> None:
+    runtime = ChatRuntime(
+        config=_config(recognition_grounded_graph=True),
+        engine_state_path=tmp_path,
+    )
+    recognizer = _recognizer()
+    runtime._recognizer_registry.register(recognizer)
+
+    result = CognitiveTurnPipeline(runtime).run("A baker has 24 loaves", max_tokens=4)
+
+    assert result.epistemic_graph is not None
+    assert result.epistemic_graph.recognizer_id == recognizer.teaching_set_id
+    assert result.epistemic_graph.nodes[0].node_id.startswith(
+        f"{recognizer.teaching_set_id}:"
+    )
+
+
+def test_flag_off_registry_ignored(tmp_path) -> None:
+    runtime = ChatRuntime(
+        config=_config(recognition_grounded_graph=False),
+        engine_state_path=tmp_path,
+    )
+    runtime._recognizer_registry.register(_recognizer())
+
+    result = CognitiveTurnPipeline(runtime).run("A baker has 24 loaves", max_tokens=4)
+
+    assert result.epistemic_graph is None
+
+
+def test_first_admitted_returns_none_on_empty() -> None:
+    assert RecognizerRegistry().first_admitted() is None
+
+
+def test_first_admitted_returns_registered() -> None:
+    registry = RecognizerRegistry()
+    recognizer = _recognizer()
+
+    registry.register(recognizer)
+
+    assert registry.first_admitted() == recognizer
+
+
+def test_record_and_checkpoint_derives_recognizer(tmp_path) -> None:
+    runtime = ChatRuntime(
+        config=_config(recognition_grounded_graph=True),
+        engine_state_path=tmp_path,
+    )
+    for tokens, bundle in _examples():
+        runtime.record_recognition_example(tokens, bundle)
+
+    runtime.checkpoint_engine_state()
+
+    assert len(runtime._recognizer_registry) == 1
+    persisted = EngineStateStore(tmp_path).load_recognizers()
+    assert persisted == runtime._recognizer_registry.all()