Memory faculty: root entry surfaces; retire *_wiki names everywhere

agents/conductors/bug/_bug.py

@@ -377,7 +377,7 @@ def emit_human(mode: str, d: dict):
     if d["memory_context"]:
         print("Relevant context (PyAutoMemory — consult, do not invent):")
         for wiki, kws in d["memory_context"].items():
-            print(f"  - {wiki}/index.md  ({', '.join(kws)})")
+            print(f"  - PyAutoMemory/wiki/{wiki}/index.md  ({', '.join(kws)})")
     else:
         print("Relevant context:     none matched")
     print(f"Fix locus:            {d['fix_locus']['locus']}")

agents/conductors/feature/MIND_TAXONOMY.md

@@ -67,11 +67,11 @@ context when memory has material. Sub-wikis (source: `PyAutoMemory/index.md`):
 
 | Sub-wiki | Domain | Triggered by (examples) |
 |----------|--------|--------------------------|
-| `lensing_wiki/` | strong gravitational lensing | lens, deflection, source reconstruction, subhalo, time delay, cosmography, SLACS/TDCOSMO |
-| `smbh_wiki/` | supermassive black holes | black hole, SMBH, binary, recoil, NANOGrav |
-| `cti_wiki/` | charge-transfer inefficiency | CTI, trap, arctic, VIS calibration |
-| `methods_wiki/` | statistical / computational methods | Bayesian, sampler, JAX, NUFFT, SBI, graphical models, deep learning |
-| `galaxies_wiki/` | galaxy formation & evolution | bulge/disk, MGE, morphology, IFU, kinematics |
+| `wiki/lensing/` | strong gravitational lensing | lens, deflection, source reconstruction, subhalo, time delay, cosmography, SLACS/TDCOSMO |
+| `wiki/smbh/` | supermassive black holes | black hole, SMBH, binary, recoil, NANOGrav |
+| `wiki/cti/` | charge-transfer inefficiency | CTI, trap, arctic, VIS calibration |
+| `wiki/methods/` | statistical / computational methods | Bayesian, sampler, JAX, NUFFT, SBI, graphical models, deep learning |
+| `wiki/galaxies/` | galaxy formation & evolution | bulge/disk, MGE, morphology, IFU, kinematics |
 
 Library targets also pull a default sub-wiki (`autolens`→lensing,
 `autogalaxy`→galaxies, `autofit`/`autoarray`/`autoconf`→methods). PyAutoMemory is

agents/conductors/feature/_feature.py

@@ -305,7 +305,7 @@ def emit_human(mode: str, decision: dict):
     if d["memory_context"]:
         print("Relevant context (PyAutoMemory — consult, do not invent):")
         for wiki, kws in d["memory_context"].items():
-            print(f"  - {wiki}/index.md  ({', '.join(kws)})")
+            print(f"  - PyAutoMemory/wiki/{wiki}/index.md  ({', '.join(kws)})")
     else:
         print("Relevant context:     none matched (no scientific context required)")
     print(f"Phase decision:       {d['phase_decision']}")

agents/faculties/memory/_memory.py

@@ -2,11 +2,16 @@
 """agents/faculties/memory/_memory.py — the recall substrate.
 
 The **memory faculty** is a read-only opinion sink: given a topic/question, it
-greps the organism's knowledge surfaces — PyAutoMemory's sub-wikis,
-autolens_assistant's skills/wiki pages, and Mind's complete/ records — and
-returns a **cited digest**: ranked pages with matching snippets. The consulting
-agent reads only the listed pages and synthesises; this script never dumps
-content, never writes, and builds no index.
+greps the organism's knowledge surfaces — PyAutoMemory's sub-wikis and its
+root entry pages (index.md, reading-queue.md, bibliography/README.md, the
+wiki schema), autolens_assistant's skills/wiki pages, and Mind's complete/
+records — and returns a **cited digest**: ranked pages with matching
+snippets. The consulting agent reads only the listed pages and synthesises;
+this script never dumps content, never writes, and builds no index.
+
+Deliberately NOT a surface: ``bibliography/*.bib`` — the canonical BibTeX
+file is far over ``MAX_FILE_BYTES`` and raw entries are metadata, not digest
+text; ``bibliography/README.md`` carries the workflow instead.
 
 Exit codes: 0 digest · 4 no hits or no surface · 5 usage.
 """
@@ -32,6 +37,15 @@ def terms_of(query: str) -> list[str]:
 def surfaces(memory: Path | None, assistant: Path | None, mind: Path | None):
     """Yield (surface-name, root, md-file) triples, discovered at query time."""
     if memory and memory.is_dir():
+        # The repo's own entry surfaces: the hand-written index, the reading
+        # queue, the bibliography workflow, and the wiki schema. Without
+        # these the front door PyAutoMemory/AGENTS.md points every agent at
+        # ("index first") was invisible to this faculty (PyAutoBrain#239).
+        for rel in ("index.md", "reading-queue.md",
+                    "bibliography/README.md", "wiki/CLAUDE.md"):
+            f = memory / rel
+            if f.is_file():
+                yield "PyAutoMemory/root", memory, f
         # wiki/<domain>/ is the current layout; root-level *_wiki/ is the
         # pre-2026-07 layout, kept for older checkouts and forks.
         wikis = sorted(d for d in memory.glob("wiki/*") if d.is_dir())

agents/faculties/samplers/AGENTS.md

@@ -140,7 +140,7 @@ not inventoried.
   (prototypes + `_metrics.py` + `output/comparison.txt`),
   `autofit_workspace_developer/searches/` (removed-sampler archive),
   `autofit_workspace_test/scripts/searches/` (integration scripts).
-- **The science** — `PyAutoMemory/methods_wiki`:
+- **The science** — `PyAutoMemory/wiki/methods`:
   `concepts/hamiltonian-monte-carlo.md`, `concepts/gpu-nested-sampling.md`,
   `concepts/initialization-chaining.md`, `concepts/sampler-benchmarks.md`
   (the campaign record), `concepts/nested-sampling.md`. **Privacy seam:**

agents/faculties/samplers/_samplers.py

@@ -217,7 +217,7 @@ def digest(autofit, developer, test, lens_developer=None, profiling=None) -> dic
         "instruction": "reason with agents/faculties/samplers/AGENTS.md "
                        "(sampler<->likelihood match, gradients/JAX, "
                        "initialization chaining); the science lives in "
-                       "PyAutoMemory/methods_wiki — internal use only, "
+                       "PyAutoMemory/wiki/methods — internal use only, "
                        "citations never reach public output",
     }
     minimal = integration = []

config/policy.yaml

@@ -26,21 +26,22 @@ sizing_categories:
 extra_workspace_targets: [workspaces]
 extra_organism_targets: [autohands]
 
-# Keyword -> PyAutoMemory sub-wiki (also the science vocabulary difficulty
-# scoring keys off). Source of truth for the wiki list: PyAutoMemory/index.md.
+# Keyword -> PyAutoMemory sub-wiki (the wiki/<domain>/ layout; also the science
+# vocabulary the difficulty scoring keys off). Source of truth for the wiki
+# list: PyAutoMemory/index.md.
 memory_wikis:
-  lensing_wiki: [lens, deflection, source reconstruction, caustic, einstein,
+  lensing: [lens, deflection, source reconstruction, caustic, einstein,
                  subhalo, substructure, time delay, cosmography, shear,
                  multipole, mass sheet, slacs, tdcosmo, h0licow, macromodel]
-  smbh_wiki: [black hole, smbh, binary, recoil, nanograv,
+  smbh: [black hole, smbh, binary, recoil, nanograv,
               gravitational wave background]
-  cti_wiki: [charge transfer, cti, trap, arctic, vis calibration]
-  methods_wiki: [bayesian, sampler, nautilus, dynesty, emcee, mcmc,
+  cti: [charge transfer, cti, trap, arctic, vis calibration]
+  methods: [bayesian, sampler, nautilus, dynesty, emcee, mcmc,
                  nested sampling, likelihood, jax, nufft, interpolat,
                  graphical model, expectation propagation, deep learning, sbi,
                  simulation based inference, probabilistic, optimis, gradient,
                  regularis]
-  galaxies_wiki: [galaxy formation, bulge, disk, morphology, mge, stellar halo,
+  galaxies: [galaxy formation, bulge, disk, morphology, mge, stellar halo,
                   ifu, kinematic, elliptical, cosmos]
 
 # Intake keyword -> target repo routing signals.
@@ -62,11 +63,13 @@ target_signals:
 
 # Feature conductor: target -> the sub-wiki consulted by default.
 target_default_wiki:
-  autolens: lensing_wiki
-  autogalaxy: galaxies_wiki
-  autofit: methods_wiki
-  autoarray: methods_wiki
-  autoconf: methods_wiki
+  autolens: lensing
+  autogalaxy: galaxies
+  autofit: methods
+  autoarray: methods
+  autoconf: methods     # legacy name — the Nerves repo was PyAutoConf
+  autonerves: methods
+  pyautonerves: methods
 
 # Refactor conductor: normalised repo -> the test dir that witnesses
 # behaviour preservation.

docs/adoption/config_surfaces.md

@@ -13,7 +13,7 @@ framework organs is domain-free logic you pull from upstream.
 | `agents/faculties/sizing/_sizing.py` | Brain | The library/workspace repo sets and aliases used to size and route tasks. |
 | `agents/conductors/intake/_intake.py` — `TARGET_SIGNALS` | Brain | Keyword → target-repo routing for raw ideas. |
 | `agents/conductors/release/` — the library tuple + `nightly.sh` | Brain | Which libraries the release path drives. |
-| `agents/faculties/memory/` — the wiki keyword map | Brain | Topic keywords → your Memory's sub-wikis. |
+| `config/policy.yaml` — `memory_wikis` + `target_default_wiki` | Brain | Topic keywords → your Memory's sub-wikis (`wiki/<domain>/`), consumed by the sizing faculty and the feature/bug conductors. |
 | `heart/readiness.py` — `DEFAULT_LIBRARIES`; `heart/checks/version_skew.py` — the workspace→(library, package) map | Heart | The release-gate repo sets. |
 | `autohands/run_all.py`, `autohands/slow_skip_check.py`, navigator maps | Hands | Which workspaces the validation pipeline runs and skips. |
 

skills/sampler_pipeline/SKILL.md

@@ -26,7 +26,7 @@ Read `agents/faculties/samplers/AGENTS.md` for the judgment tables (sampler ↔
 likelihood match, gradient/JAX constraints, initialization chaining, promotion
 criteria). If the candidate is already prototyped or promoted, resume at the
 right stage instead of starting over. The deeper science is in
-`PyAutoMemory/methods_wiki` (`hamiltonian-monte-carlo`, `gpu-nested-sampling`,
+`PyAutoMemory/wiki/methods` (`hamiltonian-monte-carlo`, `gpu-nested-sampling`,
 `initialization-chaining`, `sampler-benchmarks`) — internal use only; its
 citations never reach public user-facing output.
 
@@ -72,7 +72,7 @@ The prototype contract (what makes its row comparable):
 - If the candidate targets real likelihoods (lensing MGE / pixelization /
   interferometer), extend to `autolens_profiling` use-case runs before any
   promotion argument — the 1D Gaussian alone never justifies promotion.
-- Record durable findings in `PyAutoMemory/methods_wiki` (update
+- Record durable findings in `PyAutoMemory/wiki/methods` (update
   `concepts/sampler-benchmarks.md`).
 
 ## 3. Promote (or archive)

tests/test_memory_surfaces.py

@@ -0,0 +1,87 @@
+"""tests/test_memory_surfaces.py — the memory faculty's corpus (PyAutoBrain#239).
+
+First regression net for `agents/faculties/memory/_memory.py`. What matters:
+the root entry surfaces (index.md, reading-queue.md, bibliography/README.md,
+wiki/CLAUDE.md) are part of the corpus — they were invisible before #239 even
+though PyAutoMemory/AGENTS.md tells every agent to index them first; surface
+labels stay repo-name-first (digest()'s exit-4 logic splits on '/'); ranking
+and the no-hits path behave; and the conductors' printed PyAutoMemory pointers
+name paths that exist under the wiki/<domain>/ layout.
+"""
+
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parents[1]
+                       / "agents" / "faculties" / "memory"))
+
+import _memory  # noqa: E402
+
+
+def _memory_repo(tmp_path: Path) -> Path:
+    m = tmp_path / "PyAutoMemory"
+    (m / "wiki" / "demo" / "concepts").mkdir(parents=True)
+    (m / "bibliography").mkdir()
+    (m / "index.md").write_text("# Index\nthe entry surface mentions quasars\n")
+    (m / "reading-queue.md").write_text("# Reading queue\n\nquasar paper one\n")
+    (m / "bibliography" / "README.md").write_text("# Bib\nadding a quasar paper\n")
+    (m / "wiki" / "CLAUDE.md").write_text("# Schema\nstatus flags\n")
+    (m / "wiki" / "demo" / "concepts" / "quasars.md").write_text(
+        "---\nstatus: drafted\n---\nquasars lens light\n")
+    return m
+
+
+def test_root_entry_surfaces_are_in_the_corpus(tmp_path):
+    m = _memory_repo(tmp_path)
+    triples = list(_memory.surfaces(m, None, None))
+    root_files = {f.name for name, _, f in triples if name == "PyAutoMemory/root"}
+    assert root_files == {"index.md", "reading-queue.md", "README.md", "CLAUDE.md"}
+    # the wiki pages still arrive under their sub-wiki label
+    assert any(name == "PyAutoMemory/demo" for name, _, _ in triples)
+
+
+def test_labels_stay_repo_name_first(tmp_path):
+    m = _memory_repo(tmp_path)
+    for name, _, _ in _memory.surfaces(m, None, None):
+        assert name.split("/")[0] == "PyAutoMemory"
+
+
+def test_digest_ranks_and_cites_the_queue(tmp_path):
+    m = _memory_repo(tmp_path)
+    d = _memory.digest("quasar reading queue", m, None, None, limit=8)
+    assert d["surfaces_present"] == ["PyAutoMemory"]
+    pages = {p["page"] for p in d["pages"]}
+    assert "reading-queue.md" in pages
+    assert all(p["hits"] >= 1 and p["snippets"] for p in d["pages"])
+
+
+def test_no_hits_yields_empty_pages(tmp_path):
+    m = _memory_repo(tmp_path)
+    d = _memory.digest("nonexistent-elephant-topic", m, None, None, limit=8)
+    assert d["pages"] == [] and d["surfaces_present"] == ["PyAutoMemory"]
+
+
+def test_bib_file_itself_is_not_a_surface(tmp_path):
+    m = _memory_repo(tmp_path)
+    (m / "bibliography" / "demo.bib").write_text("@article{Quasar2020}\n")
+    files = {f.name for _, _, f in _memory.surfaces(m, None, None)}
+    assert "demo.bib" not in files  # metadata, not digest text (by design)
+
+
+def test_conductor_pointers_name_real_layout_paths(tmp_path):
+    """The feature/bug conductors print `PyAutoMemory/wiki/<d>/index.md`
+    pointers built from the policy wiki names — those names must be the
+    wiki/<domain>/ layout, and the printed shape must resolve in a real
+    checkout when one is present."""
+    sys.path.insert(0, str(Path(__file__).resolve().parents[1]
+                           / "agents" / "faculties" / "sizing"))
+    import _sizing
+    live = Path(__file__).resolve().parents[2] / "PyAutoMemory"
+    for wiki in _sizing.MEMORY_WIKIS:
+        assert not wiki.endswith("_wiki")
+        if live.is_dir():
+            assert (live / "wiki" / wiki / "index.md").is_file(), (
+                f"policy names sub-wiki '{wiki}' but "
+                f"PyAutoMemory/wiki/{wiki}/index.md does not exist")

tests/test_policy_seams.py

@@ -48,8 +48,10 @@ def test_aliases_normalise_known_mentions():
 
 def test_memory_wikis_route_science_vocabulary():
     wikis = _sizing.MEMORY_WIKIS
-    assert "lens" in wikis["lensing_wiki"]
-    assert "sampler" in wikis["methods_wiki"]
+    # Keys are the wiki/<domain>/ names (the *_wiki root layout is retired).
+    assert "lens" in wikis["lensing"]
+    assert "sampler" in wikis["methods"]
+    assert not any(k.endswith("_wiki") for k in wikis)
     assert _sizing.SCIENCE_KEYWORDS  # derived, non-empty
 
 
@@ -65,7 +67,9 @@ def test_feature_default_wiki_loads():
     sys.path.insert(0, str(BRAIN_HOME / "agents" / "conductors" / "feature"))
     import _feature
 
-    assert _feature.TARGET_DEFAULT_WIKI["autolens"] == "lensing_wiki"
+    assert _feature.TARGET_DEFAULT_WIKI["autolens"] == "lensing"
+    # The Nerves rename is covered both ways (legacy autoconf + autonerves).
+    assert _feature.TARGET_DEFAULT_WIKI["autonerves"] == "methods"
 
 
 def test_refactor_test_witness_loads():