Merge pull request 'docs+chore: audit findings + tooling (base of the remediation stack)' (#15) from audit/loop-1 into main

2026-06-16 04:52:58 +00:00
parent 5dfa6e4bae 64e3ee188c
commit fb06c6f65a
10 changed files with 1159 additions and 30 deletions
--- a/.gitignore
+++ b/.gitignore
@@ -18,6 +18,8 @@ build/

 # Environment secrets — NEVER commit
 .env
+.env.*
+!.env.example

 # Type-checker and linter caches
 .mypy_cache/
--- a/codex/sources/semanticscholar.py
+++ b/codex/sources/semanticscholar.py
@@ -96,10 +96,7 @@ def fetch_references(paper_id: str) -> list[Citation]:
        context: str | None = contexts[0] if contexts else None

        cited_id: str = (
-            external_ids.get("DOI")
-            or external_ids.get("ArXiv")
-            or cited_paper.get("paperId")
-            or ""
+            external_ids.get("DOI") or external_ids.get("ArXiv") or cited_paper.get("paperId") or ""
        )
        if cited_id:
            citations.append(Citation(citing_id=paper_id, cited_id=cited_id, context=context))
--- a/codex/synthesis.py
+++ b/codex/synthesis.py
@@ -365,11 +365,7 @@ def find_connections(
                continue
            seen_pairs.add(pair)  # type: ignore[arg-type]

-            pair_chunks = [
-                c
-                for c in chunks
-                if c.get("bibkey") in {bibkey, other}
-            ]
+            pair_chunks = [c for c in chunks if c.get("bibkey") in {bibkey, other}]
            prompt = _CONNECTION_PROMPT.format(chunks_block=_format_chunks(pair_chunks))
            raw = _safe_llm_generate(llm, prompt, settings.synthesis_llm_model)
            if not raw.strip():
--- a/codex/wiki.py
+++ b/codex/wiki.py
@@ -240,9 +240,36 @@ def _parse_claims(markdown: str) -> list[Claim]:

 _STOPWORDS: frozenset[str] = frozenset(
    {
-        "the", "a", "an", "in", "on", "of", "to", "is", "are", "was", "were",
-        "and", "or", "but", "for", "with", "this", "that", "it", "we", "they",
-        "be", "as", "at", "by", "from", "has", "have", "not", "which",
+        "the",
+        "a",
+        "an",
+        "in",
+        "on",
+        "of",
+        "to",
+        "is",
+        "are",
+        "was",
+        "were",
+        "and",
+        "or",
+        "but",
+        "for",
+        "with",
+        "this",
+        "that",
+        "it",
+        "we",
+        "they",
+        "be",
+        "as",
+        "at",
+        "by",
+        "from",
+        "has",
+        "have",
+        "not",
+        "which",
    }
 )

@@ -360,12 +387,8 @@ def _detect_conflicts(
        for bk2 in bibkeys[i + 1 :]:
            # Check if any chunk from bk1 contains a conflict keyword
            # and any chunk from bk2 also does — heuristic signal only
-            bk1_has_conflict = any(
-                _CONFLICT_KEYWORDS.search(c) for c in by_bib[bk1]
-            )
-            bk2_has_conflict = any(
-                _CONFLICT_KEYWORDS.search(c) for c in by_bib[bk2]
-            )
+            bk1_has_conflict = any(_CONFLICT_KEYWORDS.search(c) for c in by_bib[bk1])
+            bk2_has_conflict = any(_CONFLICT_KEYWORDS.search(c) for c in by_bib[bk2])
            if bk1_has_conflict and bk2_has_conflict:
                conflicts.append((slug, bk1, bk2))

--- a/docs/audit/AUDIT-2026-06-15-full-repo.md
+++ b/docs/audit/AUDIT-2026-06-15-full-repo.md
@@ -0,0 +1,616 @@
+# AUDIT — Full Repository (iterative)
+
+**Auditor:** Audit-Loop (Opus)
+**Started:** 2026-06-15
+**Scope:** entire `main` tree, audited iteratively module-by-module.
+**Companion doc:** [`AUDIT-2026-06-15-loop-1.md`](AUDIT-2026-06-15-loop-1.md) —
+the F-15 / dev-loop-1 findings (S-1…U-4). This document continues the **same
+finding-ID register** (so `C-4` here follows `C-3` there) and covers the rest of
+the repo.
+
+> Severity: **HIGH** fix before relying · **MED** fix soon · **LOW** polish ·
+> **INFO** accepted/no-action. Findings only — remediation planned separately.
+
+## Coverage tracker
+
+| Part | Area | Status |
+|------|------|--------|
+| (loop-1) | F-15 graph, db/schema, ingest, discover, mcp, provenance-scan, cli(search+graph), config(graph) | ✅ documented |
+| **A** | Grounding guard (`wiki.py` `_parse_claims`/`_run_grounding_guard`, quarantine) — shared by `synthesis.py` | ✅ documented |
+| **B** | `sources/` (arxiv, openalex, semanticscholar) | ✅ documented |
+| **C** | `parsing/` (grobid, nougat, mathpix, figures, tex) | ✅ documented |
+| **D** | `synthesis.py` orchestration (find_connections/gaps/improvements, conjectures, write_leads) | ✅ documented |
+| **E** | `wiki.py` rest (retrieve, compile_concept, cross-refs, index/log, embed) | ✅ documented |
+| **F** | `embed.py`, `quality.py`, `models.py` | ✅ documented |
+| **G** | `cli.py` rest (ingest/wiki/synthesis/provenance commands), `config.py` rest | ✅ documented |
+
+---
+
+## Part A — Grounding guard (the "zero-fact-leak" core)
+
+The guard lives in `wiki.py` (`_parse_claims:216`, `_run_grounding_guard:299`)
+and is imported by `synthesis.py:57-59`, so **one implementation gates both**
+the wiki pages and the synthesis leads. A claim is "grounded" iff the **last
+sentence** before its `[BibKey #loc]` citation contains a run of 5 consecutive
+non-stopword tokens that appears as a substring of any chunk of the same bibkey.
+
+### C-4 — Zero-claim pages bypass quarantine and publish as "compiled" · **HIGH**
+- **Evidence:** `wiki.py:730` — `if total_claims > 0 and grounding_rate < threshold:`
+  quarantines; the `else` (`:737-742`) **publishes to `wiki/` and records the
+  hash** as successfully compiled.
+- **Bug:** when the LLM output has **no parseable citations** (`total_claims == 0`,
+  so `grounding_rate == 0.0`), the `total_claims > 0` guard is false → it takes
+  the *publish* branch. The quarantine gate is skipped exactly when grounding
+  evidence is **weakest** (zero citations).
+- **Trigger paths:** (a) the LLM emits uncited prose; (b) **LLM outage** — the
+  module header guarantees "LLM unavailable → empty ConceptPage, no crash"
+  (`wiki.py:13-14`), so a transient Ollama outage writes near-empty stub pages to
+  `wiki/`, marks them compiled, and stores their hash. On the next
+  `changed_only` run the hash matches → the stub is **skipped forever** until a
+  forced recompile. A research wiki silently fills with empty/uncited pages.
+
+### C-5 — "Every claim is grounded" holds only for the last cited sentence · **MED**
+- **Evidence:** `_CLAIM_RE` (`wiki.py:211-213`) captures `text = [^\[]+?` (the
+  span back to the previous `]`/start); `_last_sentence` (`:279-291`) then keeps
+  only the final sentence; `_run_grounding_guard` checks that sentence only.
+- **Two holes in the guarantee:**
+  1. **Paragraph passengers** — any sentence that is *not* the last one before a
+     citation is never grounding-checked and never `⚠`-marked. "Hallucination one.
+     Hallucination two. Grounded tail [Bib]." passes with only the tail verified.
+  2. **Uncited text is invisible** — a sentence with no `[BibKey]` is not a Claim
+     at all (`_parse_claims` only yields cited spans), so it is never grounded
+     nor marked. The LLM can emit fully uncited assertions into a published page.
+- **Impact:** the effective invariant is "*the last sentence immediately before
+  each citation has lexical overlap with the cited paper*" — materially weaker
+  than the N-02 "zero-fact-leak enforcement" framing. The module **header is
+  honest** ("Substring-Match MVP", `wiki.py:4,13`); the commit/ADR language is
+  not. Recommend aligning the claim or strengthening the guard (check every
+  sentence; flag uncited sentences).
+
+### C-6 — Substring match does not enforce the documented "5 consecutive tokens" · **LOW**
+- **Evidence:** `wiki.py:347-349` — `gram = " ".join(content[i:i+5])`;
+  `if any(gram in src for src in sources)` where `src` is the space-joined
+  content-word string of the chunk.
+- **Issue:** a *substring* test on the joined string lets the first/last gram
+  token match as a **prefix/suffix of a longer chunk word** (e.g. gram
+  `"cat dog …"` substring-matches inside `"scat dogma …"`). The docstring claims
+  matching "as 5 consecutive non-stopword **tokens**" — true sequence matching
+  would compare token lists with a sliding window (or add boundary sentinels).
+  Low real-world impact at 5-gram length, but the claimed invariant is not what
+  the code enforces.
+
+### R-1 — Tokenizer brittleness causes false-negative over-quarantine · **MED**
+- **Evidence:** `_content_words` (`wiki.py:294-296`) lowercases and `.split()`s on
+  whitespace only — **no punctuation stripping**. `_last_sentence` splits on
+  `.!?` (`:276`).
+- **Impact:** punctuation stays glued to tokens, so a faithful paraphrase grounds
+  only on *exact* surface form — `"volume,"` ≠ `"volume"`, `"L(gamma1)"` must
+  match character-for-character, and a trailing decimal like `"V = 1.5"` makes
+  `_last_sentence` return the fragment `"5"` (split at the `.`), which then has
+  `< 5` content words → **forced ungrounded** (`:339-344`). Math-heavy and terse
+  claims are systematically marked ungrounded even when verbatim, inflating the
+  ungrounded ratio and pushing real pages into quarantine (and into C-4's
+  blind spot when it drives `total_claims` interactions). The passing test
+  `test_grounding_guard_…` only succeeds because its fixture reproduces the chunk
+  text *exactly*.
+
+### R-2 — Reference-list filter can drop theorem-dense chunks · **LOW**
+- **Evidence:** `_retrieve_chunks` (`wiki.py:188-199`) skips a chunk when
+  `>60 %` of its lines match `^\s*(\[\d+\]|[A-Z][a-z]+,?\s+[A-Z]\.)`.
+- **Issue:** the second alternative (`[A-Z][a-z]+,?\s+[A-Z]\.`) also matches prose
+  like `"Theorem A."`, `"Lemma B."`, `"Section C."`. A chunk formatted as a list
+  of theorems/lemmas can exceed 60 % and be discarded as a "reference list" —
+  silently dropping exactly the mathematical content the wiki wants.
+
+### R-3 — Conflict detection has near-zero precision · **LOW**
+- **Evidence:** `_detect_conflicts` (`wiki.py:367-…`) flags a pair `(bk1, bk2)` if
+  **each** has *any* chunk containing a hedging keyword
+  (`but|however|in contrast|…`, `_CONFLICT_KEYWORDS:361-364`).
+- **Issue:** it never checks the keywords relate to a shared topic or to each
+  other. In a corpus where most papers contain "however"/"but" somewhere, this
+  flags ~O(n²) pairs as "conflicting" — noise surfaced in `CompileReport`.
+  Labeled "heuristic signal only", but as-is it is not actionable.
+
+**Part A net:** 1 HIGH (C-4), 2 MED (C-5, R-1), 3 LOW (C-6, R-2, R-3). The guard
+is a reasonable lexical MVP, but (a) it has a concrete publish-bypass bug for
+zero-citation output, and (b) the "zero-fact-leak" framing overstates what a
+last-sentence substring check can guarantee.
+
+---
+
+## Part B — `sources/` (external API clients)
+
+Solid foundations: all three clients set httpx timeouts; OpenAlex and S2 retry
+429/5xx with exponential back-off (tenacity); S2 enforces a ≥1 req/s floor with a
+monotonic clock under a lock. The findings are about **ID format consistency**
+and a couple of dead/inert paths.
+
+### C-7 — `papers.id` is stored in OpenAlex *URL* form, breaking the bare-ID contract · **HIGH**
+- **Evidence:** `openalex.py:94` — `_map_paper` sets
+  `id = data.get("doi") or data.get("id") or ""`. Real OpenAlex returns `doi` as
+  a **full URL** (`https://doi.org/10.x`) and `id` as `https://openalex.org/W…`
+  (the repo's own test confirms the URL form for `openalex_id`,
+  `test_openalex.py:92`). `ingest.py:82` takes `paper.id` from this unchanged and
+  upserts it as `papers.id` (`ingest.py:138`).
+- **Why it's currently masked:** the test fixture `_SAMPLE_WORK["doi"]` is a
+  **bare** DOI `"10.1145/3592430"` (`test_openalex.py:18`) — not the URL form real
+  OpenAlex emits — and **no test asserts `paper.id`** at all. So the suite is
+  green while production stores `papers.id = "https://doi.org/10.1145/3592430"`.
+- **Impact:**
+  1. **Bare-ID lookups fail.** `ingest_all.sh` ingests `"10.1145/1964921.1964997"`,
+     but `papers.id` becomes the URL; `codex graph related 10.1145/…`,
+     `discover citing 10.1145/…`, search-by-id, etc. find nothing.
+  2. **Second ID-mismatch axis (compounds C-1).** `cited_id` from the S2/GROBID
+     paths is a **bare** DOI/arXiv (`semanticscholar.py:98-99`,
+     `ingest.py:187`), so a bare-DOI reference can never equal a URL-form
+     `papers.id` — bare-DOI cross-citations stay "dangling" regardless of the
+     F-15 W-ID resolution (which only matches `papers.openalex_id`, the URL form).
+- **Action:** confirm with `SELECT id FROM papers LIMIT 5` on the Jetson DB; if
+  URL-form, normalize `papers.id` to a canonical bare form at ingest (and the
+  `cited_id` writers to match), then add a representative fixture + `paper.id`
+  assertion (see T-3).
+
+### T-3 — OpenAlex test fixture is unrepresentative and skips `paper.id` · **MED**
+- **Evidence:** `_SAMPLE_WORK["doi"] = "10.1145/3592430"` (bare) vs the real
+  `https://doi.org/…` URL form; `test_fetch_paper_*` assert title/year/authors/
+  abstract/`openalex_id` but **never** `paper.id`.
+- **Impact:** the id-mapping — the exact surface of C-7 — is untested, and the
+  fixture actively hides the URL-form behaviour. Any normalization fix needs this
+  closed first or it will regress invisibly.
+
+### R-5 — S2 `fetch_references` breaks on legacy arXiv IDs containing `/` · **LOW**
+- **Evidence:** `semanticscholar.py:80` interpolates `paper_id` into the path
+  `…/paper/{paper_id}/references`. The ingest fallback passes `arXiv:math/0603097`
+  (`ingest.py:91`) — the embedded `/` yields a malformed path → 404 →
+  `fetch_references` silently returns `[]`. Legacy-arXiv papers that miss the
+  OpenAlex path get **zero references** with no error.
+
+### R-4 — `arxiv.fetch_source` no-op try/except + no retry · **LOW**
+- **Evidence:** `arxiv.py:41-44` — `try: httpx.get(...) except httpx.RequestError: raise`
+  catches and immediately re-raises, doing nothing. Unlike OpenAlex/S2, the arXiv
+  client has **no** retry/back-off, so a transient network blip aborts ingest.
+  Minor resilience asymmetry.
+
+### C-9 — `semanticscholar.fetch_recommendations` is unreachable app code · **LOW**
+- **Evidence:** referenced only by `tests/sources/test_semanticscholar.py`; no
+  `codex/` module or CLI calls it (`grep fetch_recommendations`). Tested but
+  never wired — dead feature surface (decide: wire a `discover recommend`
+  command, or drop).
+
+**Part B net:** 1 HIGH (C-7, ID-form contract — the big one; confirm on live DB),
+1 MED (T-3, fixture masks it), 3 LOW (R-4, R-5, C-9). The retry/timeout/rate-limit
+hygiene is genuinely good; the weakness is the **un-normalized, heterogeneous ID
+space** (URL vs bare, W-ID vs DOI vs arXiv vs S2-hash) threaded through `papers.id`
+and `citations.cited_id` — the same root cause behind C-1 and the F-15 ID dance.
+
+---
+
+## Part C — `parsing/` (tex, grobid, nougat, mathpix, figures)
+
+> **Scope reality:** the live corpus is ingested entirely as **`.txt`**
+> (`ingest_all.sh` maps every paper to a `*.txt` file), so the `.tex`-clean,
+> GROBID, Nougat, Mathpix, and Figures paths are **latent** — only
+> `tex.chunk_text` + `quality.*` run on the live data. The findings below are
+> real but mostly fire only once `.tex`/`.pdf --rich` ingest is used.
+>
+> **Not a finding:** the broad `except Exception` in `mathpix.py:109,151` and
+> `figures.py:125,152` are *justified* per-item graceful degradation
+> (`fitz.open` fails → return empty; pix2tex/image fails → skip that item), not
+> swallowed bugs.
+
+### C-10 — Formulas/figures key on the PDF *filename*, not `paper.id` · **MED** (latent, `.pdf --rich`)
+- **Evidence:** `figures.py:119` and the mathpix extractor set
+  `paper_id = Path(pdf_path).stem`. `ingest.py:289,303` insert
+  `f.paper_id` / `fig.paper_id` **unchanged** into `formulas`/`figures`, whose
+  schema declares `paper_id TEXT REFERENCES papers(id)` (`schema.sql:97,115`).
+- **Impact:** the filename stem (e.g. `springborn-2008-weighted-delaunay…`) will
+  not equal `papers.id` (an OpenAlex URL/bare id — see C-7), so the FK insert
+  **fails and aborts rich ingest**, or (without FK enforcement) writes orphaned
+  rows. Identity is derived from the *file*, not the *paper*. Fix: pass `paper.id`
+  into the extractors, or have ingest overwrite `f.paper_id = paper.id` before insert.
+
+### R-7 — `_clean_latex` misses `\(…\)` / `\[…\]` math + strips all math from chunks · **LOW** (`.tex` path)
+- **Evidence:** `tex.py:21-56` strips `%comment`, `$…$`, `$$…$$`,
+  `equation`/`align` envs, `\cite/\label/\ref` — but **not** the `\(…\)` / `\[…\]`
+  math delimiters. Papers using those leave raw `\( … \)` markup in the chunk text.
+- **Impact:** (a) leftover math markup pollutes embeddings; (b) *all* math is
+  removed from chunk text, so chunk-level semantic search can never match formula
+  content — it relies entirely on the separate formula FTS (F-09), which is itself
+  latent here. Acceptable as a separation-of-concerns choice, but worth stating.
+
+### R-6 — PDF-path external calls are unwrapped / un-retried · **LOW** (`.pdf` path)
+- **Evidence:** `ingest.py:185` calls `extract_references()` outside any
+  `try`; GROBID (`grobid.py:54,133`) and the mathpix HTTP calls `raise_for_status`
+  with **no** retry/back-off (unlike OpenAlex/S2). A transient GROBID/Mathpix
+  5xx aborts the whole PDF ingest.
+- **INFO:** `grobid.py:56` uses stdlib `ET.fromstring` on the server's XML.
+  Fine for a trusted self-hosted GROBID; switch to `defusedxml` if it is ever
+  pointed at an untrusted endpoint.
+
+**Part C net:** 1 MED (C-10), 2 LOW (R-6, R-7), all **latent** for the current
+`.txt` corpus. `nougat.py` was skimmed only (54 lines, `.pdf`-only wrapper) — no
+deep review yet; flagged in the ledger. Parsing hygiene is otherwise good.
+
+> **Request (a) — second audit pass — is now complete:** §8 of the loop-1 doc
+> (grounding guard ✓ Part A, `sources/` ✓ Part B, `parsing/` ✓ Part C) is closed.
+> Remaining whole-repo parts D–G are the broader sweep.
+
+---
+
+## Part D — `synthesis.py` orchestration
+
+The lead pipeline (connections → gaps → improvements → conjectures) reuses the
+Part-A grounding guard for the three grounded kinds and degrades gracefully on
+LLM/DB failure (`_safe_llm_generate`/`_list_paper_titles` swallow + log). The
+conjecture invariants are genuinely solid (status hard-set `"unverified"`,
+mandatory falsification path or the lead is dropped, written to `conjectures/`
+only). Two real defects and a doc contradiction:
+
+### C-11 — Lead-ID collision silently overwrites synthesis output · **HIGH**
+- **Evidence:** every stage restarts its counter at `seq = 1`
+  (`synthesis.py:341` connections, `:415` gaps, `:530` improvements,
+  `:634` conjectures) and ids are `L-0001, L-0002, …` (`_make_lead_id`). The CLI
+  concatenates the grounded kinds — `all_leads = connections + gaps + improvements`
+  (`cli.py:421`) — and `write_leads` routes **all** grounded kinds to the *same*
+  `grounded/<id>.md` (`synthesis.py:783-796`), writing in list order.
+- **Bug:** connection `L-0001`, gap `L-0001`, improvement `L-0001` all target
+  `grounded/L-0001.md`; later writes overwrite earlier → **improvements clobber
+  gaps clobber connections** at each number. Surviving files =
+  `max(#connections, #gaps, #improvements)`, **not** the sum. With 5/3/4 leads you
+  get 5 files, not 12; connections are lost first. Silent data loss of the
+  synthesis engine's core output, masked because "files do get written".
+- **Fix direction:** namespace ids by kind (`L-C-0001` / `L-G-0001` / `L-I-0001`)
+  or use a single shared counter across stages.
+
+### R-8 — Default-topic gap detection is a false-positive generator · **MED**
+- **Evidence:** `find_gaps` defaults `probe_topics` to paper **titles** when
+  `topics is None` (`synthesis.py:418-420`); a title query retrieves mostly its
+  *own* paper's chunks → `bibkeys_for_topic` ≈ 1, which is `< synthesis_gap_min_coverage`
+  (`:450-453`) → the topic is flagged as a gap and the LLM is asked to articulate
+  one.
+- **Impact:** nearly every ingested paper becomes a spurious "Gap: '<title>'
+  covered by only 1 bibkey(s)" candidate — the LLM invents a gap for material that
+  is actually well covered. The grounding guard catches some, but the default
+  signal is mostly noise. (User-supplied `--topics` avoid this.)
+
+### D-4 — `write_leads` vs `_update_index` docstrings contradict each other · **LOW**
+- **Evidence:** `write_leads` (`synthesis.py:762-765`) claims it "maintains an
+  **append-only** `INDEX.md`: a previously-recorded id is kept on subsequent runs",
+  but `_update_index` (`:802-803`) states — and implements — a "**full rebuild**,
+  not append-only" by globbing the on-disk dirs.
+- **Impact:** there is no append-only / id-retention logic; the index purely
+  reflects current disk state. The guarantee in the `write_leads` docstring is
+  false (and interacts with C-11: overwritten ids simply vanish from the index).
+
+**Part D net:** 1 HIGH (C-11, silent lead overwrite), 1 MED (R-8, gap noise),
+1 LOW (D-4, doc contradiction). The conjecture path is exemplary; the grounded
+path's persistence layer is where it breaks.
+
+---
+
+## Part E — `wiki.py` rest (compile, cross-refs, state, index/log)
+
+Good: `yaml.safe_load`, deterministic `_chunk_hash` (sorted by chunk id) driving a
+sound incremental-skip, append-only `log.md`, inline-code protection during
+cross-ref injection. The issues are math-corruption, a dead step, and a marking
+order bug.
+
+### R-9 — Cross-ref injection corrupts LaTeX math · **MED**
+- **Evidence:** `_inject_cross_refs` (`wiki.py:426`) protects only inline-code
+  spans (`_INLINE_CODE_RE`); it then runs whole-word, case-insensitive title/alias
+  → `[[slug]]` replacement over everything else (`:429-438`). **Math spans
+  (`$…$`, `$$…$$`, `\(…\)`) are not protected.**
+- **Impact:** the synthesis prompt explicitly asks for LaTeX `$ … $`
+  (`wiki.py:495`), so bodies contain math. A concept title/alias that appears
+  inside a formula (e.g. a concept named "Energy" inside `$E_{\text{Energy}}$`,
+  or a single-word alias) is rewritten to `[[slug]]`, corrupting the LaTeX. The
+  risk scales with how common the concept titles/aliases are as math tokens.
+
+### C-12 — `_try_embed_formulas` is a no-op that still costs a DB round-trip per concept · **LOW**
+- **Evidence:** `wiki.py:623-637` opens a connection, queries
+  `information_schema.tables` for `formulas`, and returns — the actual embed is
+  `"(Future: …)"`. Called unconditionally per concept (`compile_concept:609`).
+- **Impact:** step 4 of `compile_concept`'s docstring ("Embed formula chunks")
+  does nothing, and each concept compile pays an extra DB connection + query for
+  no effect. Placeholder masquerading as a feature.
+
+### C-13 — Ungrounded-claim marking can miss claims mutated by cross-ref injection · **LOW**
+- **Evidence:** order in `compile_concept` — claims parsed from `raw_output`
+  (`:602-603`), then `raw_output` is rewritten by `_inject_cross_refs` (`:606`),
+  then `_render_page_markdown` (`:612`) re-finds the *original* `claim.text` by
+  regex in the **injected** body (`:544-552`).
+- **Impact:** if an ungrounded claim's text contained a term that got replaced
+  with `[[slug]]`, the original text no longer matches → the `⚠` mark is silently
+  not applied. (Also note: wiki's marking lacks the `(?<!⚠ )` guard that
+  synthesis's `_mark_ungrounded` has — two divergent mark implementations.)
+
+### R-10 — Non-atomic compile-state write · **LOW**
+- **Evidence:** `_save_compile_state` (`wiki.py:475-477`) writes
+  `.compile-state.json` in place. A crash mid-write corrupts it; `_load`
+  then catches `JSONDecodeError` → `{}` → every concept recompiles. Graceful but
+  loses incrementality. Write to a temp file + `os.replace` for atomicity.
+
+### R-11 — `load_concepts` raises `KeyError` on malformed YAML · **LOW**
+- **Evidence:** `entry["slug"]` / `entry["title"]` (`wiki.py:135-136`) are
+  unguarded; a concepts.yaml entry missing a key crashes the whole compile with a
+  bare `KeyError` rather than a actionable message.
+
+> **Minor:** `compile_all` builds the index from `state.keys()` (`:745`), which
+> accumulates slugs across runs — a concept removed from `concepts.yaml` lingers
+> in `index.md`/state until manually pruned. Low.
+
+**Part E net:** 1 MED (R-9, math corruption), 4 LOW (C-12, C-13, R-10, R-11). The
+incremental-compile machinery is sound; the rendering/cross-ref layer has the rough
+edges. `OllamaClient.generate` (no retry, 120 s timeout, `raise` on error) feeds
+directly into **C-4** (its exception → empty page → published), so C-4's blast
+radius includes every transient Ollama hiccup.
+
+---
+
+## Part F — `embed.py`, `quality.py`, `models.py`
+
+The best-built corner of the repo. `embed.py` L2-normalises correctly (cosine ⇒
+dot product), guards empty input and zero-norm rows, and caches a process
+singleton. `quality.py` thresholds are all config-driven, `run_quality_pass` is
+CLI-wired (`cli.py:513-517`) and atomic (single commit). `models.py` dataclasses
+mirror the schema cleanly. Two dead/low-signal items and one corroboration:
+
+### C-14 — The sparse-embedding ("hybrid") path is dead code · **LOW**
+- **Evidence:** `Embedder.encode_sparse` / `encode` / `_coerce_sparse` are called
+  nowhere in `codex/` (grep for `encode_sparse` / `.encode(` finds only
+  `str.encode` in `wiki.py:455`). Only `encode_dense` is wired (ingest, search,
+  wiki). There is no sparse column in the schema.
+- **Impact:** the `embed.py` header and ADR-0002 advertise "**Hybrid dense +
+  sparse** embeddings via BGE-M3" as a headline, but the sparse half is
+  unreachable — the actual "hybrid" elsewhere is dense + Postgres FTS. Either wire
+  sparse retrieval or drop the capability + correct the docs.
+
+### R-12 — `classify_section` is low-signal as fed · **LOW**
+- **Evidence:** `classify_section` matches section headers in the **first 200
+  chars** (`quality.py:103-106`), but `tex.chunk_text` produces overlapping
+  *word-window* chunks that almost never begin at a header. So most chunks fall
+  through to `"body"` (or the DOI-density `"bibliography"` fallback).
+- **Impact:** the `chunks.section` column is mostly `"body"` — limited value for
+  any section-aware retrieval that depends on it. Not wrong, just weak signal
+  given the chunker.
+
+### Corroboration of C-7 (not a new finding)
+- `models.Paper.id` docstring (`models.py:20-21`) states the canonical id is "an
+  arXiv ID … **or a DOI (e.g. `10.1145/3592430`)**" — the **bare** form. Production
+  fills it from OpenAlex in **URL** form (C-7), so the code violates its own model
+  contract. This makes C-7 a documented-contract breach, not a judgement call.
+
+**Part F net:** 2 LOW (C-14, R-12) + a C-7 corroboration. These three modules are
+clean; no correctness defects.
+
+---
+
+## Part G — `cli.py` rest + `config.py`
+
+Most CLI commands are thin Typer wrappers delegating to already-audited modules
+(`ingest`, `search`, `discover`, `wiki`, `synthesis`, `provenance`, `quality`,
+`graph`). Settings are pydantic-validated with sensible bounds (`gt=0`, `0..1`).
+Findings are config-coupling and output hygiene.
+
+### C-15 — `embedding_dim` is a footgun decoupled from the hardcoded schema · **MED**
+- **Evidence:** `embedding_dim` is configurable (`config.py:67-74`, default 1024)
+  and `.env.example`/`schema.sql:7` invite other models (Qwen3 1024, **Jina v4
+  2048**). But `schema.sql` hardcodes `vector(1024)` on every embedding column
+  (`:25,38,102,119`), and `Embedder` derives the real output dim from the *model*,
+  using the `dim` setting only for empty-input zero-vectors (`embed.py:50,89`).
+- **Impact:** set `EMBEDDING_DIM=2048` (or pick a non-1024 model) and inserts
+  fail on a pgvector dimension mismatch — silently for empty abstracts
+  (`(1,2048)` zero-vector into `vector(1024)`), structurally for real vectors.
+  Nothing couples the knob to the DDL; the setting reads as supported but is not.
+
+### R-13 — `export_bib` does no BibTeX escaping and forces `@article` · **LOW**
+- **Evidence:** `provenance.py:174-181` interpolates `title`/`author`/`year` raw
+  into `@article{…}`; no escaping/brace-balancing of BibTeX specials
+  (`{ } \ % # $ _ &`).
+- **Impact:** a title with an unbalanced brace or a stray special char produces a
+  malformed `.bib` entry that breaks Doxygen/BibTeX. Also every paper is emitted
+  as `@article` although the corpus includes theses/books/chapters
+  (`ingest_all.sh`) — bibliographically wrong (functional for `@cite` keys though).
+
+### S-4 — Secrets stored as plain `str`, not `SecretStr` · **LOW**
+- **Evidence:** `database_url`, `mathpix_app_key`, `mcp_auth_token`
+  (`config.py:29,148,249`) are plain `str`. Any `repr(Settings())` or debug log of
+  the settings object prints the DB password / API key / bearer token in clear.
+- **Impact:** log/exception-dump leakage surface (complements S-1). Use
+  `pydantic.SecretStr` for credential fields.
+
+### R-14 — `mcp_transport` enum not enforced · **LOW**
+- **Evidence:** `config.py:229-236` documents "stdio or http" but no validator;
+  `main()` (`mcp_server.py:323`) treats anything non-`http` as stdio. A typo
+  (`htttp`) silently starts stdio instead of failing loudly.
+
+### D-5 — Inconsistent `validation_alias` on `nougat_url` only · **LOW**
+- **Evidence:** `nougat_url` adds `AliasChoices("NOUGAT_URL", "nougat_url")`
+  (`config.py:47`) that no other field has — redundant under
+  `case_sensitive=False`, and the inconsistency invites confusion.
+
+> **Corroboration of C-7:** `cli.py:45` ingest does **no** id normalization and
+> echoes `result.paper_id` (the URL-form id) straight back — no normalization
+> layer exists at any tier (source → ingest → CLI).
+
+**Part G net:** 1 MED (C-15, dim/schema footgun), 4 LOW (R-13, S-4, R-14, D-5).
+CLI/config are otherwise solid and well-documented.
+
+---
+
+# Executive Summary — whole-repo audit complete (loop-1 + Parts A–G)
+
+**Tally: 45 findings — 8 HIGH, 11 MED, 26 LOW/INFO** (after live-DB verification:
+**7 HIGH, 12 MED** — M-1 reclassified, see the verification section at the bottom).
+Test suite 330 green, `ruff`/`mypy` clean. The codebase is well-structured and idiomatic; the defects
+cluster into five root themes, not random scatter.
+
+### The 8 HIGH findings
+| ID | Theme | One-liner |
+|----|-------|-----------|
+| S-1 | secrets | live DB password in un-ignored `.env.jetson-ingest` (mitigated this session) |
+| M-1 | migration | `paper_identifiers` has no migration path; `schema.sql` not re-applyable, `apply_schema` never called |
+| C-1 | identity | `discover.py` keeps the ID-mismatch bug F-15 fixed in `graph.py` (over-reports ingested papers as leads) |
+| C-7 | identity | `papers.id` stored in OpenAlex **URL** form, violating the bare-ID contract (breaks lookups + cross-citation match) |
+| C-4 | silent failure | zero-citation/LLM-outage wiki pages **bypass quarantine** and publish as "compiled" |
+| C-11 | silent data loss | synthesis leads collide on `L-000N` across stages → improvements overwrite gaps overwrite connections |
+| D-1 | doc integrity | ADR-F15 GO rests on spike numbers measured **before** the shipped resolution fix |
+| T-1 | verification | the F-15 resolution SQL (the substance of the last 2 commits) has **zero** real-DB coverage |
+
+### Five root-cause themes
+1. **Un-normalized identity (the dominant theme).** C-1, C-7, C-10, M-1, T-3,
+   and most of the F-15 ID dance are one problem: a paper's identity exists in
+   many un-reconciled forms (bare vs URL · DOI vs arXiv vs OpenAlex-W vs S2-hash
+   vs PDF-filename) with **no canonical normalization layer**. One normalization
+   boundary at ingest would retire ~5 HIGH/MED findings at once.
+2. **"Green but unverified."** T-1/T-2/T-3 — the most error-prone code (resolution
+   SQL, migrations, id-mapping) is mocked or tested with unrepresentative fixtures,
+   so real bugs (C-4, C-7, C-11) sit behind 330 passing tests.
+3. **Silent failure / data loss under degradation.** C-4 (publish empty), C-11
+   (overwrite leads), R-1 (over-quarantine) corrupt *outputs* instead of erroring
+   — the worst failure mode for a research tool whose value is trust.
+4. **Overstated guarantees.** D-1 (stale ADR), D-2/D-4 (doc↔code drift), C-5
+   ("zero-fact-leak" ≫ a last-sentence substring check), C-14 ("hybrid" sparse is
+   dead) — the docs/commit-language claim more than the code delivers.
+5. **Latent rich-path debt.** C-10, R-6, R-7 are real but dormant because the live
+   corpus is `.txt`; they detonate on the first `.pdf --rich` ingest.
+
+### Suggested remediation sequencing (themes, not a plan — planning deferred)
+- **First:** S-1 (rotate credential — secret already protected), M-1 (migration
+  path), C-7+C-1 (one identity-normalization layer), T-1 (real-DB test) — these
+  unblock trust in the live corpus.
+- **Then:** C-4 + C-11 (stop silent output corruption), D-1 (re-validate ADR-F15).
+- **Then:** MED batch (R-1, R-8, R-9, C-15, C-10) and the LOW polish.
+
+### Coverage honesty
+Every `codex/` module was read. After the follow-up pass **all are deep-reviewed**
+(`nougat.py` included — see Runtime Validation). Runtime state that was initially
+*inferred from code* (C-7 id-form, M-1 table existence, and the C-4/C-11
+behaviours) has since been **empirically confirmed** — see the *Live DB
+Verification* and *Runtime Validation* sections below.
+
+---
+
+# Live DB Verification — 2026-06-15 (read-only, via existing SSH tunnel)
+
+Ran read-only queries against the live Jetson corpus (`papers=36`,
+`citations=811`). This **confirms the two identity HIGHs with hard numbers** and
+**corrects an over-statement in M-1**.
+
+### C-7 — CONFIRMED ✅ (severity stands: HIGH)
+`papers.id` form distribution: **`doi_url=35`, `bare_arxiv=1`, everything else 0**
+(of 36). Samples: `https://doi.org/10.48550/arxiv.math/0603097`,
+`https://doi.org/10.48550/arxiv.2310.17529`. Even arXiv papers are stored as their
+arXiv-DOI **URL**, never the bare `math/0603097` / `2301.x` the schema, `models.Paper`,
+and `ingest_all.sh` assume. The single bare id is the one paper OpenAlex 404'd
+(fell to `ingest.py:94`). → bare-ID lookups fail for 35/36 papers. Confirmed.
+
+### C-1 — CONFIRMED ✅ with impact (severity stands: HIGH)
+- `citations.cited_id` is uniformly **`oa_url=811`** (all OpenAlex URLs) — the live
+  corpus was ingested via the OpenAlex path, so the heterogeneous bare-DOI/S2 mix
+  theorised in Part B is **not present here** (it would appear only via S2/GROBID,
+  i.e. `.pdf`/fallback ingest).
+- **`discover.py` wrongly flags 13 already-ingested papers as "dangling / not
+  ingested"** (their `cited_id` is an OpenAlex URL whose paper *is* in the corpus
+  via `openalex_id`). Meanwhile **`graph.py` correctly resolves 57 cross-citation
+  edges** via the `openalex_id` JOIN. The two "dangling" implementations
+  measurably disagree — exactly the C-1 divergence, now quantified.
+
+### M-1 — REFINED ⤵ (downgrade acute severity; structural finding stands)
+`paper_identifiers` **exists on the live DB with 39 rows** — the catastrophic
+"graph JOIN crashes with relation-does-not-exist" scenario **does not occur**;
+the table was applied (manually or via DB re-create). The *structural* gap stands:
+`apply_schema` is still non-idempotent and never invoked, so the **next** schema
+change has no migration path — but the acute "live graph is broken" framing was
+**over-stated**. Re-rank M-1 from HIGH → **MED** (latent migration debt, not an
+active outage).
+
+### C-2 — PARTIALLY REFUTED ⤵
+`paper_identifiers` has **39 rows vs ~35 papers-with-openalex_id**, so ~4 papers
+carry a diverging alias → the `pi` JOIN is **not** fully inert (it resolves a
+handful of cases the `p` JOIN can't). Keep as LOW/INFO: the table earns its keep
+for a few papers, though the cost/benefit at this corpus size is still marginal.
+
+### Net adjustment to the tally
+HIGH **8 → 7** (M-1 → MED). C-7 and C-1 are now **confirmed**, not inferred — they
+should lead remediation. The dominant root theme (**un-normalized identity**) is
+empirically validated: 35/36 ids in the "wrong" form, 13 papers mis-classified by
+the older code path, 57 edges rescued by the newer one.
+
+---
+
+# Runtime Validation — 2026-06-15 (C-4 and C-11 reproduced end-to-end)
+
+The two HIGH findings flagged "code-certain, runtime-confirmable" were exercised
+against the **real** code paths (injected fake LLM; no Ollama, no DB needed).
+
+### C-11 — CONFIRMED ✅ (lead-ID collision → silent loss; stays HIGH)
+Reproduced the exact CLI assembly (`cli.py:421` `connections + gaps +
+improvements`, each stage numbered from `L-0001`): **4 distinct leads** (2
+connection, 1 gap, 1 improvement) → `write_leads` produced **2 files**.
+`grounded/L-0001.md` survived as the **improvement** ("IMPROVEMENT-one"); the
+connection and gap `L-0001` were silently overwritten. **2 of 4 leads lost.**
+
+### C-4 — CONFIRMED ✅ (zero-claim page bypasses quarantine; stays HIGH)
+Drove the real `compile_all` with a fake LLM returning confident **uncited** prose
+(→ 0 parseable claims). With `wiki_min_grounding_rate = 0.5` and an actual
+grounding rate of `0.0`, the page was **published to `wiki/` and marked
+compiled** (`report.compiled=['test-concept']`, `report.quarantined=[]`) — not
+quarantined to `wiki/draft/`. The `total_claims > 0` guard (`wiki.py:730`) is the
+exact bypass; the bug fires precisely when grounding evidence is weakest.
+
+### `nougat.py` — now fully reviewed (no new finding)
+Folds into **R-6** (PDF-path resilience). Detail: `pdf_to_markdown` retries **only**
+`httpx.ConnectError` with `wait_fixed(0)` (no 5xx retry, no back-off) and is
+unwrapped at `ingest.py:183` — a Nougat 5xx aborts PDF ingest.
+
+**Closure:** every `codex/` module is deep-reviewed; the two runtime-confirmable
+HIGHs are empirically reproduced. The **audit / finding-identification pass is
+complete**. Remediation planning is the open next phase.
+
+---
+
+# Migration Incident — 2026-06-15 (M-1 confirmed live, new privilege dimension)
+
+While running the C-7 re-ingest migration (`infra/reingest_canonical_ids.sh`),
+M-1 manifested exactly as predicted — and exposed a dimension the static audit
+did **not** capture.
+
+**What happened.** The helper ran `TRUNCATE papers CASCADE` (corpus wiped to 0),
+then `ingest_all.sh` failed on the first chunk insert:
+`UndefinedColumn: column "section" of relation "chunks" does not exist`. The live
+`chunks` table predates the F-16 `section` column, and nothing had ever applied
+that DDL to the live DB (M-1: `apply_schema` is never called; `schema.sql` is not
+re-applyable). The ingest code was ahead of the live schema.
+
+**New finding — privilege separation (extends M-1).** The fix
+`ALTER TABLE chunks ADD COLUMN IF NOT EXISTS section TEXT` then failed with
+`InsufficientPrivilege: must be owner of table chunks`. The live ownership:
+
+| table | owner |
+|-------|-------|
+| papers, chunks, citations, code_links, formulas, figures | `postgres` |
+| paper_identifiers | `researcher` (app user) |
+
+The app role `researcher` has DML + `TRUNCATE` but **not** DDL on the
+postgres-owned core tables. So a `codex migrate` / `apply_schema` invoked with the
+application's `DATABASE_URL` would fail on every `ALTER` / `CREATE INDEX` against
+those tables. **Implication for the M-1 fix:** making `schema.sql` idempotent is
+necessary but *not sufficient* — the migration must run under a privileged role
+(owner/superuser), via a separate admin connection, not the least-privilege app
+user. This is now part of M-1's remediation scope.
+
+**Consequences observed.**
+- Corpus left wiped (papers = 0) until the owner adds the column and re-ingest is
+  re-run — a destructive-then-fail sequence.
+- Helper hardened (`fix(infra)` on `fix/audit-wave-2`): a **preflight** verifies
+  `chunks.section` exists *before* the `TRUNCATE`, aborting early with the
+  owner-level `ALTER` to run, so a schema gap can no longer cost the corpus.
+
+**Unblock (owner action):**
+`sudo -u postgres psql -d papers -c "ALTER TABLE chunks ADD COLUMN IF NOT EXISTS section TEXT;"`
+then re-run `bash ingest_all.sh`.
+
+This incident is the strongest possible validation of M-1 and should anchor its
+Wave-3 fix: **idempotent `schema.sql` + a `codex migrate` command that connects as
+a privileged role**, plus a documented owner/grant model for the app user.
--- a/docs/audit/AUDIT-2026-06-15-loop-1.md
+++ b/docs/audit/AUDIT-2026-06-15-loop-1.md
@@ -0,0 +1,227 @@
+# AUDIT — Development Loop 1 (main branch)
+
+**Auditor:** Audit-Loop (Opus)
+**Date:** 2026-06-15
+**Scope:** entire `main` branch (HEAD `5dfa6e4`)
+**Method:** static review of source + schema + tests, branch-wide pattern scan
+(`ruff`, `mypy`, defect-class greps), full test run (`330 passed`).
+
+> Status legend — **HIGH** = fix before relying on the feature ·
+> **MED** = fix soon, correctness/maintainability · **LOW** = polish ·
+> **INFO** = accepted risk / no action, documented for the record.
+
+This document **only records findings**. Remediation planning is a separate
+step (next phase of the audit loop).
+
+---
+
+## 0. Headline
+
+The most recent dev-loop work (**F-15 citation graph** + the follow-up fixes
+`cited_id→papers.id` resolution and the `paper_identifiers` alias table) is
+**functionally green but unverified where it matters**: the SQL that *is* the
+fix has zero real-database coverage, the live-corpus validation in the ADR
+predates the fix, and the same ID-mismatch bug the fix solves still lives in a
+parallel code path (`discover.py`). Separately, a **live DB password sits in an
+un-ignored file** and the new table has **no migration path onto the live DB**.
+
+Test suite: **330 passed**, `ruff check` clean, `ruff format --check` clean,
+`mypy` clean on reviewed modules.
+
+---
+
+## 1. Security
+
+### S-1 — Live DB credential in an un-ignored file · **HIGH**
+- **Evidence:** `.env.jetson-ingest` contains
+  `DATABASE_URL=postgresql://researcher:05761b87…@localhost:5433/papers`.
+  `.gitignore` ignores only the literal `.env` (line: `# Environment secrets — NEVER commit` → `.env`).
+  `git check-ignore .env.jetson-ingest` → **NOT IGNORED**.
+- **Impact:** one `git add -A && git commit` leaks a real Postgres password to
+  history. The file is currently untracked (`??`), so the leak has **not**
+  happened yet — this is a near-miss, not an incident.
+- **Note:** the same gap exposes any future `.env.<profile>` file.
+
+### S-2 — Non-constant-time token comparison (MCP HTTP) · **LOW**
+- **Evidence:** `codex/mcp_server.py:340` — `if auth != f"Bearer {token}":`.
+- **Impact:** theoretical timing side-channel on the Bearer token. Low on a
+  trusted LAN, but trivially fixable with `secrets.compare_digest`.
+
+### S-3 — DNS-rebinding protection disabled · **INFO (accepted)**
+- **Evidence:** `codex/mcp_server.py:35-37`,
+  `TransportSecuritySettings(enable_dns_rebinding_protection=False)` (commit `d29dcf2`).
+- **Assessment:** documented and mitigated — HTTP transport requires a Bearer
+  token (`_TokenAuth`) and is firewalled to LAN IPs (UFW). A rebound browser
+  request lacks the token. Residual risk only if the firewall assumption fails.
+  No action; recorded so the trade-off stays visible.
+
+---
+
+## 2. Correctness
+
+### C-1 — `discover.py` still has the ID-format-mismatch bug F-15 fixed · **HIGH**
+- **Evidence:** `codex/discover.py:20` —
+  `WHERE cited_id NOT IN (SELECT id FROM papers)`; same pattern in
+  `cocited_papers` (`discover.py:52-61`), `citing_papers`, `cited_by`.
+- **Root cause:** `citations.cited_id` is a **mix** of OpenAlex W-IDs (OpenAlex
+  API path, `ingest.py:229-231` ← `sources/openalex.py:155`) and DOIs/arXiv
+  (GROBID/S2 paths). `papers.id` is the canonical DOI/arXiv. An ingested paper
+  cited by its **OpenAlex ID** is *not* in `(SELECT id FROM papers)`, so it is
+  wrongly reported as a discovery lead.
+- **Impact:** `codex discover leads`, `codex discover cocited`, and the MCP
+  `discover_leads` tool over-report — foundational papers the user has **already
+  ingested** show up as "cited but not ingested". This is exactly the
+  D-05-class issue the F-15 graph path resolves via the
+  `papers.openalex_id` JOIN (`graph.py:48-56`) — but the fix was **not
+  propagated** to `discover.py`. Two divergent "dangling" implementations now
+  disagree on what "ingested" means.
+
+### C-2 — `paper_identifiers` JOIN is effectively inert · **MED**
+- **Evidence:** `graph.py:50-56` second LEFT JOIN; populated only at
+  `ingest.py:153-161`, which inserts **the same** `paper.openalex_id` that
+  `papers.openalex_id` already holds.
+- **Impact:** because ingest never records an alias that differs from
+  `papers.openalex_id`, the `pi` JOIN can only resolve what the `p` JOIN already
+  resolves. It adds value **only** in the narrow case where a paper's
+  `openalex_id` changes across re-ingests (old alias retained via
+  `ON CONFLICT DO NOTHING`). The table + unique index + seed INSERT + migration
+  burden buy near-zero behaviour today — speculative generality. No test
+  exercises an alias that diverges from `papers.openalex_id`.
+
+### C-3 — MCP `wiki_read` returns the slug as its own "sources" · **MED**
+- **Evidence:** `codex/mcp_server.py:138-149` — comment says "Collect cited
+  bibkeys", code sets `sources = [concept_slug]`.
+- **Impact:** the `sources` field is a placeholder masquerading as data; any MCP
+  client trusting it to list cited bibkeys gets the concept's own slug.
+
+---
+
+## 3. Documentation / Code Drift
+
+### D-1 — ADR-F15 validation predates the shipped graph · **HIGH (integrity)**
+- **Evidence:** ADR-F15 (commit `a1a6f45`) lands **before** the resolution fixes
+  `664906f` (`cited_id→papers.id`) and `5dfa6e4` (`paper_identifiers`).
+- **Stale claims:**
+  - "Spike Result": *478 dangling nodes*, specific PageRank scores,
+    *Bobenko/Springborn rank 7*, *5.7 % spread* — all measured on the
+    **pre-resolution** (bipartite DOI→OpenAlex) graph. Post-fix, in-KB cited
+    papers resolve to DOIs, so dangling count and PageRank distribution change.
+  - "Decision → Graph Construction": "inter-KB citations are structurally
+    invisible" — **now false**.
+  - "Known Limitation: ID-Format Mismatch": "cross-citations… not represented"
+    — **now resolved by the code**, contradicting the section.
+- **Impact:** the **GO** decision rests on numbers the shipping code no longer
+  produces. Validation must be re-run on the post-fix graph (or the ADR marked
+  validation-pending).
+
+### D-2 — `graph_cite_boost_alpha` doc describes the *fixed* bug · **MED**
+- **Evidence:** `config.py:296-300` describes
+  `Final score = dense_score * (1 + alpha * pagerank_score)` — a **multiply**.
+  Actual code (`cli.py:107`) divides distance:
+  `boosted = distance / (1.0 + alpha * pr_score)`. The ADR explicitly states the
+  multiply form was the "critical bug corrected during the F-15 review gate".
+- **Impact:** the config docstring still documents the inverted (buggy) formula.
+
+### D-3 — `paper_identifiers` tagged "F-17" in schema, shipped as F-15 fix · **LOW**
+- **Evidence:** `infra/schema.sql:129` comment "F-17 Paper identifier aliases",
+  but the table is introduced by the F-15 graph follow-up (`5dfa6e4`) and the
+  ADR-F15 work. Feature-tag inconsistency; ADR-F15 doesn't mention the table at all.
+
+---
+
+## 4. Schema / Migration
+
+### M-1 — New table has no migration path onto the live DB · **HIGH (operational)**
+- **Evidence chain:**
+  1. `apply_schema` docstring (`db.py:46-49`) claims "all DDL statements use
+     CREATE … IF NOT EXISTS" — **false**: `papers`, `chunks`, `citations`,
+     `code_links` and their indexes use bare `CREATE TABLE/INDEX`
+     (`schema.sql:16,33,56,68,30,41,…`). Only F-09/F-16/F-17 DDL is idempotent.
+  2. `apply_schema` is **never called** anywhere in the app — no `init`/`migrate`
+     CLI command (`grep apply_schema` → only `db.py` + a mocked unit test).
+  3. `ingest_all.sh` does **not** apply the schema; it only runs `codex ingest`.
+- **Impact:** re-running `infra/schema.sql` against the **existing** Jetson DB
+  fails on the first `CREATE TABLE papers` (already exists) and rolls back the
+  whole script in one implicit transaction — so the new `paper_identifiers`
+  table (and its seed INSERT) is **never created** on a DB that predates it.
+  `build_citation_graph` then fails with `relation "paper_identifiers" does not
+  exist` on the live corpus. The migration step is currently manual and undocumented.
+
+---
+
+## 5. Test Integrity
+
+### T-1 — The F-15 resolution SQL has zero real coverage · **HIGH**
+- **Evidence:** every test in `tests/graph/test_graph.py` and
+  `tests/graph/test_cli.py` builds the graph from a `MagicMock` conn
+  (`_make_conn`, `_make_conn_with_paper_ids`). `test_cross_citation_uses_doi_when_in_kb`
+  (`test_graph.py:83-99`) **pre-resolves** the rows in Python and asserts the
+  graph builds — it tests the mock, not the `COALESCE(p.id, pi.paper_id, …)` +
+  two LEFT JOINs that *are* the fix.
+- **Impact:** the substance of the two most recent commits is unverified. 330
+  green tests give false confidence about the resolution logic, the `p.id IS NULL`
+  correlated guard, and the `paper_identifiers` unique-index behaviour. Needs an
+  integration test against a real (or `pytest`-managed) Postgres.
+
+### T-2 — `apply_schema` idempotency untested · **MED**
+- **Evidence:** `tests/scaffold/test_db.py:24-32` mocks the conn and only checks
+  `execute`/`commit` are called. The (false) idempotency claim in M-1 is never
+  exercised.
+
+---
+
+## 6. Design / Usability
+
+### U-1 — `graph report` prints bare IDs, no titles · **LOW**
+- `cli.py:582-588` emits `score  <id>` and bare dangling IDs — a mix of DOIs and
+  `W…` OpenAlex IDs with no title/year. The ADR's readable author/title table is
+  not what the CLI produces; output is hard to interpret.
+
+### U-2 — `--cite-boost` can only reorder within the top-`limit` page · **LOW**
+- `cli.py:84-94` applies `ORDER BY … LIMIT limit` in SQL **before** the boost
+  (`cli.py:96-109`). A high-PR paper just outside the semantic cut can never be
+  pulled in. With α=0.3 and PR≈0.002 the effect is ~0.06 % — effectively a
+  within-page tie-breaker. Consistent with ADR intent, but worth stating as a
+  limitation (the flag does far less than "weight results by PageRank" implies).
+
+### U-3 — `graph report --json` drops the small-corpus warning · **LOW**
+- `cli.py:557-570` returns before the `graph_min_corpus_size` warning
+  (`cli.py:572-577`). JSON consumers get neither the warning nor a flag for it.
+
+### U-4 — `graph.find_co_cited` implemented + tested but not wired to a CLI · **LOW**
+- `graph.find_co_cited` (`graph.py:142`) is unit-tested but exposed by no `graph`
+  sub-command (only the separate SQL-based `discover cocited` is wired). ADR-F15
+  R-43's co-citation is only half-surfaced.
+
+---
+
+## 7. State of Tree (INFO — no defect)
+
+- The uncommitted diff (`wiki.py`, `synthesis.py`, `semanticscholar.py`,
+  `tests/synthesis/test_cli.py`, `tests/wiki/test_compile.py`) is a **legitimate
+  `ruff format` normalization** to `line-length=100`. `ruff format --check`
+  reports all 25 files formatted *with* these changes applied, meaning the
+  **committed** code was non-compliant. These edits should be **committed**, not
+  reverted. (Not a finding; clarifies the working-tree churn.)
+
+---
+
+## 8. Coverage of this audit (honesty ledger)
+
+**Deep-reviewed:** `graph.py`, `db.py`, `infra/schema.sql`, `ingest.py`,
+`discover.py`, `provenance.py` (scan), `mcp_server.py`, `cli.py` (search +
+graph), `config.py` (graph settings), F-15 tests, ADR-F15; branch-wide pattern
+scan (SQL injection, bare except, eval/exec/subprocess, secrets, timeouts).
+
+**Not yet deep-reviewed (no findings asserted):**
+- `synthesis.py` grounding guard / claim parser (`_run_grounding_guard`,
+  `_parse_claims`) — the "zero-fact-leak (N-02)" core. Prompts reviewed; guard
+  logic not yet traced.
+- `wiki.py` grounding guard + conflict detection (only the format diff seen).
+- `sources/` (arxiv, openalex, semanticscholar) beyond citation shapes.
+- `parsing/` (grobid, nougat, mathpix, figures, tex) — error handling in
+  `mathpix.py`/`figures.py` uses broad `except Exception` without `noqa`
+  (`mathpix.py:109,151`, `figures.py:125,152`) — flagged for a later pass.
+- `embed.py`, `quality.py`, `models.py`.
+
+A second pass should close these before the audit is declared complete.
--- a/docs/audit/DATA-QUALITY-2026-06-15.md
+++ b/docs/audit/DATA-QUALITY-2026-06-15.md
@@ -0,0 +1,160 @@
+# DATA-QUALITY AUDIT — codex knowledge base (handoff for a fresh session)
+
+**Status:** baseline scan done 2026-06-15; four investigation items open (DQ-1…DQ-4).
+**Author:** Audit-Loop (Opus). **Intended reader:** a *cold* session with no memory
+of the audit conversation — everything needed to continue is in this file.
+
+---
+
+## Why this exists — the second audit axis
+
+The code/pipeline audit (`AUDIT-2026-06-15-loop-1.md`, `-full-repo.md`) verified the
+**loader is correct**: IDs are canonical, the citation graph is consistent, the
+ingest pipeline does what it claims. It did **not** ask whether the *information
+in the database is actually good*. A correct loader can faithfully load thin,
+incomplete, or unrepresentative data. This document is that second axis — a
+**data-quality** assessment of the live corpus, independent of code correctness.
+
+---
+
+## How to reach the data (self-contained)
+
+- **Live DB:** Jetson PostgreSQL via an SSH tunnel (open it first):
+  ```
+  ssh -f -N -L 5433:localhost:5432 alfred@192.168.178.103
+  ```
+- **Credentials:** `.env.jetson-ingest` (gitignored) holds `DATABASE_URL`
+  (`postgresql://researcher:…@localhost:5433/papers`). The `researcher` role is
+  DML-only (read freely; no DDL).
+- **Run probes:** `set -a; source .env.jetson-ingest; set +a` then
+  `PYTHONPATH=. .venv/bin/python` with `psycopg` + `psycopg.rows.dict_row`.
+  Reusable domain helpers: `codex.quality` (`is_quality_chunk`, `_bib_score`,
+  `classify_section`), `codex.graph`, `codex.discover`.
+- **Corpus snapshot (2026-06-15, post canonical-id migration):**
+  **29 papers, 1507 chunks, 590 citations.**
+
+---
+
+## Baseline — what is already GOOD (measured, no action needed)
+
+- **Chunk content is clean.** Re-running the F-16 quality gate over all stored
+  chunks: **0 / 1507 fail** `is_quality_chunk` (none too-short, none low-alpha /
+  OCR-ish, none bibliography-like). The gate did its job on the `.txt` ingest.
+- **Chunk sizing healthy.** chars: min 478 / median 2913 / max 4154; **0 exact
+  duplicate** chunk bodies.
+- **`section` column is low-signal** (98 % `body`; 13 `bibliography`, 9 `theorem`,
+  2 `intro`, 1 `proof`). This confirms code-audit **R-12** — word-window chunks
+  rarely start at a section header. Not a data defect; the column is just weak.
+
+---
+
+## Findings (open investigation items)
+
+### DQ-1 — Citation coverage: 12 / 29 papers (41 %) have ZERO citations · **HIGH**
+- **Measured:** 12 papers contribute **no** out-edges; the 590-edge citation graph
+  comes from only ~17 papers. The whole F-15 layer (PageRank / coupling /
+  co-citation / discovery leads) therefore runs on ~59 % of the corpus.
+- **Sharpening — it is mostly a source-coverage limit, not an ingest bug:**
+  **11 of the 12** zero-citation papers *do* have an `openalex_id`, i.e. ingest
+  called `openalex.fetch_citations(openalex_id)` and OpenAlex returned an **empty
+  `referenced_works`** list. Only `1911.00966` has no `openalex_id` (OpenAlex
+  404 → arXiv/S2 fallback, which also yielded nothing).
+- **Zero-citation ids:** `1005.2698`, `1505.01341`, `1911.00966` (no oa),
+  `2206.13461`, `2305.10988`, `2310.17529`, `2601.22903`, `math/0001176`,
+  `math/0503219`, `math/0603097`, `10.14279/depositonce-20357`,
+  `10.14279/depositonce-5415`. (Pattern: arXiv preprints + the two `depositonce`
+  theses — works OpenAlex indexes without parsed references.)
+- **To investigate next:**
+  1. Confirm it is OpenAlex coverage, not a `fetch_citations` bug: for 2–3 of the
+     ids, hit `GET https://api.openalex.org/works/<openalex_id>` and check whether
+     `referenced_works` is genuinely empty server-side.
+  2. If genuinely empty: enrich via the **Semantic Scholar references** path
+     (`semanticscholar.fetch_references("arXiv:<id>")`) as a *supplement* for
+     OpenAlex-empty papers — S2 often has references where OpenAlex doesn't. Note
+     S2 cited_ids are bare DOI/arXiv, so they flow through the existing
+     `RESOLVED_CITATIONS_SQL` resolver (audit C-1) fine.
+  3. Decide whether the F-15 `graph_min_corpus_size` warning should also flag *low
+     citing-paper coverage*, not just paper count.
+- **Acceptance:** either (a) citation coverage materially improves after an S2
+  supplement, or (b) documented as an inherent OpenAlex-coverage limit with the
+  graph caveated accordingly.
+
+### DQ-2 — Metadata gaps: 3 no-abstract, 1 fully metadata-less · **MED**
+- **Measured:**
+  - **No abstract (3):** `10.1007/978-3-642-17413-1_7`,
+    `10.1007/s00454-019-00132-8`, `1911.00966`. These get a **zero-vector**
+    `abstract_emb`, so paper-level semantic search can't place them.
+  - **No bibkey / year (1):** `1911.00966` only — and its authors array is empty
+    too. This paper is **fully degraded** (OpenAlex 404 → `Paper(id, title="")`
+    fallback): no abstract, no bibkey, no year, empty authors, no citations. With
+    no bibkey it cannot be `@cite`d and is invisible to wiki grounding (which keys
+    on bibkey).
+- **To investigate next:**
+  1. `1911.00966` — confirm the arXiv id is correct and whether OpenAlex/S2 has it
+     under a different id (DOI?); if recoverable, re-ingest to populate metadata.
+     If not, decide: keep as a degraded node or drop.
+  2. The 2 no-abstract DOIs — check if OpenAlex has an `abstract_inverted_index`
+     that the mapper missed, or if the abstract is genuinely absent upstream.
+- **Acceptance:** every paper has at least a bibkey + non-zero abstract embedding,
+  or the exceptions are documented with rationale.
+
+### DQ-3 — Content fidelity (chunks vs source `.txt`): NOT YET VERIFIED · **MED**
+- **Open question:** does the stored chunk set faithfully reconstruct each source
+  file, or did the chunker / `filter_chunks` silently drop material (e.g. an
+  abstract, a section, math-heavy passages)? The whole corpus was ingested from
+  `.txt` (`PAPERS_DIR=/Users/tarikmoussa/Desktop/ConformalLabpp/papers/txt`), so
+  the `.tex`/`.pdf` paths never ran.
+- **To investigate next:** for a sample of ~5 papers, compare
+  `len("".join(stored chunks))` against `len(source.txt)` (coverage %), and
+  eyeball the first/last chunk vs the file head/tail. Flag papers where coverage
+  is low (content lost) — F-16 dropping >X % of a paper is a signal.
+- **Acceptance:** sampled papers retain ≳ the expected fraction of source text;
+  no paper is silently gutted by the quality gate.
+
+### DQ-4 — Retrieval quality: NOT YET MEASURED · **MED**
+- **Open question:** end-to-end, do real queries return *relevant* results? Clean
+  chunks + a working index don't guarantee useful retrieval.
+- **To investigate next:** run a handful of domain queries through the actual
+  search path (`codex search paper "<q>"` and the chunk-level MCP `search`), e.g.
+  "discrete conformal map", "circle packing rigidity", "combinatorial Yamabe
+  flow", "discrete Laplace-Beltrami operator", and judge whether the top-5 hits
+  are on-topic and point at the right papers. Cross-check a couple against
+  `--cite-boost` to see if the boost helps or hurts on this corpus.
+- **Acceptance:** a short relevance table (query → top-5 → on/off-topic) good
+  enough to trust the KB for lookups, or a list of failure modes to fix.
+
+---
+
+## Priority for the next session
+1. **DQ-1** (biggest lever — 41 % of the corpus is invisible to the graph; the S2
+   supplement is concrete and reuses existing code).
+2. **DQ-4** (cheap, high-information — tells you if the KB is actually usable).
+3. **DQ-2**, then **DQ-3**.
+
+All four are *data* work (queries + maybe a re-ingest/enrichment), not code-audit
+work — the code is already remediated (see the AUDIT-* docs and PRs #12–#14).
+
+---
+
+## Appendix — reproduce the baseline scan
+
+```python
+import os, psycopg, statistics
+from collections import Counter, defaultdict
+from psycopg.rows import dict_row
+from codex.config import Settings
+from codex.quality import is_quality_chunk, _bib_score
+
+s = Settings()
+with psycopg.connect(os.environ["DATABASE_URL"], row_factory=dict_row) as c:
+    papers = c.execute("SELECT id, bibkey, title, abstract, year, authors, openalex_id FROM papers").fetchall()
+    chunks = c.execute("SELECT paper_id, content, section FROM chunks").fetchall()
+    cit = c.execute("SELECT citing_id, count(*) n FROM citations GROUP BY citing_id").fetchall()
+
+# zero-citation papers (DQ-1)
+cited = {r["citing_id"] for r in cit}
+zero = [p["id"] for p in papers if p["id"] not in cited]
+# metadata gaps (DQ-2): p["abstract"] empty / p["bibkey"] None / not p["authors"]
+# F-16 re-gate (baseline): [ch for ch in chunks if not is_quality_chunk(ch["content"], settings=s)]
+# section dist (R-12): Counter(ch["section"] or "(null)" for ch in chunks)
+```
--- a/ingest_all.sh
+++ b/ingest_all.sh
@@ -0,0 +1,116 @@
+#!/usr/bin/env bash
+# Batch ingest — writes directly to Jetson PostgreSQL via SSH tunnel.
+#
+# Prerequisites:
+#   ssh -f -N -L 5433:localhost:5432 alfred@192.168.178.103
+#   uv sync  (done once)
+#
+# Usage: bash ingest_all.sh 2>&1 | tee ingest_all.log
+set -euo pipefail
+
+PAPERS_DIR="/Users/tarikmoussa/Desktop/ConformalLabpp/papers/txt"
+CODEX_DIR="$(cd "$(dirname "$0")" && pwd)"
+ENV_FILE="$CODEX_DIR/.env.jetson-ingest"
+TUNNEL_PORT=5433
+
+# ── 1. Verify SSH tunnel is active ───────────────────────────────────────────
+if ! nc -z localhost "$TUNNEL_PORT" 2>/dev/null; then
+  echo "ERROR: SSH tunnel not active on port $TUNNEL_PORT"
+  echo "Run: ssh -f -N -L 5433:localhost:5432 alfred@192.168.178.103"
+  exit 1
+fi
+echo "✓ SSH tunnel active on port $TUNNEL_PORT"
+
+# ── 1b. Clear macOS UF_HIDDEN flag on editable-install .pth files ─────────────
+# Python 3.13 skips .pth files with this flag set (new in 3.13); uv sets it.
+# --no-sync below prevents uv from re-syncing (which would re-set the flag).
+SP="$CODEX_DIR/.venv/lib/python3.13/site-packages"
+for pth in "$SP"/_editable_impl_codex*.pth; do
+  [[ -f "$pth" ]] && chflags nohidden "$pth" 2>/dev/null || true
+done
+echo "✓ Cleared hidden flags on editable .pth files"
+
+# ── 2. Load env (DATABASE_URL → Jetson via tunnel) ───────────────────────────
+# pydantic-settings reads DATABASE_URL directly from the environment.
+# DOTENV_PATH is not supported; export each var explicitly.
+set -a; source "$ENV_FILE"; set +a
+export DATABASE_URL GROBID_URL OLLAMA_BASE_URL EMBEDDING_MODEL EMBEDDING_DIM OPENALEX_MAILTO
+echo "✓ DATABASE_URL: $DATABASE_URL"
+
+# ── 3. Paper ID → source file mapping ────────────────────────────────────────
+# Format: "PAPER_ID|filename.txt"
+# SKIP: farkas-kra-1992-riemann-surfaces.txt (textbook, no OpenAlex entry)
+declare -a PAPERS=(
+  # arXiv papers
+  "math/0603097|springborn-2008-weighted-delaunay-hyperideal.txt"
+  "1005.2698|bobenko-pinkall-springborn-2015-discrete-conformal-maps.txt"
+  "math/0306167|luo-2004-combinatorial-yamabe-flow.txt"
+  "math/0203250|bobenko-springborn-2004-circle-patterns.txt"
+  "math/0503219|bobenko-springborn-2007-discrete-laplace-beltrami.txt"
+  "2310.17529|bobenko-lutz-2025-non-euclidean-dce.txt"
+  "2305.10988|bobenko-lutz-2024-decorated-conformal-maps.txt"
+  "2206.13461|lutz-2023-canonical-tessellations.txt"
+  "2601.22903|bowers-bowers-lutz-2026-koebe-rigidity.txt"
+  "1911.00966|pinkall-springborn-2021-liouville.txt"
+  "1505.01341|born-bucking-springborn-2015-quasiconformal.txt"
+  "0906.1560|glickenstein-2011-discrete-conformal-variations.txt"
+  "math/0001176|rivin-schlenker-2000-schlafli-formula-arxiv-preprint.txt"
+  # DOI papers
+  "10.1007/s00454-019-00132-8|springborn-2020-ideal-hyperbolic-polyhedra.txt"
+  "10.1145/1964921.1964997|alexa-wardetzky-2011-discrete-laplacians-polygonal.txt"
+  "10.1111/cgf.13931|bunge-herholz-kazhdan-botsch-2020-polygon-laplacian.txt"
+  "10.1145/3450626.3459763|gillespie-springborn-crane-2021-discrete-conformal-equivalence.txt"
+  "10.1145/3306346.3323042|sharp-soliman-crane-2019-navigating-intrinsic-triangulations.txt"
+  "10.1145/3197517.3201367|soliman-slepcv-crane-2018-optimal-cone-singularities.txt"
+  "10.1145/3132705|sawhney-crane-2017-boundary-first-flattening.txt"
+  "10.1145/2767000|knoppel-crane-pinkall-schroder-2015-stripe-patterns.txt"
+  "10.5555/1070432.1070581|erickson-whittlesey-2005-greedy-homotopy-homology.txt"
+  "10.1145/1185657.1185665|desbrun-kanso-tong-2006-discrete-differential-forms.txt"
+  "10.1080/10586458.1993.10504266|pinkall-polthier-1993-computing-discrete-minimal-surfaces.txt"
+  "10.1007/s11040-021-09394-2|bobenko-bucking-2021-period-matrices.txt"
+  "10.14279/depositonce-20357|lutz-2024-thesis.txt"
+  "10.14279/depositonce-5415|sechelmann-2016-thesis.txt"
+  # Book chapters (txt contains full book; ID = cited chapter)
+  "10.1007/0-387-29555-0_13|precopa-molnar-eds-2006-non-euclidean-geometries-book.txt"
+  "10.1007/978-3-642-17413-1_7|bobenko-klein-eds-2011-computational-approach-riemann-surfaces-book.txt"
+)
+
+# ── 4. Ingest loop ────────────────────────────────────────────────────────────
+OK=0; FAIL=0; SKIP=0
+FAILED_IDS=()
+
+cd "$CODEX_DIR"
+
+for entry in "${PAPERS[@]}"; do
+  PAPER_ID="${entry%%|*}"
+  FILENAME="${entry##*|}"
+  SOURCE="$PAPERS_DIR/$FILENAME"
+
+  if [[ ! -f "$SOURCE" ]]; then
+    echo "SKIP (file missing): $FILENAME"
+    (( SKIP++ )) || true
+    continue
+  fi
+
+  echo ""
+  echo "── Ingesting: $PAPER_ID"
+  echo "   source: $FILENAME"
+
+  if PYTHONPATH="$CODEX_DIR" "$CODEX_DIR/.venv/bin/codex" ingest "$PAPER_ID" --source "$SOURCE"; then
+    (( OK++ )) || true
+  else
+    echo "FAILED: $PAPER_ID"
+    FAILED_IDS+=("$PAPER_ID")
+    (( FAIL++ )) || true
+  fi
+done
+
+# ── 5. Summary ────────────────────────────────────────────────────────────────
+echo ""
+echo "════════════════════════════════════════"
+echo "Ingest complete: $OK OK, $FAIL FAILED, $SKIP SKIPPED"
+if [[ ${#FAILED_IDS[@]} -gt 0 ]]; then
+  echo "Failed IDs:"
+  for id in "${FAILED_IDS[@]}"; do echo "  $id"; done
+fi
+echo "════════════════════════════════════════"
--- a/tests/synthesis/test_cli.py
+++ b/tests/synthesis/test_cli.py
@@ -55,9 +55,7 @@ def isolated_leads_dir(monkeypatch: pytest.MonkeyPatch, tmp_path: Path) -> Path:

    monkeypatch.setattr("codex.cli.get_settings", lambda: settings, raising=False)
    monkeypatch.setattr("codex.synthesis.get_settings", lambda: settings)
-    monkeypatch.setattr(
-        "codex.config.get_settings", lambda: settings, raising=False
-    )
+    monkeypatch.setattr("codex.config.get_settings", lambda: settings, raising=False)
    return leads_path


--- a/tests/wiki/test_compile.py
+++ b/tests/wiki/test_compile.py
@@ -201,9 +201,7 @@ def test_grounding_guard_last_sentence_grounded_earlier_hallucination_ignored()
    This test verifies that a claim whose text = the last sentence grounds correctly.
    """
    # The last sentence of the claim text is directly grounded in the fixture chunk.
-    grounded_sentence = (
-        "the hyperbolic volume is V = L(gamma1) + L(gamma2) + L(gamma3)"
-    )
+    grounded_sentence = "the hyperbolic volume is V = L(gamma1) + L(gamma2) + L(gamma3)"
    claim = Claim(
        text=grounded_sentence,
        bibkey="Springborn2008",
@@ -523,9 +521,7 @@ def test_idempotent_no_rewrite_when_unchanged(
    call_count = 0
    original_compile = compile_concept

-    def counting_compile(
-        concept: Concept, chunks: list[Any], **kwargs: Any
-    ) -> ConceptPage:
+    def counting_compile(concept: Concept, chunks: list[Any], **kwargs: Any) -> ConceptPage:
        nonlocal call_count
        call_count += 1
        return original_compile(concept, chunks, **kwargs)
@@ -569,9 +565,7 @@ def test_force_all_recompiles_even_when_unchanged(
    call_count = 0
    original_compile = compile_concept

-    def counting_compile(
-        concept: Concept, chunks: list[Any], **kwargs: Any
-    ) -> ConceptPage:
+    def counting_compile(concept: Concept, chunks: list[Any], **kwargs: Any) -> ConceptPage:
        nonlocal call_count
        call_count += 1
        return original_compile(concept, chunks, **kwargs)