refactor(027)!: drop deprecation shims — hard cutover to abstractatlas/aacli

Per requester: no deprecation shims. Removes:
- src/ohbm2026/__init__.py (the import meta-path shim) — `import ohbm2026`
  now raises ModuleNotFoundError.
- the `ohbmcli` [project.scripts] entry + `_ohbmcli_deprecated_main` in
  cli.py — `ohbmcli` is now `command not found`.

Also renames the remaining `ohbmcli` string literals in src (+ tests):
argparse `prog=` (aacli --help now shows `usage: aacli`), provenance
`command_line` strings, help/error text; and fixes a dynamic
`__import__("ohbm2026")` in test_enrich_stage that the module-ref passes
missed. Reinstalled the venv (uninstall ohbm2026 dist → install -e .) so a
prior installed `ohbm2026` can't mask the cutover.

Spec/plan/tasks/contracts (FR-003, SC-007, R3, rename-map, cli-aacli) +
CLAUDE pointer updated to describe the hard cutover.

Verified: full suite `Ran 1034 — OK (skipped=1)`; `import ohbm2026` →
ModuleNotFoundError; `ohbmcli` → command not found; `aacli` works.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Author: satra

CLAUDE.md

@@ -175,8 +175,10 @@ things: (1) **renames** the component `ohbm2026`→`abstractatlas` (package
 `src/ohbm2026/`→`src/abstractatlas/` via `git mv` + import rewrite across
 ~86 files) and the CLI `ohbmcli`→**`aacli`** (module form
 `python -m abstractatlas.cli`; legacy `ohbm-*` scripts→`aa-*` or dropped),
-with `ohbmcli`/`import ohbm2026` kept as **labeled deprecation shims** for
-one cycle; and (2) **generalizes ingestion** into a pluggable architecture —
+with `ohbmcli`/`import ohbm2026` **removed outright (hard cutover, no
+shims)** — they fail loudly (`command not found` / `ModuleNotFoundError`);
+the venv is reinstalled so a prior pip-installed `ohbm2026` dist doesn't
+mask it; and (2) **generalizes ingestion** into a pluggable architecture —
 a NEW `src/abstractatlas/ingest/` subpackage with an `Ingestor` ABC
 (`pull`→`normalize`→`validate`), a runtime-discoverable `registry`
 (never a hardcoded source list — CA-007), and a standardized **LinkML
@@ -199,7 +201,7 @@ fidelity. Companions: `research.md`, `data-model.md`,
 `contracts/{rename-map,ingestor-interface,ingest-schema.linkml.yaml,cli-aacli}.md`,
 `quickstart.md`. NOTE for future sessions: after this ships, the package is
 `abstractatlas` and the CLI is `aacli` — update muscle memory (`ohbmcli`
-still works via a deprecation shim slated for removal).
+and `import ohbm2026` are GONE; they error, no shim).
 
 The immediately-prior Stage 26 plan is at
 `specs/026-neuroscape-year-density/plan.md`. Stage 26 (Track A,

pyproject.toml

@@ -39,15 +39,13 @@ abstracts_book = [
 ]
 
 [project.scripts]
-# Canonical CLI (spec 027: ohbmcli → aacli).
+# Canonical CLI (spec 027: ohbmcli → aacli). Hard cutover — no `ohbmcli`
+# alias and no `ohbm2026` import shim: the old names are removed and fail
+# loudly (command-not-found / ModuleNotFoundError). The legacy per-stage
+# `ohbm-*` console scripts were also dropped (most pointed at modules moved
+# in the Stage-5 reorg); their functions remain importable via
+# `abstractatlas.*` and runnable as `aacli` subcommands.
 aacli = "abstractatlas.cli:main"
-# Deprecated alias — warns then delegates to aacli. Kept one transition
-# cycle; remove next release. (Spec 027 FR-003 / research R3.)
-ohbmcli = "abstractatlas.cli:_ohbmcli_deprecated_main"
-# The legacy per-stage `ohbm-*` console scripts were dropped in spec 027:
-# most already pointed at modules moved in the Stage-5 reorg
-# (ohbm2026.enrichment, ohbm2026.openalex). Their functions remain
-# importable via `abstractatlas.*` and runnable as `aacli` subcommands.
 
 [build-system]
 requires = ["setuptools>=69"]

specs/027-abstractatlas-rename-ingestors/contracts/cli-aacli.md

@@ -30,11 +30,11 @@ subcommand (SC-001).
 > surface that makes future sources first-class without new bespoke
 > subcommands.
 
-## Deprecation
+## Hard cutover (no deprecation)
 
-- `ohbmcli <args>` → stderr: `ohbmcli is deprecated; use 'aacli'. Delegating…`
-  then runs `aacli <args>`. Removed next cycle.
-- `python -m ohbm2026.cli` → shim module warns + delegates.
+- `ohbmcli` → removed; `command not found`.
+- `import ohbm2026` / `python -m ohbm2026.cli` → `ModuleNotFoundError: No module named 'ohbm2026'`.
+- The sole interface is `aacli` / `python -m abstractatlas.cli`.
 
 ## Verification
 

specs/027-abstractatlas-rename-ingestors/contracts/rename-map.md

@@ -17,15 +17,17 @@ green) + SC-003 (byte-identical published data).
 | Docs | README, CLAUDE.md, constitution naming refs | updated to `abstractatlas`/`aacli` (FR-012) |
 | Site test strings | `ohbm2026` refs in `site/src/**` tests | updated where they name the package (NOT the `ohbm2026` data/site-mode/route which is the source identity) |
 
-## Deprecated shims (one transition period)
+## Hard cutover — no shims (per requester)
 
-| Legacy | Behavior |
-|--------|----------|
-| `ohbmcli` entry point | prints "renamed to `aacli`" deprecation notice, then delegates to `aacli` |
-| `import ohbm2026` | thin shim re-exporting `abstractatlas` with a `DeprecationWarning`; labeled + follow-up to remove |
+| Legacy | Behavior after cutover |
+|--------|------------------------|
+| `ohbmcli` | removed from `[project.scripts]` → `command not found: ohbmcli` |
+| `import ohbm2026` | package dir gone + venv reinstalled → `ModuleNotFoundError: No module named 'ohbm2026'` |
 
-Shims MUST be labeled in code with root cause + removal follow-up
-(Constitution VI). Neither may partially work in a confusing way (FR-003).
+No deprecation shim is provided. Both fail loudly and immediately (FR-003 /
+SC-007) — never a silent or partial success. The venv is reinstalled
+(`uv pip uninstall ohbm2026 && uv pip install -e .`) so a previously
+pip-installed `ohbm2026` dist can't mask the cutover.
 
 ## NOT renamed (data identity — FR-004)
 

specs/027-abstractatlas-rename-ingestors/research.md

@@ -43,20 +43,22 @@ option) — rejected by the requester for continuity.
 
 ## R3 — Legacy name policy (`ohbmcli`, `import ohbm2026`)
 
-**Decision**: `aacli` / `abstractatlas` are canonical. Provide a
-**deprecated shim** for one transition period: `ohbmcli` remains as an entry
-point that prints a clear "renamed to `aacli`" deprecation notice then
-delegates; `import ohbm2026` resolves via a thin shim module that re-exports
-`abstractatlas` with a `DeprecationWarning`. Both are labeled and slated for
-removal.
-
-**Rationale**: Avoids silently breaking existing scripts/cron while making
-the new name canonical (FR-003, SC-007, edge case). A hard break is also
-acceptable per the spec, but a labeled shim is lower-friction and still
-loud. The shim carries a root-cause comment + follow-up per Constitution VI.
-
-**Alternatives considered**: hard removal with an error — acceptable but
-more disruptive; chose the shim as the default, removable next cycle.
+**Decision (updated per requester): HARD CUTOVER — no deprecation shims.**
+`aacli` / `abstractatlas` are the sole names. The `ohbmcli` entry point and
+any `ohbm2026` import path are removed; using them fails loudly with the
+standard not-found error (`command not found: ohbmcli` /
+`ModuleNotFoundError: No module named 'ohbm2026'`). The venv is reinstalled
+(`uv pip uninstall ohbm2026 && uv pip install -e .`) so a previously
+pip-installed `ohbm2026` dist doesn't mask the cutover.
+
+**Rationale**: The requester opted for a clean break — no shim to maintain
+or later remove. A bare not-found error is loud and unambiguous (satisfies
+FR-003/SC-007's "never silent/partial"), and there is no external
+consumer that needs a grace period.
+
+**Alternatives considered**: a labeled deprecation shim (`ohbmcli` warns +
+delegates; `ohbm2026` meta-path redirect) — implemented first, then
+**removed** at the requester's direction in favour of the hard cutover.
 
 ## R4 — Ingestor contract: wrap, don't rewrite
 

specs/027-abstractatlas-rename-ingestors/spec.md

@@ -102,7 +102,7 @@ error rather than silently passed downstream.
 ### Edge Cases
 
 - **Naming divergence for preserved data**: the package is `abstractatlas` but the published parquet keeps the name `ohbm2026.parquet` (that source's data name). The system must tolerate — and clearly document — this intentional divergence rather than "fix" the data name (which would force a re-publish).
-- **Legacy imports / command**: existing scripts, cron jobs, or muscle-memory using `ohbmcli` / `import ohbm2026` must fail with a clear pointer to the new name (or work via a clearly-labeled deprecated alias for one transition period), never a confusing partial success.
+- **Legacy imports / command**: existing scripts, cron jobs, or muscle-memory using `ohbmcli` / `import ohbm2026` fail loudly with the standard not-found error (hard cutover — the old names are removed), never a confusing partial success.
 - **State-keyed caches/checkpoints**: caches and checkpoints keyed by names that embed `ohbm2026` must remain readable so the rename does not silently invalidate expensive prior work.
 - **Source-type asymmetry**: a conference record has no DOI/journal; a literature record has no poster/session. The schema must make each set optional-by-source without letting a conference record masquerade as literature or vice versa.
 - **Two sources, one atlas**: an ingestor's records carry which source they came from, so downstream can attribute, filter, and de-duplicate across sources.
@@ -114,7 +114,7 @@ error rather than silently passed downstream.
 
 - **FR-001**: The Python package MUST be renamed `ohbm2026` → `abstractatlas`, and all internal imports, references, and docs updated to the new name.
 - **FR-002**: The canonical CLI MUST be renamed `ohbmcli` → `aacli`; every existing subcommand MUST remain available with identical behavior and options.
-- **FR-003**: Legacy entry points (`ohbmcli`, `import ohbm2026`) MUST either be removed with a precise "renamed to …" error or provided as an explicitly-deprecated alias for one transition period; they MUST NOT partially work in a confusing way.
+- **FR-003**: Legacy entry points (`ohbmcli`, `import ohbm2026`) MUST be removed outright (hard cutover — no deprecation shims); using them MUST fail loudly with the standard not-found error (`command not found` / `ModuleNotFoundError: No module named 'ohbm2026'`), never partially work in a confusing way.
 - **FR-004**: Existing on-disk data artifact paths, artifact/state-key naming, cache/checkpoint keys, and published data-package names (including `ohbm2026.parquet`) MUST be preserved so no data is regenerated and the live site requires no re-publish (byte-identical data).
 - **FR-005**: Ingestion MUST be organized as a pluggable architecture: a common **ingestor contract** (pull records from a source → normalize) with a runtime-discoverable registry of named ingestors.
 - **FR-006**: The two existing sources — the OHBM conference (Oxford Abstracts) and the PubMed/NeuroScape corpus — MUST be ported to become the first two ingestor instances behind the common contract, each reproducing its former normalized output exactly.
@@ -153,13 +153,13 @@ error rather than silently passed downstream.
 - **SC-004**: Both existing sources are expressed as named ingestors behind the common contract; porting them changed **zero** lines of downstream stage logic (measured by the diff touching only ingestor + rename code).
 - **SC-005**: 100% of records emitted by each ingestor validate against the LinkML ingest schema; a deliberately malformed record is rejected with a precise, source-attributed error (0 malformed records reach downstream).
 - **SC-006**: A maintainer can onboard a new source by implementing only the ingestor contract + registering it, with no edits to downstream stages — demonstrated by a documented "add an ingestor" guide and the two ports as worked examples.
-- **SC-007**: Using a legacy name (`ohbmcli` / `import ohbm2026`) yields a clear rename pointer or a labeled-deprecated alias — never a silent or partial success.
+- **SC-007**: Using a legacy name (`ohbmcli` / `import ohbm2026`) fails loudly and immediately with the standard not-found error — never a silent or partial success. (Hard cutover: the old names are gone.)
 
 ## Assumptions
 
 - "Foundation only": this change renames + establishes the ingestor architecture + LinkML ingest schema + ports the two existing sources. It does **not** add arXiv/bioRxiv/medRxiv or new-conference ingestors — those are follow-on specs enabled by this foundation.
 - Data continuity is paramount: on-disk artifact layout and published data-package names are preserved verbatim (including the historical `ohbm2026.parquet` name), so the rename is a code/CLI/package/docs change with no data regeneration or site re-publish.
 - The standardized LinkML ingest schema is designed to **capture the existing normalized shape** (so the ported OHBM output validates and downstream stays byte-identical), while being general enough to admit literature-index sources — it does not reshape already-published data.
-- A deprecated `ohbmcli` alias (and/or an `ohbm2026` import shim) may be provided for one transition period; the canonical, documented interface is `aacli` / `abstractatlas`.
+- Hard cutover: NO deprecation shims. `ohbmcli` and `import ohbm2026` are removed and fail loudly; the sole interface is `aacli` / `abstractatlas`. (The local venv must be reinstalled so a previously pip-installed `ohbm2026` dist is replaced by `abstractatlas`.)
 - Governance docs (constitution) contain naming references (`ohbmcli`, `src/ohbm2026/`, `data/abstracts.json`); updating those references is in scope for doc-sync, coordinated with the project's constitution-amendment process where required.
 - Downstream stages already treat the corpus generically enough that a standardized ingest contract can be introduced without changing their outputs for the existing sources.

specs/027-abstractatlas-rename-ingestors/tasks.md

@@ -57,7 +57,7 @@ renamed package and must follow it. Sequence: Setup → baseline → US1 → US2
 
 - [x] T003 [US1] `git mv src/ohbm2026 src/abstractatlas`, then rewrite `ohbm2026`→`abstractatlas` across all imports/refs in `src/` (per `contracts/rename-map.md`). Do NOT touch data-identity strings in the rename-map "NOT renamed" table (data paths, state-keys, `ohbm2026.parquet`, `/ohbm2026/` route, `SITE_MODE='ohbm'`, `OHBM2026_*`).
 - [x] T004 [US1] Update `pyproject.toml`: `name` `ohbm2026`→`abstractatlas`; `[project.scripts]` `ohbmcli`→`aacli = "abstractatlas.cli:main"`; rename `ohbm-*` entry points to `aa-*` where still used or drop them (record disposition per rename-map). Keep `package-dir`/`packages.find` under `src`.
-- [x] T005 [US1] Add labeled deprecation shims (FR-003, research R3): `ohbmcli` entry point that warns "renamed to aacli" then delegates; an `ohbm2026` import shim module re-exporting `abstractatlas` with a `DeprecationWarning`. Label both in code with root cause + removal follow-up (Constitution VI).
+- [x] T005 [US1] HARD CUTOVER — no deprecation shims (per requester). Remove the `ohbmcli` entry point (pyproject keeps only `aacli`), the legacy `ohbm-*` scripts, and rename all `ohbmcli` string literals in `src/` (argparse `prog=`, provenance `command_line`, help/error text) → `aacli`. Reinstall the venv package so a previously pip-installed `ohbm2026` dist is replaced by `abstractatlas` (`uv pip uninstall ohbm2026 && uv pip install -e .`), so `import ohbm2026` fails with `ModuleNotFoundError` and `ohbmcli` is `command not found` (FR-003/SC-007).
 - [x] T006 [P] [US1] Rewrite `ohbm2026`→`abstractatlas` imports/refs across `tests/` and `scripts/`; update `PYTHONPATH=src -m ohbm2026.cli`→`-m abstractatlas.cli` in scripts + `.github/workflows/**` + `.githooks/`.
 - [x] T007 [P] [US1] Update the few `ohbm2026` string refs in `site/src/**` tests that name the PACKAGE (not the `ohbm2026` data/route/site-mode identity). — VERIFIED NO-OP: audited `site/src` — every `ohbm2026` ref is data/route/storage-key identity (`/ohbm2026/` routes, `ohbm2026.parquet`, `ohbm2026.ui.*`/`ohbm2026.analytics.*` localStorage keys, conference `kind`); the TypeScript site never imports the Python package, so nothing to rename (2 stale comment path-mentions left as historical, like the src ones).
 - [x] T008 [US1] Verify the rename: `PYTHONPATH=src .venv/bin/python -m unittest discover -s tests` fully green (SC-002); re-run the T002 fixture under `aacli` and diff artifacts — identical (SC-001); `ohbmcli`/`import ohbm2026` emit the labeled deprecation and still succeed (SC-007); confirm an existing `data/cache/**` (or checkpoint) entry created before the rename is still discovered/reused — no forced recompute from the name change (Constitution III / spec edge case); `git grep -n 'ohbm2026\|ohbmcli'` returns only intentional data-identity names, shims, and historical spec docs.

src/abstractatlas/analyze/__init__.py

@@ -23,7 +23,7 @@
 - `analyze.umap` — Stage 4 UMAP fit + `project_into_umap` (US2).
 - `analyze.rollup` — Canonical UI rollup writer (annotations.parquet
   + sqlite).
-- `analyze.stage` — Stage 4 orchestrator (`ohbmcli analyze-matrix`).
+- `analyze.stage` — Stage 4 orchestrator (`aacli analyze-matrix`).
 
 **No package-level re-exports.** Per spec clarification Session
 2026-05-15 Q2, every caller imports from the explicit submodule that

src/abstractatlas/analyze/stage.py

@@ -1,4 +1,4 @@
-"""Stage 4 orchestrator — `ohbmcli analyze-matrix`.
+"""Stage 4 orchestrator — `aacli analyze-matrix`.
 
 Iterates the `(model, input_source, analysis_kind)` matrix, dispatches
 to per-kind runners, writes per-bundle artifacts atomically, and emits
@@ -7,7 +7,7 @@
 
 Per spec FR-001 / FR-002 / FR-003 / FR-014 / FR-017:
 
-- Single canonical CLI entrypoint (`ohbmcli analyze-matrix`).
+- Single canonical CLI entrypoint (`aacli analyze-matrix`).
 - Default matrix is **34 bundles** (5 models × 2 inputs × 4 kinds, with
   `neuroscape_clusters` auto-skipped for minilm/openai/pubmedbert
   because the published Stage-2 lens is Voyage-dim-specific).
@@ -365,7 +365,7 @@ def _run_entries_parallel(
     # internal thread pool should be 1. Setting these env vars BEFORE
     # the `Parallel` call ensures the loky-spawned subprocesses inherit
     # them. Existing user-set values win (the explicit env-var prefix
-    # `OMP_NUM_THREADS=N PYTHONPATH=src ohbmcli …` still applies).
+    # `OMP_NUM_THREADS=N PYTHONPATH=src aacli …` still applies).
     if config.n_jobs != 1:
         for var in (
             "OMP_NUM_THREADS",
@@ -594,7 +594,7 @@ def _git_rev_parse_head() -> str:
 
 def build_parser() -> argparse.ArgumentParser:
     p = argparse.ArgumentParser(
-        prog="ohbmcli analyze-matrix",
+        prog="aacli analyze-matrix",
         description="Stage 4 canonical analysis & annotation matrix.",
     )
     p.add_argument("--env-file", type=Path, default=Path(".env"))
@@ -692,7 +692,7 @@ def main(argv: list[str] | None = None) -> int:
         sqlite_path=args.rollup_path.with_suffix(".sqlite") if args.rollup_path else None,
         neuroscape_centroids_dir=args.neuroscape_centroids,
         code_revision=args.code_revision or _git_rev_parse_head(),
-        command_line=" ".join(sys.argv if argv is None else ["ohbmcli", "analyze-matrix"] + list(argv)),
+        command_line=" ".join(sys.argv if argv is None else ["aacli", "analyze-matrix"] + list(argv)),
         env_file=args.env_file,
         dry_run=args.dry_run,
         n_jobs=args.n_jobs,

src/abstractatlas/atlas_hosting/cli.py

@@ -4,7 +4,7 @@
 ``contracts/cli-upload-atlas-package.md`` +
 ``contracts/cli-compare-data-hosting.md``.
 
-Each subcommand exposes a ``build_*_parser`` (the top-level ``ohbmcli``
+Each subcommand exposes a ``build_*_parser`` (the top-level ``aacli``
 copies its actions in) and a ``*_main`` argv entry point returning the
 documented exit code. Typed Stage-20 failures map to distinct non-zero
 codes; secrets are never echoed.
@@ -40,7 +40,7 @@
 
 def build_upload_parser() -> argparse.ArgumentParser:
     parser = argparse.ArgumentParser(
-        prog="ohbmcli upload-atlas-package",
+        prog="aacli upload-atlas-package",
         description=(
             "Upload a built atlas-package's parquets to Cloudflare R2 under "
             "content-addressed, immutable keys; emit a registry channel entry "
@@ -89,7 +89,7 @@ def upload_main(argv: list[str] | None = None) -> int:
     parser = build_upload_parser()
     args = parser.parse_args(argv)
 
-    command_line = " ".join(["ohbmcli", "upload-atlas-package", *(argv or sys.argv[1:])])
+    command_line = " ".join(["aacli", "upload-atlas-package", *(argv or sys.argv[1:])])
 
     try:
         settings = r2_client.load_settings(args.env_file)
@@ -117,7 +117,7 @@ def upload_main(argv: list[str] | None = None) -> int:
 
 def build_compare_parser() -> argparse.ArgumentParser:
     parser = argparse.ArgumentParser(
-        prog="ohbmcli compare-data-hosting",
+        prog="aacli compare-data-hosting",
         description=(
             "Probe the Dropbox- and R2-served copies of each artifact for "
             "byte-parity, HTTP Range support, and CORS; write a pass/fail report."