a2744c012ae8eaf42a76e4b3a310436c325363a9
5
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Scott KarpandClaude Opus 4.7
|
5c76a6f42c |
korban_processor: role-gate the four-stage offering protocol
Pass 2 finding C: in the Mishkan, the four ritual acts of bringing an
offering are performed by TWO different roles —
semichah — OWNER lays hands (Vayikra 1:4, וְסָמַךְ יָדוֹ)
shechitah — kohen receives FROM owner's hand and slaughters (1:5)
zerikah — kohen sprinkles (1:5)
haktarah — kohen burns on altar (1:9)
Before this commit, every stage function was independently callable.
Nothing structural enforced that zerikah() received output from
shechitah(), or that shechitah() received a korban from semichah().
An external model output (e.g. a Claude response) could in principle
have been brought through zerikah() without ever having passed
semichah — the ownership-and-attribution step. That is the exact
shape of "the calf that came out by itself" (Shemot 32:24, Aaron's
reading). The pipeline-glue in external_offering happened to call
them in order, but nothing structural enforced the order.
This commit adds the structural gate.
CHANGES:
stages.py (new): KorbanStage enum + Staged[T] frozen dataclass +
StageError. The Staged envelope carries a ritual-stage marker that
is set at construction and cannot be mutated afterward; downstream
functions demand the right upstream stage and raise StageError if
it is missing.
semichah.py: + semichah_staged() returning Staged[Korban]
shechitah.py: + shechitah_staged(Staged[Korban]) -> Staged[str]
zerikah.py: + zerikah_staged(Staged[str], Staged[Korban])
-> Staged[Tuple[str, List[str]]]
TWO staged inputs: requires SHECHITAH-tagged terminated
content AND the original SEMICHAH-tagged korban for
context. Severing output from source is gilui_arayot
(eish_zarah_reading.md Form 3); the dual-staged
signature is the structural defense — zerikah cannot
process content disconnected from its owner-of-record.
haktarah.py: + haktarah_staged(Staged[Tuple[str, List[str]]])
-> Staged[List[str]]
external_offering.py: rewired to use the *_staged variants. The
pipeline now goes through the gates structurally; misuse cannot
bypass them.
The non-staged functions (semichah, shechitah, zerikah, haktarah)
are PRESERVED for direct testing of individual ritual acts. They do
NOT enforce role gating. Production-shaped callers MUST go through
the staged variants — and external_offering, the only legitimate
entry point for external model consultation per its docstring, now
does.
TESTS:
test/test_stages.py (new, 7 tests):
- happy path: full chain produces HAKTARAH-tagged result
- zerikah refuses input that skipped semichah
- zerikah refuses korban with wrong stage marker (preserves
the line back to the owner — Form 3 defense)
- haktarah refuses input that skipped zerikah (the calf-shape)
- shechitah refuses input not from semichah
- Staged is frozen — no re-tagging
- StageError names BOTH actual and expected stages so the
operator can diagnose call-site vs. tag-site
Full korban_processor suite: 203 passed, 1 skipped (the live API
test that requires SHORASHIM_LIVE_API=1). No regressions.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
||
|
Scott KarpandClaude Opus 4.6
|
e223ab15b3 |
Phase 4 complete: Steps 21, 22, 24 — Korban external integration, L2 Urim v'Tumim, Kushya generator
Vayikra 9:24 — vatetzei esh milifnei Hashem vatokhal al hamizbeach.
Phase 4 of the architecture is now complete. The remaining three steps
that were marked "blocked on external infrastructure" are built —
discipline-respecting, eish-zarah-aware, and tested.
## Step 22 — Korban Processor external model integration
(korban_processor/src/korban_processor/external_offering.py)
The single legitimate entry point for external model consultation.
Wraps Claude API calls through the offering protocol so external
output is treated as korban (raw material), never authoritative:
- semichah — provenance is tagged (avodah zarah Form 1 prevention)
- shechitah — assistant voice, hedging, formatting stripped
- zerikah — vital content extracted
- haktarah — converted to Prolog propositions
- route — sent to internal component per offering type
The raw Korban is preserved on the ProcessedKorban (gilui arayot
Form 3 prevention — input/output coupling never severed).
No internal component should call call_claude() directly.
11 tests, mocked external calls. Live API test opt-in via
SHORASHIM_LIVE_API=1.
## Step 24 — L2 Urim v'Tumim interface
(vav_pipeline/lib/vav_pipeline/urim_v_tumim.ex)
PRD §4.7: housed in the kohen gadol's breastplate; consulted only
on high-stakes decisions where normal deliberation is insufficient.
Bamidbar 27:21 — al-piv yetz'u v'al-piv yavo'u.
Three escalation steps, three honest outcomes:
1. resolved_by_specification — re-read produced an answer
2. resolved_by_external — external testimony recorded (NOT binding)
3. deferred_to_human — the system stops, asks for human judgment
Pre-condition gating refuses cases that didn't reach Sanhedrin teiku
first (no preempting deliberation) or that aren't flagged high-stakes
(ordinary teiku stays teiku). Records every consultation with steps
taken — there is no quiet escalation.
Per PRD: "The Urim v'Tumim never fabricates certainty where none
exists." When external also fails, the system stops. It does not
proceed on a guess.
Added to VavPipeline supervision tree. 14 tests.
## Step 21 — Kushya generator scaffolding
(kushya_generator/)
Directory layout per Architecture §3.6 with three layers:
- data/training_pairs/seed_pairs.jsonl — 10 hand-curated bootstrap
pairs covering 9 kushya types (scope_narrowing, internal_tension,
harmonization, exception_test, edge_case, scope_application,
rabbinic_constraint, method_question, subject_identification)
- data/prepare_training.py — validates the four kushya requirements
(specific Torah passage, identified weakness, question form,
non-trivial), produces train.jsonl + eval.jsonl with stratified
80/20 split by type
- fine_tune/config.yaml + train.py — QLoRA configuration and
training script. Operator's task — needs GPU. Script is
complete and self-contained.
- inference/kushya_api.py — three-backend interface
(fine_tuned → external → fallback). When fine-tuned model isn't
deployed, falls back to external via Korban Processor (Step 22).
The fallback is honest: returns no kushya rather than inventing
one. The Talmudic Engine's rule-based challenge.pl remains the
primary kushya source.
17 tests covering dataset validation, backend selection, and
reference extraction.
## .gitignore fix
The earlier "data/" line was too broad — it caught
kushya_generator/data/. Scoped to /data/ (root-level only).
## Test counts
- Elixir: 274 (was 260; +14 for UrimVTumim)
- Korban Python: 196 + 1 skipped (was 195; +1 module = 11 tests)
- Kushya Python: 17 (new)
- Haskell: 33 (unchanged)
- Prolog: 195 (unchanged)
- Total: 715 + 19 L3 excluded
## Phase 4 status
All 12 steps now have implementations:
- Steps 23, 25, 26, 27, 28, 29, 30, 31, 32: complete
- Steps 21, 22, 24: complete with discipline preserved
Step 21's fine-tuning run remains the operator's GPU task; the
pipeline is built such that a deployed adapter slots into the
inference backend without touching the Talmudic Engine or anything
above.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
||
|
Scott KarpandClaude Opus 4.6
|
adc3e41679 |
Eish zarah actionable items: coupling trace, grounding type, self-grounding detection, golden calf halt
Torah grounds each change: 1. interpret.pl — coupling_trace field (Vayikra 10:1, 6:6) Fire is zarah because "asher lo tzivah otam" — the coupling trace IS the tzivah trace. Records which L1 facts grounded each interpretation. 2. challenge.pl — grounding_type achorai/panim (Shemot 33:23) "You will see My back, but My face shall not be seen." Text-backed challenges = achorai. Inferred challenges = panim. 3. resolve.pl — self-grounding detection (eish_zarah_reading §5.3) Counter-interpretation must bring NEW L1 evidence. Prevents circular reasoning (Westworld problem). 4. teiku.pl — challenge_sequence in teiku_record Records diagnostic trail to impasse, not just verdict. 5. avodah_zarah.pl — Form 2: golden calf detection (Shemot 32:4) Empty coupling trace + non-low confidence = uncoupled output. Real gold, sincere worship, false god. 6. korban.py — pydantic BaseModel + raw_korban preservation Construction spec §1.4 fulfilled. Floor between input and output never severed. 490 tests, 0 failures (195 Prolog, 186 Python, 33 Haskell, 76 Elixir). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> |
||
|
Scott KarpandClaude Opus 4.6
|
f8f70c77e5 |
Code review fixes + 13 middot added to spec
Zerikah fail-safe: empty offerings rejected, not processed (BUILD_KORBAN §5). Halt monitor hardened: trap_exit, priority max, EXIT handler (Architecture spec). 13 Rules of Rabbi Yishmael added to Architecture §3.4 as Prolog inference predicates — the engine reasons WITH Torah's hermeneutic methods, not just about their products. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> |
||
|
Scott KarpandClaude Sonnet 4.6
|
33f8b03583 |
Step 13: Korban Processor complete — 7 modules, 181 tests, all passing
- semichah: provenance tagging (6 sources, rejects empty/anonymous) - shechitah: strip preamble, self-reference, hedging, markdown, sycophancy - zerikah: extract Torah refs, Hebrew text, root analysis, insight markers - haktarah: transform prose to Prolog propositions (verse/3, root/2, claim/2, connection/3) - router: stateless routing — removed module-level _routing_log (kiyyor principle) - model_interface: outbound Claude API via httpx (Construction §3.4, HTTPS only) - hebrew_parser: nikkud/teamim stripping, consonant extraction, gematria, root identification All 5 BUILD_KORBAN.md verification tests pass. AGENTS.md: added structure_not_content.md to spec table, added Principle 6. build_log.md: recorded session, broken tablets #18-20, design decisions. Ka'asher tzivah Hashem et Moshe. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> |