Files
bikoret/vav_pipeline/test/urim_v_tumim_test.exs
T
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>
2026-04-23 17:12:03 -04:00

196 lines
6.8 KiB
Elixir

defmodule VavPipeline.UrimVTumimTest do
@moduledoc """
Tests for L2 — Urim v'Tumim (Architecture §2, PRD §4.7).
The Urim v'Tumim is the decision oracle interface, consulted only on
high-stakes teiku cases the Sanhedrin cannot resolve. It NEVER fabricates
certainty — it returns one of 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, human judgment required
Bamidbar 27:21 — al-piv yetz'u v'al-piv yavo'u.
By the Urim's word the people go out and come in. Not the system's word.
"""
use ExUnit.Case, async: false
alias VavPipeline.UrimVTumim
setup do
# Reset state by replacing hooks with defaults each test.
GenServer.call(UrimVTumim, {:set_hook, :specification_consult, &UrimVTumim.default_specification_consult/1})
GenServer.call(UrimVTumim, {:set_hook, :external_consult, &UrimVTumim.default_external_consult/1})
:ok
end
defp valid_input(overrides \\ %{}) do
Map.merge(
%{
case_id: "case_#{:rand.uniform(1_000_000)}",
question: "When two principles conflict, which takes precedence?",
context: %{torah_refs: ["Shemot 35:3", "Vayikra 18:5"]},
stakes: :high,
sanhedrin_teiku: true
},
overrides
)
end
# ============================================================
# Pre-condition gating
# ============================================================
describe "pre-condition gating" do
test "refuses cases that did not reach Sanhedrin teiku first" do
# Urim is consulted AFTER teiku, never instead of Sanhedrin.
input = valid_input(%{sanhedrin_teiku: false})
assert {:error, :not_eligible} = UrimVTumim.consult(input)
end
test "refuses ordinary-stakes cases" do
# Only national-importance cases escalate to the Urim.
input = valid_input(%{stakes: :ordinary})
assert {:error, :not_eligible} = UrimVTumim.consult(input)
end
test "refuses missing case_id" do
input = valid_input() |> Map.delete(:case_id)
assert {:error, _} = UrimVTumim.consult(input)
end
test "refuses empty question" do
input = valid_input(%{question: ""})
assert {:error, :invalid_input} = UrimVTumim.consult(input)
end
end
# ============================================================
# Outcome 1: resolved by specification
# ============================================================
describe "specification re-read" do
test "specification consult resolves the case" do
GenServer.call(UrimVTumim, {:set_hook, :specification_consult,
fn _input ->
{:resolved, %{verses: ["Shemot 35:3"], reading: "Shabbat is absolute"}}
end
})
input = valid_input()
assert {:ok, {:resolved_by_specification, evidence}} = UrimVTumim.consult(input)
assert evidence.reading == "Shabbat is absolute"
end
test "specification resolution is recorded with steps_taken=[:specification]" do
GenServer.call(UrimVTumim, {:set_hook, :specification_consult,
fn _input -> {:resolved, %{ok: true}} end
})
input = valid_input()
{:ok, _} = UrimVTumim.consult(input)
{:ok, record} = UrimVTumim.get_record(input.case_id)
assert record.steps_taken == [:specification]
end
end
# ============================================================
# Outcome 2: resolved by external (testimony, not ruling)
# ============================================================
describe "external consultation" do
setup do
GenServer.call(UrimVTumim, {:set_hook, :external_consult,
fn _input ->
{:ok, %{
provenance: :claude,
reading: "Pikuach nefesh overrides — testimony only",
propositions: ["overrides(pikuach_nefesh, shabbat_lo_taaseh)"]
}}
end
})
:ok
end
test "external consultation succeeds when specification fails" do
input = valid_input()
assert {:ok, {:resolved_by_external, consultation}} = UrimVTumim.consult(input)
assert consultation.provenance == :claude
end
test "steps_taken records both specification attempt and external consult" do
input = valid_input()
{:ok, _} = UrimVTumim.consult(input)
{:ok, record} = UrimVTumim.get_record(input.case_id)
assert record.steps_taken == [:specification, :external]
end
end
# ============================================================
# Outcome 3: deferred to human (the honest refusal)
# ============================================================
describe "deferral to human" do
test "defers when both specification and external fail" do
# Both hooks return their default failure modes.
input = valid_input()
assert {:ok, {:deferred_to_human, reason}} = UrimVTumim.consult(input)
assert is_binary(reason)
assert reason =~ "Human judgment required"
end
test "deferred record has steps_taken with :deferred at the end" do
input = valid_input()
{:ok, _} = UrimVTumim.consult(input)
{:ok, record} = UrimVTumim.get_record(input.case_id)
assert record.steps_taken == [:specification, :external, :deferred]
end
test "the Urim refuses to fabricate certainty (PRD 4.7)" do
# When both consultations fail, the Urim does NOT return any
# invented answer — it deliberately stops and asks for human judgment.
input = valid_input()
{:ok, outcome} = UrimVTumim.consult(input)
assert match?({:deferred_to_human, _}, outcome)
end
end
# ============================================================
# Records & stats
# ============================================================
describe "records and stats" do
test "consultations are recorded in chronological order" do
before_count = length(UrimVTumim.records())
input1 = valid_input(%{case_id: "rec_test_1"})
input2 = valid_input(%{case_id: "rec_test_2"})
{:ok, _} = UrimVTumim.consult(input1)
{:ok, _} = UrimVTumim.consult(input2)
records = UrimVTumim.records()
assert length(records) >= before_count + 2
ids = Enum.map(records, & &1.case_id)
assert "rec_test_1" in ids
assert "rec_test_2" in ids
end
test "stats track outcomes" do
stats = UrimVTumim.stats()
assert is_integer(stats.total)
assert is_integer(stats.resolved_by_specification)
assert is_integer(stats.resolved_by_external)
assert is_integer(stats.deferred_to_human)
assert stats.total ==
stats.resolved_by_specification + stats.resolved_by_external +
stats.deferred_to_human
end
test "get_record returns :not_found for unknown case_id" do
assert :not_found = UrimVTumim.get_record("nonexistent_xyz")
end
end
end