main
3
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Scott KarpandClaude Fable 5.1
|
c0d0555aed |
the name: Bikoret — code, contracts, images
Headers X-Bikoret-Key / -Rebuke / -Verdict / -Bedikah; the `bikoret` object and the "— bikoret …" fringe on a gate reply; environment variables BIKORET_*; the image user and paths (/home/bikoret, /tmp/bikoret, /var/lib/bikoret/manna, /etc/bikoret/secrets, /opt/bikoret/spec); the tmp-bikoret pod volume; local databases bikoret_dev/_test. Kubernetes resource names, the shorashim_app role and the domain are unchanged (Tamid §1.1, "the names that still carry the old word"). Doorpost c963c4fe…. Elixir 604/0, L3 10/0, korban 295, Prolog bedikah 44/44. Gitea repo renamed heaviness510/bikoret (old URL redirects). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> |
||
|
Scott KarpandClaude Fable 5.1
|
f4e8e5afc8 |
specs §1.1/§1.5 + Doorpost: the pins say what is built, and the build obeys them
The pin audit (2026-10-04) found Construction §1 wrong in three ways.
1. §1.5's mix.exs block was not the real mix.exs.
phoenix ~> 1.7.11 was never a dependency (the API is Plug on
Bandit); `ecto` is `ecto_sql`; telemetry is transitive (locked
1.4.2, outside the stated ~> 1.2.1); bandit, plug and the sanhedrin
path dependency were missing. The block now matches mix.exs and adds
the versions mix.lock actually builds. Terumah's Elixir table
matched the wrong block and also said postgrex ~0.17.4; corrected.
2. §1.1's component table described a machine that is not the one
running. PostgreSQL 16.2 via apt → 17, managed (17.11 today).
Podman 4.9.3 → Kubernetes/containerd, managed; Podman is used
nowhere. Ubuntu 24.04.1 → what each image is in fact built on
(Ubuntu 22.04 for pipeline and engine, Debian 12 for korban and
talmudic) and the managed node (24.04). rebar3 3.23.0 → 3.22.0, the
version in the build image. Azazel Processor added as a component.
The spec is corrected to the deployment, not the reverse.
3. §1.1's rule — "No >=. No +. No ranges." — was not true of the spec's
own Elixir deps, and "lock files are committed" was not true of two
components. Both halves fixed:
- The rule now says what is the case: a manifest may give a range;
the committed lock is the pin; every image is built from its
lock; base images are pinned by tag; managed services are pinned
to a major version and patched by the provider.
- And the build was made to obey it. The Korban image installed
three packages named by hand and let pip pick every transitive
one on the day of the build, plus an unpinned poetry-core. It now
installs the lock's main group by exact version AND hash
(scripts/lock_to_requirements.py; --require-hashes --no-deps;
pip check). The image holds the lock's 12 packages and nothing
else.
- azazel_processor/poetry.lock: generated (it had none).
- kushya_generator/pyproject.toml + poetry.lock: the four §1.4 pins
(llama-cpp-python 0.2.55, peft 0.9.0, transformers 4.38.2,
datasets 2.17.1) were pinned nowhere in the repo. They are now
declared and locked (74 packages). The lock resolves; it has NOT
been installed or run — Step 21 is unbuilt and needs a GPU.
Its test config gains the korban and azazel source paths, which
the external backend needs: 3 tests there had been failing since
the azazel build, unnoticed because that suite is in no total.
Doorpost @expected_hash 39b66c31… → 1cd07079…, in this commit, with the
witnessed file that moved it (vayakhel).
Read before committing: Elixir 474/0 (seeds 0, 42) · korban 222 passed
1 skipped and azazel 125 passed, each with dependencies installed from
its lock by hash · kushya 17 passed (was 14 of 17) · korban image
starts, imports korban_processor + azazel_processor, pip check clean ·
pipeline image reports specification_identity verified 1cd07079….
Left as written: architecture §2's technology table still names
Phoenix, PostgreSQL 16, Ubuntu 24.04 and rootless Podman for Sanhedrin
isolation, and terumah §B still lists systemd, squashfs, Caddy,
nftables for a self-hosted server. Those describe a design, not a pin.
Co-Authored-By: Claude Fable 5.1 <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>
|