CLI: export active SPL corpus and replay with fingerprint verification #27

New issue

Open

opened 2026-07-16 06:48:50 +00:00 by anuna-02 · 0 comments

anuna-02 commented

2026-07-16 06:48:50 +00:00

(Migrated from codeberg.org)

Summary

Add native export of the active SPL corpus and a replay/verification workflow for
reproducible semantic closure.

Motivation from the grounded two-machine run

The live journal contained 53 entries, but a clean semantic replay required exactly 38
active content assertions. The harness had to manually:

exclude theory/member setup entries;
exclude six retracted assertions and six retraction records;
preserve all active proponent and real machine-B statements;
create a fresh identity and theory;
assert each statement individually;
recompute and compare a custom closure fingerprint.

The replay succeeded, but reconstructing the active corpus by hand is error-prone and
not a reusable CLI contract.

Proposed CLI

For example:

elephant theory export rootclaim-grounded --active --format spl > corpus.spl
elephant theory export rootclaim-grounded --manifest export.json

elephant theory replay corpus.spl \
  --alias rootclaim-grounded-replay \
  --expect-fingerprint ee41f61d... --json

Useful export selectors could include --active, --include-meta,
--include-members, --include-retracted, and --signer. A manifest can preserve
provenance and original entry IDs without embedding non-SPL data into the replayable
corpus.

Acceptance criteria

--active --format spl emits only active, parseable SPL statements and is stable
across repeated exports of the same journal state.
Meta/member/retracted inclusion is explicit and documented.
Exported statements can be replayed one at a time into a fresh theory without custom
parsing.
Replay reports created theory ID, assertion count, failures, resulting fingerprint,
and fingerprint match in JSON.
Replay is safe to resume or uses the proposed idempotent assertion primitive.
A provenance manifest maps exported statements to source entry IDs/signers while the
semantic replay remains identity-independent.
Tests reproduce the same non-member closure under a different replay identity.

Environment / evidence

Observed with Elephant 0.1.2. The hand-maintained corpus and replay script are:

### Summary Add native export of the active SPL corpus and a replay/verification workflow for reproducible semantic closure. ### Motivation from the grounded two-machine run The live journal contained 53 entries, but a clean semantic replay required exactly 38 active content assertions. The harness had to manually: - exclude theory/member setup entries; - exclude six retracted assertions and six retraction records; - preserve all active proponent and real machine-B statements; - create a fresh identity and theory; - assert each statement individually; - recompute and compare a custom closure fingerprint. The replay succeeded, but reconstructing the active corpus by hand is error-prone and not a reusable CLI contract. ### Proposed CLI For example: ```console elephant theory export rootclaim-grounded --active --format spl > corpus.spl elephant theory export rootclaim-grounded --manifest export.json elephant theory replay corpus.spl \ --alias rootclaim-grounded-replay \ --expect-fingerprint ee41f61d... --json ``` Useful export selectors could include `--active`, `--include-meta`, `--include-members`, `--include-retracted`, and `--signer`. A manifest can preserve provenance and original entry IDs without embedding non-SPL data into the replayable corpus. ### Acceptance criteria - `--active --format spl` emits only active, parseable SPL statements and is stable across repeated exports of the same journal state. - Meta/member/retracted inclusion is explicit and documented. - Exported statements can be replayed one at a time into a fresh theory without custom parsing. - Replay reports created theory ID, assertion count, failures, resulting fingerprint, and fingerprint match in JSON. - Replay is safe to resume or uses the proposed idempotent assertion primitive. - A provenance manifest maps exported statements to source entry IDs/signers while the semantic replay remains identity-independent. - Tests reproduce the same non-member closure under a different replay identity. ### Environment / evidence Observed with Elephant 0.1.2. The hand-maintained corpus and replay script are: - https://github.com/anuna-cooperative/flf-elephant-experiment/blob/96ca63675edd88cf4982449c5f78b9441c320a39/corpus/rootclaim-grounded.spl - https://github.com/anuna-cooperative/flf-elephant-experiment/blob/96ca63675edd88cf4982449c5f78b9441c320a39/runs/2026-07-16/grounded-machineB/verify-replay.sh

No labels

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

anuna-research/elephant#27

No description provided.

Rows
Columns