RAGE #12: bounded indexed classifier query + restart #28

Closed

lost-rob0t wants to merge 0 commits from rage/12-classifier-query-red into main

lost-rob0t commented

2026-08-31 10:14:34 +00:00

(Migrated from github.com)

Consumes the next bounded acceptance slice under #12 after merged classifier substrate PR #27.

Canonical ARADR: research/LLM-LOG-RESEARCH-037-request-classifier-indexed-query-restart.org.
Implementation evidence: evidence/LLM-LOG-EVIDENCE-018-request-classifier-indexed-query-restart.org.

Executed RED: 2b95c54dd4a1436f06eccbc11e8b6d48cfb512da, hosted run 33381478690. The missing query_classification_history operation and missing explicit ambiguity semantics failed as intended before production mutation.

Current exact head: 94c8980cb306f4a6c3964de7617eaf8ffb71974a.
Hosted run 33387423599 is GREEN on that exact head across Common Lisp/Tek9/SWI-Prolog integration, Python expert-policy/adapter/transport/launch, and Home Manager expert wiring.

Behavior in this slice:

Common Lisp owns typed bounded history-query validation, Tek9 index registration/access, provenance joins, and result shaping.
SWI-Prolog owns explicit ambiguity rules; no Python or Common Lisp rule engine was added.
history retrieval requires a finite limit and a selective request/user-message/provider/model equality filter;
retrieval uses existing bounded Tek9 secondary-index range APIs and never falls back to whole-corpus scans;
classifier history survives expert-host restart and preserves source/rule/expert provenance;
empty unbounded history queries are rejected;
ambiguity remains analytical metadata and grants no authority.

Tek9 repository changes are not authorized or required: the pinned public API is sufficient. No dotfiles consumer mutation is included in this PR.

#12 remains open after this bounded slice: timestamp/time-range history acceptance still needs its own RED contract before the classifier issue can be considered complete.

Consumes the next bounded acceptance slice under #12 after merged classifier substrate PR #27. Canonical ARADR: `research/LLM-LOG-RESEARCH-037-request-classifier-indexed-query-restart.org`. Implementation evidence: `evidence/LLM-LOG-EVIDENCE-018-request-classifier-indexed-query-restart.org`. Executed RED: `2b95c54dd4a1436f06eccbc11e8b6d48cfb512da`, hosted run 33381478690. The missing `query_classification_history` operation and missing explicit ambiguity semantics failed as intended before production mutation. Current exact head: `94c8980cb306f4a6c3964de7617eaf8ffb71974a`. Hosted run 33387423599 is GREEN on that exact head across Common Lisp/Tek9/SWI-Prolog integration, Python expert-policy/adapter/transport/launch, and Home Manager expert wiring. Behavior in this slice: - Common Lisp owns typed bounded history-query validation, Tek9 index registration/access, provenance joins, and result shaping. - SWI-Prolog owns explicit ambiguity rules; no Python or Common Lisp rule engine was added. - history retrieval requires a finite limit and a selective request/user-message/provider/model equality filter; - retrieval uses existing bounded Tek9 secondary-index range APIs and never falls back to whole-corpus scans; - classifier history survives expert-host restart and preserves source/rule/expert provenance; - empty unbounded history queries are rejected; - ambiguity remains analytical metadata and grants no authority. Tek9 repository changes are not authorized or required: the pinned public API is sufficient. No dotfiles consumer mutation is included in this PR. #12 remains open after this bounded slice: timestamp/time-range history acceptance still needs its own RED contract before the classifier issue can be considered complete.

nsaspy changed title from ~~WIP: RAGE #12: bounded indexed classifier query + restart~~ to RAGE #12: bounded indexed classifier query + restart

2026-09-04 14:35:42 +00:00

nsaspy referenced this pull request

2026-09-04 14:35:42 +00:00

RAGE #12: bounded indexed classifier query + restart #29