Add adaptive research DAG and approve scoped ZeroMQ runtime #13

Merged
lost-rob0t merged 6 commits from agent/adaptive-research-dag into main 2026-07-19 23:54:56 +00:00
lost-rob0t commented 2026-07-19 23:50:15 +00:00 (Migrated from github.com)

Adaptive research pipeline

  • add STAR-RESEARCH-000 R&D-Agent Pipeline Analysis
  • add approved STAR-RESEARCH-PIPELINE-001 Adaptive Research DAG and Evaluation Harness
  • define append-only run and candidate ledgers
  • define bounded branch supervision, memory packets, deterministic validation, independent reviewers, score aggregation, vetoes, and one promotion gate
  • preserve the existing zero-or-one implementation-slot invariant
  • add a dedicated automatic-research project index

Runtime approvals

  • record nsaspy approval of the scoped domain-gserver architecture
  • record approval of explicit ZeroMQ boundaries and ownership rules
  • retain Sento mailboxes for ordinary local actor traffic
  • require one ZeroMQ context per process and one exclusive socket owner
  • keep Star Router above the transport layer

Binding audit

  • start SENTO-ZMQ-RESEARCH-001 PZMQ Binding Audit
  • inspect orivej/pzmq before adoption
  • use PyZMQ as a behavioral and interoperability reference, not as the Common Lisp production dependency
  • define an evidence-based adopt, fork, or replace gate
  • require source, memory ownership, context shutdown, multipart, polling, monitoring, CURVE, socket-option, SBCL, Nix, and cross-language tests

The PZMQ README itself states that much of the surface is untested and that its automatic tests are barebone, so blind adoption is explicitly rejected.

Navigation

  • link the approved designs, approval record, PZMQ audit, and automatic-research index from STAR-INDEX-000

Validation

  • branch is 6 commits ahead of main and 0 behind
  • six Org files are added or updated
  • relative links were reviewed manually
  • local scripts/sync.py --check, Pages export, and link audit were not executable in this connector-only runtime; the draft PR should run repository CI before merge
## Adaptive research pipeline - add `STAR-RESEARCH-000 R&D-Agent Pipeline Analysis` - add approved `STAR-RESEARCH-PIPELINE-001 Adaptive Research DAG and Evaluation Harness` - define append-only run and candidate ledgers - define bounded branch supervision, memory packets, deterministic validation, independent reviewers, score aggregation, vetoes, and one promotion gate - preserve the existing zero-or-one implementation-slot invariant - add a dedicated automatic-research project index ## Runtime approvals - record nsaspy approval of the scoped domain-gserver architecture - record approval of explicit ZeroMQ boundaries and ownership rules - retain Sento mailboxes for ordinary local actor traffic - require one ZeroMQ context per process and one exclusive socket owner - keep Star Router above the transport layer ## Binding audit - start `SENTO-ZMQ-RESEARCH-001 PZMQ Binding Audit` - inspect `orivej/pzmq` before adoption - use PyZMQ as a behavioral and interoperability reference, not as the Common Lisp production dependency - define an evidence-based adopt, fork, or replace gate - require source, memory ownership, context shutdown, multipart, polling, monitoring, CURVE, socket-option, SBCL, Nix, and cross-language tests The PZMQ README itself states that much of the surface is untested and that its automatic tests are barebone, so blind adoption is explicitly rejected. ## Navigation - link the approved designs, approval record, PZMQ audit, and automatic-research index from `STAR-INDEX-000` ## Validation - branch is 6 commits ahead of `main` and 0 behind - six Org files are added or updated - relative links were reviewed manually - local `scripts/sync.py --check`, Pages export, and link audit were not executable in this connector-only runtime; the draft PR should run repository CI before merge
Sign in to join this conversation.
No description provided.