Design government data acquisition actors #50

Merged

lost-rob0t merged 2 commits from design/government-data-acquisition-actors into research/us-government-data-acquisition

2026-07-28 21:08:25 +00:00

lost-rob0t commented

2026-07-28 08:46:19 +00:00

(Migrated from github.com)

Stack

Depends on #48, which adds the government-data acquisition research basis.

Summary

add STAR-RESEARCH-PIPELINE-003 Government Data Acquisition Actors
define one government-data domain gserver rather than one actor per jurisdiction
separate long-lived one-writer authorities from short-lived bounded source workers
define root and per-run supervision, dispatchers, mailbox policies, keyed routing, restart budgets, and health
add versioned Common Lisp command, event, result, checkpoint, actor-manifest, and adapter-manifest contracts
define registry, identity, domain, source-profile, fingerprint, acquisition, artifact, normalization, provenance, coverage, scheduling, and gap actors
centralize per-host, vendor, tenant, and credential rate leases and circuit breakers
require raw artifacts and resolvable provenance before canonical writes
isolate browser, OCR, archive, and heavy parser workers behind process adapters
define at-least-once delivery, idempotency, replay, cancellation, source migration, and typed failure behavior
add 57 deterministic fixtures spanning registry change, platform paging, rate limiting, crashes, provenance, and records-request authorization
update the adaptive-research index with the stacked design and government-data dependency order

Core architecture

Government units remain durable records and routing keys. Reusable actors are organized by responsibility and platform family. Adapter workers execute one immutable, bounded source attempt and cannot write canonical state.

Validation

stacked branch is 2 commits ahead of research/us-government-data-acquisition
only the actor design and adaptive-research index are changed
design remains REVIEW
no implementation or approval state is claimed

## Stack Depends on #48, which adds the government-data acquisition research basis. ## Summary - add `STAR-RESEARCH-PIPELINE-003 Government Data Acquisition Actors` - define one government-data domain gserver rather than one actor per jurisdiction - separate long-lived one-writer authorities from short-lived bounded source workers - define root and per-run supervision, dispatchers, mailbox policies, keyed routing, restart budgets, and health - add versioned Common Lisp command, event, result, checkpoint, actor-manifest, and adapter-manifest contracts - define registry, identity, domain, source-profile, fingerprint, acquisition, artifact, normalization, provenance, coverage, scheduling, and gap actors - centralize per-host, vendor, tenant, and credential rate leases and circuit breakers - require raw artifacts and resolvable provenance before canonical writes - isolate browser, OCR, archive, and heavy parser workers behind process adapters - define at-least-once delivery, idempotency, replay, cancellation, source migration, and typed failure behavior - add 57 deterministic fixtures spanning registry change, platform paging, rate limiting, crashes, provenance, and records-request authorization - update the adaptive-research index with the stacked design and government-data dependency order ## Core architecture Government units remain durable records and routing keys. Reusable actors are organized by responsibility and platform family. Adapter workers execute one immutable, bounded source attempt and cannot write canonical state. ## Validation - stacked branch is 2 commits ahead of `research/us-government-data-acquisition` - only the actor design and adaptive-research index are changed - design remains `REVIEW` - no implementation or approval state is claimed