FIN Central

FIN Daily digest — 2026-07-01

4 min read

FIN Daily digest — 2026-07-01

Yesterday’s push in fin-agentic was a concentrated reliability pass: the team tightened RP581 QC replay and queue authority from several angles while also shoring up deploy and CI paths that support those flows. Building on the prior day’s momentum in the same repo, the most consequential work was the cluster of RP581 fixes that clarified source-of-truth behavior, provenance requirements, and replay handling so QC can trust what it processes. Around that core, the day also advanced comms and knowledge infrastructure, with new Company Brain MS Graph wiring and proof work landing alongside seed compatibility fixes.

Highlights

  • A substantial RP581 hardening sequence landed across QC and deploy, centering queue authority on provenance and authoritative PDF rows while protecting the source of truth in deployment paths: #1679, #1678, and #1675.

  • Replay behavior for RP581 was tightened further, with fixes for QC extraction without queue materialization and for tail-drain plus arrivals queue routing—important plumbing for making replayed document flows behave predictably end to end: #1683 and #1676.

  • The team also made QC fallback behavior more explicit by marking the load-expected path provisional, a small but meaningful signal that reduces ambiguity when the primary path is unavailable: #1682.

  • Delivery infrastructure got sturdier too: deploy now accepts same-tree proof attestations, and staging eval seeding no longer depends on psql, reducing friction in verification and environment bring-up: #1680 and #1672.

  • On the data and model side, tenant seeding was updated to write knowledge embedding model metadata compatibly, while QC’s vision training corpus contract was hardened—work that improves both environment consistency and training-data discipline: #1671, #1670, and #1668.

  • Beyond reliability work, fin-agentic expanded its comms and knowledge surface by wiring Company Brain into MS Graph intelligence and recording proof artifacts for the merged comms phase, continuing the platform’s push toward richer organizational context: #1667, #1677, and #1666.

Work started

RepoPRTitleAuthorLabels
fin-platform/fin-agenticfin-platform/fin-agentic#1683fix(qc): replay RP581 extraction without queue materializationrobertolouisgarcia-makerrun-heavy-ci
fin-platform/fin-agenticfin-platform/fin-agentic#1682fix(qc): mark load-expected fallback provisionalrobertolouisgarcia-maker—
fin-platform/fin-agenticfin-platform/fin-agentic#1681feat(comms): read merged conversation identitiesrobertolouisgarcia-maker—
fin-platform/fin-agenticfin-platform/fin-agentic#1680fix(ci): accept same-tree proof attestations for deployrobertolouisgarcia-maker—
fin-platform/fin-agenticfin-platform/fin-agentic#1679fix(qc): require RP581 provenance for queue authorityrobertolouisgarcia-maker—
fin-platform/fin-agenticfin-platform/fin-agentic#1678fix(qc): keep RP581 PDF queue rows authoritativerobertolouisgarcia-maker—
fin-platform/fin-agenticfin-platform/fin-agentic#1677docs(proof): record merged master comms phase0 proofrobertolouisgarcia-maker—
fin-platform/fin-agenticfin-platform/fin-agentic#1676Fix RP581 PDF replay tail drain and arrivals queue routingrobertolouisgarcia-maker—
fin-platform/fin-agenticfin-platform/fin-agentic#1675fix(deploy): guard RP581 queue source of truthrobertolouisgarcia-maker—
fin-platform/fin-agenticfin-platform/fin-agentic#1674Persist RP581 PDF authority replay fixes on masterrobertolouisgarcia-maker—
fin-platform/fin-agenticfin-platform/fin-agentic#1673Fix RP581 PDF authority replay for QC queuerobertolouisgarcia-maker—
fin-platform/fin-agenticfin-platform/fin-agentic#1672fix(ci): seed staging evals without psqlrobertolouisgarcia-maker—
fin-platform/fin-agenticfin-platform/fin-agentic#1671fix(seed): write knowledge embedding version compatiblyrobertolouisgarcia-maker—
fin-platform/fin-agenticfin-platform/fin-agentic#1670fix(seed): write knowledge embedding model in tenant seedrobertolouisgarcia-maker—
fin-platform/fin-agenticfin-platform/fin-agentic#1669fix(ci): force eval infra persona seedsrobertolouisgarcia-maker—
fin-platform/fin-agenticfin-platform/fin-agentic#1668feat(qc): harden vision training corpus contractrobertolouisgarcia-maker—
fin-platform/fin-agenticfin-platform/fin-agentic#1667Wire Company Brain MS Graph intelligencerobertolouisgarcia-maker—

Work completed

Merged pull requests

RepoPRTitleAuthor
fin-platform/fin-agenticfin-platform/fin-agentic#1683fix(qc): replay RP581 extraction without queue materializationrobertolouisgarcia-maker
fin-platform/fin-agenticfin-platform/fin-agentic#1682fix(qc): mark load-expected fallback provisionalrobertolouisgarcia-maker
fin-platform/fin-agenticfin-platform/fin-agentic#1680fix(ci): accept same-tree proof attestations for deployrobertolouisgarcia-maker
fin-platform/fin-agenticfin-platform/fin-agentic#1679fix(qc): require RP581 provenance for queue authorityrobertolouisgarcia-maker
fin-platform/fin-agenticfin-platform/fin-agentic#1678fix(qc): keep RP581 PDF queue rows authoritativerobertolouisgarcia-maker
fin-platform/fin-agenticfin-platform/fin-agentic#1677docs(proof): record merged master comms phase0 proofrobertolouisgarcia-maker
fin-platform/fin-agenticfin-platform/fin-agentic#1676Fix RP581 PDF replay tail drain and arrivals queue routingrobertolouisgarcia-maker
fin-platform/fin-agenticfin-platform/fin-agentic#1675fix(deploy): guard RP581 queue source of truthrobertolouisgarcia-maker
fin-platform/fin-agenticfin-platform/fin-agentic#1674Persist RP581 PDF authority replay fixes on masterrobertolouisgarcia-maker
fin-platform/fin-agenticfin-platform/fin-agentic#1672fix(ci): seed staging evals without psqlrobertolouisgarcia-maker
fin-platform/fin-agenticfin-platform/fin-agentic#1671fix(seed): write knowledge embedding version compatiblyrobertolouisgarcia-maker
fin-platform/fin-agenticfin-platform/fin-agentic#1670fix(seed): write knowledge embedding model in tenant seedrobertolouisgarcia-maker
fin-platform/fin-agenticfin-platform/fin-agentic#1669fix(ci): force eval infra persona seedsrobertolouisgarcia-maker
fin-platform/fin-agenticfin-platform/fin-agentic#1668feat(qc): harden vision training corpus contractrobertolouisgarcia-maker
fin-platform/fin-agenticfin-platform/fin-agentic#1667Wire Company Brain MS Graph intelligencerobertolouisgarcia-maker
fin-platform/fin-agenticfin-platform/fin-agentic#1666fix(comms): prove phase0 mobile native detailrobertolouisgarcia-maker

By the numbers

  • PRs opened (work started): 17
  • PRs merged (work completed): 16
  • Issues closed (work completed): 0
  • Commits (supporting signal): 16
  • Repositories touched: 1
RepoOpenedMergedClosedCommits
fin-platform/fin-agentic1716016

Previous period: github_activity_digest_daily_2026_06_30.

Provenance

Auto-generated by the scripts/maint/generate-gh-digest@1.0.0 projector from gh query output.

  • Owner scope: fin-platform (all active FIN repos)
  • Window: 2026-07-01T00:00:00Z → 2026-07-02T00:00:00Z
  • Sources: github:prs.opened, github:prs.merged, github:issues.closed, github:commits
  • Lede + Highlights: LLM editor (LLMProvider)
  • Regenerate: pnpm digest:gh --type daily --date 2026-07-01

Graph View