ProPR Gets Memory-Aware Reviews, Token Reporting, and ProCursor Source Visibility
- Meister Dev Team
- 06 Apr, 2026
The latest ProPR release makes AI review behavior more grounded in your team’s actual decisions. In release/1.0.0.alpha.0005, ProPR learns from resolved threads, suppresses repeated findings on incremental PR updates, expands token and protocol visibility, and makes ProCursor knowledge sources far easier to inspect.
This is not a cosmetic release. It changes how the reviewer behaves over time, how administrators reason about what the system did, and how confidently teams can audit review and knowledge activity in a long-lived deployment.
Reviews That Improve With Context
The biggest shift in this release is that ProPR now treats resolved review decisions as semantic memory instead of disposable text.
- Resolved review threads are stored as embeddings and reused during later reviews.
- Admin dismissals now flow into the same semantic memory pipeline instead of relying on brittle string suppression.
- Incremental reviews normalize locations and use similarity checks so updated PRs do not keep reposting the same concern in different words.
That combination makes reviews more stable. Teams get fewer repeated false positives, and the reviewer can reconsider a finding in light of what the team already accepted, rejected, or explained before.
Thread Memory now includes an activity log so administrators can audit how memory records are stored, skipped, or removed.
Pull Request Rollups That Explain The Whole Review
This release also makes ProPR far easier to inspect once reviews are running.
- A new PR Review view aggregates all jobs, token totals, model-tier usage, and memory contributions for a single pull request.
- The pull-request rollup links directly into the review passes that produced those numbers.
- Memory counts now sit alongside job counts and token totals, so teams can see when review output was shaped by prior decisions.
The result is a tighter loop between review output and operational confidence. You can move from the pull request summary to the exact review pass and then down to the specific memory or trace data that influenced it.
The new PR Review view brings jobs, token tiers, and memory signals together at the pull-request level.
Review Details And Token Usage In The Same Operational Story
Once a pull request rollup tells you something important happened, ProPR now gives you clearer evidence for why.
- Job Protocols expose review summaries, findings, and file-level detail for each pass instead of stopping at a status badge.
- The Tokens & Usage dashboard shows reviewer traffic over time, which makes rollout planning and model governance more concrete.
- Review telemetry and knowledge telemetry now fit into the same administrative loop instead of living in separate blind spots.
Job Protocols let teams move from a PR rollup to the summary and evidence for an individual review pass.
Reviewer token dashboards make long-running rollout behavior visible instead of leaving cost and load as guesswork.
ProCursor Sources Are Now Easier To Trust
ProPR now includes a knowledge-indexing service for reviewer workflows, but the important change is not just that sources exist. It is that those sources are now visible and auditable.
- Repository and wiki sources expose freshness, branch coverage, symbol mode, and source-level usage rollups.
- Per-source token totals and recent snapshots show which knowledge bases are doing real work.
- Safe recent-event inspection lets administrators audit retrieval and indexing activity without exposing prompt or response content.
ProCursor source pages now show freshness, model usage, recent snapshots, and a direct path into safe event inspection.
Source Events gives teams a safe operational log of indexing and retrieval calls, including timestamps, models, token counts, and document references.
Safer Secret Handling And Cleaner Configuration Boundaries
This release protects client-managed secrets before they are stored and ensures the admin UI never returns them in plain text. Existing credentials are backfilled during migration so teams do not need to re-enter every secret at rollout.
At the same time, ProPR removes legacy global embedding fallbacks and tightens permissions around client administration. That gives each client a clearer, safer configuration boundary for AI connections, embeddings, and review behavior.
Warning
Upgrade note. Apply the included database migrations, persist the ASP.NET Core Data Protection key ring, and verify that each client has an embedding connection configured before you enable the new memory-aware reconsideration flow in production.
Rollout Checklist
- Apply database migrations before or during deployment.
- Persist the ASP.NET Core Data Protection key ring so stored secrets remain decryptable.
- Verify automatic migration of existing dismissal records into semantic memory.
- Validate per-client embedding configuration, because legacy global embedding fallbacks are no longer used.
- Smoke-test review posting and incremental re-review behavior on a staging pull request.
- Review the new Thread Memory, PR Review, Tokens & Usage, and ProCursor source event screens after rollout.
If you want the full technical detail, the complete release notes are available on GitHub. If you want the product-level view, the updated ProPR page now walks through the new memory, observability, and configuration capabilities in one place.