changelog

latest updates, features, and improvements to brin.

##March 26, 2026

GitHub pull request scanning and registry support

###Features

  • Add GitHub pull request scanning and registry support so brin can assess PR author identity, diff behavior, content risk, and conditional deep review before merge

##March 6, 2026

Contributor scanning, refresh=true re-scans, fast web prelim scoring, ThreatFox blocklists, and API rate limiting

###Features

  • Add contributor scanning with Tier 1 identity scoring (account age, GPG keys, org memberships, email domain classification, followers, contributions) and Tier 2 behavior analysis (dormancy-then-spike detection, cross-repo velocity, activity recency) #60
  • Add ThreatFox as fourth blocklist source and seed domain blocklists from URLhaus, PhishTank, and OpenPhish #59
  • Add IP-based rate limiting to API (60 req/min, burst of 10) #52
  • Add LLM observability: persist full JSONL event stream from every Tier 3 invocation to Postgres #49
  • Implement refresh=true query parameter to force re-scan of stale artifacts #61
  • Upgrade web scanner with deterministic tier1/tier2 signals, WHOIS fallback, and YARA-X integration #47

###Improvements

  • Fast preliminary web scan: reduce cache-miss latency from ~2s to ~500ms #63

  • Add seed scripts for unscanned edge targets (repos, contributors, domains) #62

  • Reduce tier2 false positives for high-identity domains by suppressing FP threats when identity score >= 85 #56 #58

  • Fix YARA/regex false positives on SPAs and add patch-yara-fp rescore mode #55

  • Add backfill-domain-edges mode to connect domain entities in the graph #54

###Bug fixes

  • Fix O(n^2) graph subgraph query and add delete-entity mode #57
  • Skip inline preliminary scans in bulk endpoint to prevent 504 timeouts #50

##March 3, 2026

Priority queue for scan jobs and differentiated preliminary scoring for web scans

###Features

  • Add priority queue for scan jobs so seeds don't block API traffic #43
  • Add differentiated preliminary scoring for web scans #39

###Bug fixes

  • Fix PyPI repo URL extraction to catch all project_urls key variants #41
  • Fix panic on multi-byte UTF-8 chars when slicing package source samples #40

###Improvements

  • Add backfill scripts for orphaned crate HOSTED_AT edges #42

##March 2, 2026

Email artifacts support and repo URL parsing improvements

###Features

  • Add email artifacts support #36

###Bug fixes

  • Fix repo URL parsing: strip fragments, handle github: shorthand, accept non-GitHub repos #37

##February 27, 2026

Pub/Sub push-driven processing, stale-while-revalidate caching, and list artifacts endpoint

###Features

  • Replace worker polling loop with Pub/Sub push-driven processing #31
  • Reuse repo scan results for skills/MCPs to eliminate redundant scans #30
  • Add stale-while-revalidate caching for entity lookups #29
  • Add list artifacts endpoint with filtering, search, and pagination #28

###Bug fixes

  • Fix contributor count parsing to use last "page" parameter instead of "per_page" #30

###Docs

  • Add changelog for February 26, 2026 #27

##February 26, 2026

Tranco top 100K domain and MCP server seed data, package seed scripts for npm, PyPI, and crates.io

###Features

  • Add Tranco top 100K domain seed data #26
  • Add 2000 MCP server seed data #24
  • Add package seed scripts for bulk-seeding npm, PyPI, and crates.io registry scans #22

###Infra / DevOps

  • Switch Tier 3 LLM provider from Anthropic direct API to AWS Bedrock (Claude Opus) #22
  • Add automatic recovery for scans stuck in processing state (>10 min) on worker startup #22
  • Increase worker memory allocation from 1Gi to 2Gi to support concurrent Tier 3 scans #22

##February 25, 2026

Entity seed scripts and semaphore-bounded concurrent scan processing

###Features

  • Add entity seed scripts for bulk-populating the database with skills leaderboard data #21

###Improvements

  • Add semaphore-bounded concurrent scan processing to worker (up to 8 parallel scans) #18

##February 24, 2026

Safety tolerance levels, graph scoring via Memgraph, webhooks, and core scanners for packages, commits, repos, and skills

###Features

  • Add safety tolerance levels (conservative, lenient, yolo) for verdict interpretation #15
  • Add graph scoring via Memgraph for cross-entity threat analysis with relationship networks #14
  • Add webhook support for tier-by-tier scan progress callbacks #11
  • Add commit scanner with 3-tier pipeline and contribution history tracking #10
  • Add package scanner with 3-tier pipeline for npm, PyPI, and crates.io #6
  • Add MCP server scanner with 3-tier pipeline and typosquatting detection #5
  • Add support for repositories and skills #4

###Docs

  • Update README with comprehensive project information and API endpoint documentation #13

###Infra / DevOps

  • Initial setup of API and workers #1