generated: '2026-08-11'
method: searched
source: https://alphai.io/changelog
description: >-
  A dated, user-visible changelog covering both REST and MCP changes, newest
  first. Entries are dated rather than version-numbered; the only semantic
  version in play is info.version on the OpenAPI document (1.24.0 at capture).
  Cadence over the captured window is roughly every other day.

scheme: dated entries (YYYY-MM-DD); semantic version on the OpenAPI document only
current_version: '1.24.0'
url: https://alphai.io/changelog
feed: null
window_captured: '2026-07-15 to 2026-08-11'

entries:
  - date: '2026-08-11'
    kind: addition
    breaking: false
    highlights:
      - Per-ticker insider-trades endpoint GET /api/symbols/{ticker}/insider-trades/
      - 3-month and 12-month rollups plus a chart_events array
      - New transaction fields stake_change_pct and security_title
  - date: '2026-08-06'
    kind: fix
    breaking: false
    highlights:
      - Form 4 coverage gap closed; 1,489 filings backfilled from SEC indexes
      - Original publish timestamps restored on the backfilled rows
  - date: '2026-08-05'
    kind: improvement
    breaking: false
    highlights:
      - Company-name resolution improved; the query parameter now matches brand names (spacex -> SPCX)
  - date: '2026-07-31'
    kind: addition
    breaking: false
    highlights:
      - Economic calendar endpoint GET /api/calendar/ with stable UIDs and phase tracking
  - date: '2026-07-30'
    kind: addition
    breaking: false
    highlights:
      - Macro categories added (macro_economy, commodities, geopolitics)
      - Dedicated GET /api/news/macro/ endpoint
  - date: '2026-07-30'
    kind: improvement
    breaking: false
    highlights:
      - Event-level story clustering by meaning rather than headline overlap
      - sources_count field added
  - date: '2026-07-29'
    kind: fix
    breaking: false
    highlights:
      - Insider transaction values validated against SEC filings
      - avg_price_usd and total_value_usd corrected, or omitted when unverifiable
  - date: '2026-07-28'
    kind: change
    breaking: true
    highlights:
      - Pro archive depth capped at 180 days (previously unlimited)
      - Free stays 30 days, Basic stays 90 days
    note: >-
      The one entitlement reduction in the captured window. Announced in the
      changelog only — no deprecation window, no Sunset header.
  - date: '2026-07-27'
    kind: change
    breaking: false
    highlights:
      - Alert deduplication — one story delivers once per ticker per 24h across email, Telegram and webhooks
  - date: '2026-07-26'
    kind: addition
    breaking: false
    highlights:
      - Public status page launched at status.alphai.io with uptime and latency tracking
  - date: '2026-07-26'
    kind: addition
    breaking: false
    highlights:
      - Delta polling on the insider feed via sort=ingested
      - Trending items now carry story fields and symbol locale data
  - date: '2026-07-25'
    kind: addition
    breaking: false
    highlights:
      - Feed polling mode — sort=ingested returns rows by ingest order with cursor resumption
  - date: '2026-07-23'
    kind: improvement
    breaking: false
    highlights:
      - Share-class coverage expanded
      - 8-K filings reclassified by primary item
      - Trending ranks by a decay function, max 2 per ticker
  - date: '2026-07-20'
    kind: change
    breaking: false
    highlights:
      - page_size now accepts 1-20 on all tiers (50 on Pro)
  - date: '2026-07-15'
    kind: improvement
    breaking: false
    highlights:
      - Summary endpoints validate tickers
      - Crypto names resolve to the -USD suffixed form

summary:
  entries_captured: 15
  breaking_changes: 1
  cadence: roughly every other day over the captured window
  covers: [REST, MCP, data quality, entitlements]
  note: >-
    The changelog is notable for recording data-quality corrections (the Form 4
    backfill, the insider-value validation), not just feature adds. That is a
    materially different disclosure posture from a marketing release feed, and it
    is the right one for a data product where a silently-wrong number is the real
    failure mode.

related:
  lifecycle: lifecycle/alphaai-lifecycle.yml