Release notes

reconstruct flattened markdown tables across 15 blog pages (#1802)

single Astro homepage + unified collapsible FAQ + credit copy (#1801)

simplify homepage hero desc — "who's already evaluating your category" (#1798)

refresh live homepage hero — GitHub-intent positioning (#1792)

homepage CTA → Book a free consultation with founder + new booking link (#1785)

MarketingContentPage redesign — TOC sidebar + TLDR callout (#1782)

only main-CI triggers production deploy (kill dedupe-skip spam) (#1790)

design-coherence phase 2 — landing semantic-class system + tokens + tests (#1775)

delete unused build-image.yml + simplify runner-policy guard (#1789)

prevention-first quality gates (shift-left, ubuntu-latest mirror) (#1783)

restore prod version + in-app release-notes layout

move deploy chain to ubuntu-latest, retire self-hosted deploy pool (#1787)

self-serve paid-conversion hardening (checkout errors, coupon UI, event idempotency, billing edges)

alternatives converting pSEO template + phantombuster/scarf pages (#1755)

CUTOVER — production landing deploys Astro dist/ (#1707)

tools schema parity (26) + legal pages — 135/162 full parity (#1706)

send discover/admin batch deanon via Inngest instead of dead CF worker (#1754)

migrate /who-uses (30) and /blog (23) families to v2 templates (#1704)

org activity realtime channel — contract, publisher, useOrgActivity hook (slice 1) (#1742)

migrate /vs (9) and /glossary (8) families to v2 templates (#1703)

staging Inngest env guard + registration script (keys = manual mint) (#1741)

pipeline event envelope {product, apiKeyId, plan} in shared types (#1740)

enable Otta Pulse gates — acceptance + review-threads (#1750)

align deanon page layout with app shell

pipeline watchdog cron — ntfy+Sentry alert on stale funnel stages (#1730)

read array-shaped deanon cache rows

record API deanon provider costs

deliver notifications via Inngest instead of CF queue round-trip (closes #LC-828) (#1726)

widen deploy health-verify window + builder acceptance-fence template (#1731)

ICP re-score UI picks up terminal job state + numeric skipped counts (#1727)

build server image once in CI and push to GHCR (phase 1) (#1732)

send onboarding deanon via Inngest events instead of dead CF worker (#1725)

leads empty-state action hub + discover batch feedback (#1709)

Reply.io LinkedIn channel — linkedInUrl on push + LinkedIn webhook events (#1708)

Reply.io KPI metrics (pushed/delivered/opens/replies/bookings/opt-outs) (#1643)

explicit discover view param wins + fix(LC-808): settings user-menu dead click (#1683)

paywall hardening (plan-name crash, playbook gate, redaction leak, 403 shape) (#1679)

gate leads-topbar 'Find your leads' CTA on sources having loaded (#1673)

/deanon API tab documents LeadCognition API, not DataForGTM (#1672)

Build market dead-click + nav stuck in onboarding focus mode (#1671)

de-anon GH username → LinkedIn — in-app /deanon (2 cr/resolve) + DataForGTM surface (#1669)

Linear done-on-deploy sync + commit-scope-must-exist PR gate (#1666)

2FA enrollment path for OAuth-only users (closes #784) (#1662)

hermetic handler-orchestration coverage for 10 workflows + posthog-ai

persona-eval screenshots (onboarding + leads/paywall) (#1658)

companies list links to detail + surfaced 2FA setup failures (#1657)

run cull script with bun (self-hosted node too old for .mjs ESM) (#1660)

cull crash-looping PR previews on the health axis (#1659)

actionable discover failed state — restart CTA, honest stage states, friendly error copy (#1650)

onboarding 5xx error fallback + align signup-onboarding spec to 1-step contract (#1647)

consistent locked-list github_bio redaction + plan-aware insufficient-credits copy (#1654)

Sentry capture on discover/leads error paths + server-side unlock_completed event (#1646)

treat GitHub URLs as GitHub source input, not website domain, in 1-step onboarding (#1649)

null-safe discover session names in nav (app-crash on unclassified sessions) (#1653)

align Inngest Sentry tags with fnIds + top-level signal-ingest capture (#1644)

trigger CI/deploy — squash body inherited [skip ci] from branch chore commit

bump vitest 2.1.8 → 3.2.6 (GHSA-5xrq-8626-4rwp) (#1642)

fail-closed allow-list redaction for locked leads (#1640)

tini as PID1 in dev container to reap orphaned children (#1637)

upgrade better-auth suite to 1.6.15 (CVE-2026-41427) (#1635)

CJM/JTBD 4-journey bug sweep — 56 bugs (P0 PII/billing + P1 pipeline) [bundle] (#1633)

deliver daily digests to all orgs incl. no-prefs-row + cron heartbeat (#1567)

widen Playwright timeouts on contended self-hosted runner (#1634)

collapse onboarding to one step, move ICP in-app (fixes zero-repo dead-click) (#1582)

remove GraphQL email field that nulled every lead (funnel down ~90min) (#1580)

harvest free GitHub emails on ingest (profile + commit) (#1579)

generate signals.id in raw insert (BUN-1D NOT NULL) (#1577)

migrate GitHub signal ingest CF→Inngest (restore dark funnel) (#1576)

remove broken FullEnrich cost + count real emails_found from leads (#1575)

daily report — signals by loaded_at + estimate disclaimer (#1574)

drop deleted apps/web-legacy COPY from Dockerfile — unblocks prod deploys (#1571)

de-anon rate %, prev-day deltas, method breakdown in daily report (#1570)

held-out precision gate tool (relaxed mode verdict: ~62% → stays strict) (#1568)

auto-prune-disk df flag + retire dup script (#1566)

delete apps/web-legacy + finish pnpm→Bun CI cleanup (#1565)

jean.json worktree automation + auto-free-port dev launcher (#1560)

persist auto-refreshed linear token back to vault (#1553)

discover-yc --relax-filters eval mode (#1536)

use db.json for jsonb inserts — fix serp_cache + audit metadata double-encode (#1552)

auto-refresh linear-claude MCP token on 401 (#1551)

cache BrightData SERP results so de-anon re-runs never re-pay (#1549)

unwrap BrightData /request envelope so SERP de-anon returns matches (#1548)

dedupe leads_empty_state_viewed tracking (#1547)

visual-review screenshot gate (git pre-push + Claude Stop hook) + batch-results-panel fix (#1541)

emit new_lead from Inngest deanon completion (#1537)

document DataForSEO credentials (#1535)

gitignore output/ runtime cache (#1533)

track subscription_started on paid conversion, rename plan_switched intent event (#1501)

clarify match_confidence is the product "ICP fit" number (#1491)

deanon cost-control — quality gates before paid search (#1525)

lead relevance — ingestion gate + signal-strength + best-first deanon (#1524)

escape data.login in not-found banner on github-email-finder (#1526)

normalize short dispatch SHA to full OID in deploy gate (#1522)

rewrite titles/meta + leading-answer blocks on top CTR-anomaly pages (#1518)

BrightData SERP discovery — retire Serper from live deanon path (#1517)

emit signup_completed deterministically on first authenticated session (#1507)

penalize off-ICP audience-mismatch keyword false-positives at higher-tier caps (#1458)

wire fullstack auth E2E into web lane on push-to-main / run-e2e (#1482)

unify notification template styling with the app design system (#1453)

deploy-production hardening — superseded-tolerance, Inngest settle-gate, loud notifications (#1459)

automate Inngest re-registration in deploy-production (#1442)

how Discover finds buyers — source of truth + marketing

notification templates v2 — rich lead card, ICP/signal, reveal CTA + channel click-tracking (#1432)

notification deep-link + meter loading flash + digest delta body

extract shared mobile-first PlanCards (billing + activation modal) (#1420)

tighten Source Map ladder, rename Discoveries, fix pricing drift (#1415)

route New Source Map to create flow instead of upgrade wall (#1413)

un-gate new-lead notifications + per-event channel routing (#1406)

act-as members 500 + dedup view-as banner (#1392)

add devops e2e smoke marker (#1391)

reconcile previews from Coolify registry

deterministic active-org resolution for view-as consistency (#1385)

summarize release notes momentum

stop stale-org onboarding redirect loop + bounce completed users off /onboarding (#1383)

log dispatch label decision on dry-run create path (#1382)

harden Sentry auto-fix loop — impact gate, sensitive-path guard, circuit breaker, resolve-verify (#1380)

scope view-as-org raw fetches + reject stale active org (#1378)

align nav release badge with live version

show live release notes version

refresh release notes version

resolve auth review findings and simplify release workflow

gate legacy data-quality admin routes to platform admins (#1124)

use staging auth schema columns in signup health (#1246)

treat cancelled detect job as no-op in CI umbrella gate (#1373)

auto-resolve Sentry issues via API when Fixes BUN-N merges to main (closes #643) (#1367)

revert deploy-production concurrency group (broke CI guard) (#1370)

remove LLM customer-summary release-notes feature + fix prod deploy/rollback race (#1369)

add 15s settle wait before browser smoke check (#1364)

CI/CD — Node.js setup for browser smoke, RELEASE_LINE in Coolify, backfill release summaries (#1361)

harden Sentry auto-triage cron, threshold, and add SLA enforcement (#1354)

show Slack-delivery paywall instead of a silently-reverting select (#1350)

default notifications to email, not unconnected Slack (#1344)

add content safe area top padding in fullscreen mode

mount viewport before requestFullscreen in SDK v3

Mission Control UI — fix all remaining issues

request fullscreen mode in Telegram Mini App

Mission Control UI prod URL fix + tab icons + safe-area layout

setup-bun.sh version-gate early exit for self-hosted runners (#1336)

mission-control MVP — TDD package + superpowers skills (#1299)

deduplicate signup_completed for email signups (#1316)

CI cancelled-lane flake + CSP missing PostHog/Intercom sources (#1314)

hide retry buttons and add exit ramp on invalid invite (#1312)

use Linear API key for merge sync

Stripe customer crash + startup DNS retry + search tracking (#1307)

skip duplicate production deploy runs

refresh visible release version

add production browser smoke

enqueue realtime notification deliveries

finish Bun-first runner simplification

include CI lane package in production Docker install (#1298)

skip deploys for test-only runtime changes (#1295)

make Linear release sync idempotent (#1294)

repair Bun production deploy commands

fetch Sentry triage unit tests in workflow

polish notifications and Sentry watch

add notification delivery templates

run deploy dedupe on GitHub runner (#1286)

isolate web CI command cwd (#1281)

align lanes with live-health deploy dedupe (#1282)

dedupe production deploys from live health (#1280)

split CI into reusable lanes (#1276)

add notification delivery destinations (#1278)

clarify notification delivery settings

add Sentry triage manual ops

let manual deploy bypass stale dedupe (#1272)

harden notification live smoke coverage (#1271)

prefer Linear API key for Sentry triage

repair Sentry deanon requeue automation

E2E notification attribution flow (#1250)

preference enforcement (mute + digest batcher) (#1235)

add staging db runtime guardrails

post-rewrite hook to warn on rebase squash-revert pattern

leads + companies analytics events (MEDIUM batch) (#1228)

Wires `inngest-cli dev` as a third Playwright webServer when `PLAYWRIGHT_INNGEST=1` (LC-603)

wire Inngest dev server into Playwright harness + implement E2E test (#1226)

restore getModuleDb singleton + fix inngest.send → step.sendEvent (#1222)

fixme → skip on Inngest happy-path E2E test

module-level DB singleton + batch size limits + concurrency key + throttle

gate matview.refresh.requested emits on actual writes

add missing PostHog tracking events + internal email fix (#1214)

sync PostHog init + wire optOutIfInternal (#1202)

orphan-rate signup health metric (#1196)

Sentry bug fixes (BUN-K/S/R) + broaden auto-triage scope (LC-530) (#1186)

persist Add source clicks immediately (#1175)

disable ml-worker cron while ML classification is frozen

backend observability — Sentry on all Inngest workflows + PostHog pipeline events

Sentry capture inside email dispatch (supersedes #1164) (#1169)

PostHog email_sent + email_failed events (#1167)

[MIGRATION] add channel to notification_log unique dedup index (#1165)

cap notification retry at 10 attempts (avoid runaway hourly hits) (#1162)

remove green bg bleed-through on org logo in team switcher

stabilize live source activation UX

make LinkedIn URL read-paths dual-column-aware (#1129)

poll FullEnrich reverse-email lookups via correct endpoint (#1109)

LinkedIn enrichment unlock — paid tier with provenance-aware free signals (#1070)

Linear webhook receiver for Claude Code agent assignability (#1099)

wire @claude mentions via claude-code-action (#1093)

tag AI agent eval sessions in PostHog (#1083)

improve Discover paywall UX — PostHog tracking + empty-state copy (#1074)

add design-partner / early-adopter framing for founder persona (#1073)

add CSV export button to leads toolbar (#1068)

track search_performed, unlock_intent, and repo-add friction (LC-373/374/375) (#1061)

startup-time assertions for required env vars + DB tables (#1041)

restore Serper.dev provider — CSE migration silently disabled search for 6 days (#1039)

add PostHog OTel structured logging to discover pipeline

bump actions/checkout from 5.0.1 to 6.0.2 (#975)

add PostHog tracking for playbook copy, enrichment errors, signal/ICP hover dwell (#1034)

PostHog LLM tracking follow-ups — rescore span, cost fallback, session_id (#1031)

upload source maps to PostHog for readable exception stack traces (#1025)

set name property on PostHog org groups for readable display (#1017)

rename login_completed → signed_in for PostHog consistency

reduce PostHog telemetry noise, dedupe events, track discover failures

make email verification optional — users can log in unverified (LC-479)

hotfix(invite+tracking): fix false auth error on invite page and deduplicate pageviews (#1001)

switch rate limit storage to memory (fixes intermittent Google OAuth failure) (#999)

disable BQ daily cap enforcement (log only, never block) (#998)

reload session list from server after single-session revoke (#995)

add workflow_dispatch fallback to auto-tag-on-deploy (#994)

document Discovery v2 architecture

hotfix: React #300 loop, rateLimit schema, dual error reporting, stale-chunk reload, passkey list (#993)

filter invalid organizations in switcher

hotfix(LC-468): fix session.create.after hook signature — 500 on GitHub/Google OAuth (#985)

cap active sessions at 5 per user (#984)

persist rate limits to DB — rate_limit table + startup migration (#980)

auth hardening — 2FA, scrypt, email verification, JWE, rate limits, session filter (#978)

fix Google/GitHub login for existing users + auth error page (#968)

simplify discovery activation and source mapping

hotfix: revert PR #966 (LC-465 passkeys) — prod React #300 (#969)

Settings Auth panel — track auth_provider_linked on confirmed callback (closes #956) (#965)

Better Auth email flows — changeEmail config + settings UI (closes #955) (#959)

scope Sources query keys by active org ID (#954)

MCP discovery flow + repo backfill tools (closes #880) (#951)

add /.well-known/mcp-server and /.well-known/mcp/servers.json discovery endpoints (#950)

allow owner role cleanup

add Better Auth JWKS schema

ignore stale auth session cache cookie

run auth plugin schema startup migration

add organization invites and auth security

proxy tracking host through worker

route tracking host robots through worker

clean up GSC page indexing examples

clean up GSC indexing signals

prevent stale robots sitemap cache

ship AI-ready sitemap and pricing facts

unify lead score indicators

ICP/signal score consistency + LC-447 plan-gated BQ lookback

adaptive BQ backfill window to avoid cost limit errors

use org-specific lead_scores in lead detail (ICP match + signal strength) (#918)

broaden Discover market source planning (#911)

refresh org_lead_data mat view after repo removal (#917)

soften ICP coverage icon state

add ICP setup workflow and resilient re-score (#914)

harden lead avatar fallback for listing and peek

polish responsive leads and companies UX

auto-deploy release metadata bump

release: v0.82.0 Contact & UX hardening

keep release metadata aligned

sync visible release to v0.81.3

sync visible release version

align contact enrichment smoke state

update actions and repair preview reconcile

polish release notes and notifications

stop legacy API worker claiming app domain

auto-trigger lead enrichment after repo onboarding completes

retrigger CF Pages deploy after lockfile sync

sync pnpm-lock.yaml after Astro revert (#881)

Removed `backfillDiscoverBatchToMatView` from `discover-sessions.ts` — keeps the Discover People tab (batch-based) fully separated from the Leads/Companies surface (organic signals via mat view)

restore Technology Comparisons grid after static HTML deletion (#877)

CSP headers, mobile pricing, homepage copy, blog comparison table (#876)

remove stale static HTML overrides, fix alternative page layout (#874)

Astro landing migration — complete implementation (#729)

YC outreach export script — CSV per batch for email drip (#866)

preserve HTTPException status codes in sentryErrorHandler

install wget alongside curl for Coolify healthcheck

add /api/dev/sentry-test E2E throw endpoint

add Sentry error capture to Bun server

clean self-hosted checkout workspace before fetch

remove organizations GQL field — stops Inngest read:org storm (#850)

document GOOGLE_CSE_API_KEY + GOOGLE_CSE_ID in .env.example

add github_status migration + fix JSONB containment in leads auth (#846)

add idempotent startup migration for LC-347 companies columns

**Lead detail 404 fixed** — `GET /api/leads/:leadId` now falls back to `lead_unlocks` and `discover_batches` ownership checks when `org_lead_data` mat view is stale. Discover-batch leads that hadn'...

skip direct-IP TLS cert check on self-hosted runner (hairpin NAT) (#838)

fix YAML parse failure in deploy-production.yml Telegram block

remaining UX stack — ICP pre-fill, org enrichment, Discover (#834)

lower Inngest concurrency to 5 to match plan limit

GitHub GraphQL client + lead/company enrichment wiring (#819)

regenerate release data — v0.69.1 with LC-274 stack entries (#823)

strip @ prefix and null whitespace org_handle values (#822)

shared RepoSourceCard + discover owner-grouping toggle (#693)

onboarding handoff + empty states + cross-surface drill-down (#687)

add ALTER TABLE migration for lead_enrich_runs schema drift (#820)

allow Graphite stack + builder date-prefix branch names in CI

overnight data-quality enrichment — quick fixes + domain scrape + GitHub-org API

dedupe pnpm lockfile to fix drizzle-orm dual-resolution

add page_viewed event to /repos (Sources) page

keep leads and companies links stable

Revert "fix(playbook): separate seller and prospect context"

separate seller and prospect context

remove cross-repo BQ job cache contamination

PostHog foundation — SPA pageviews, internal opt-out, ?posthog=force, tracking gaps (#740)

CI preview cert gate, skill updates, and process improvements (#739)

Unify admin page headers (#685)

entrypoint writes .env.docker so MCP creds reach dev user shell

companies cleanup pass — junk, person/URL, fake-bio, trivial dupes (#569)

add source + employer_signal_type columns + deterministic backfill (#678)

MCP OAuth readiness for ChatGPT and Claude (#621)

add agent vault item lookup commands (#721)

keep agent vault profile env in parent shell (#710)

make dev container healthcheck process-existence based (#708)

avoid pgrep in dev container healthcheck (#707)

make dev container healthcheck process-based (#706)

add agent vault support to dev container (#704)

backfill existing Linear releases

run Linear backfill workflow from main

sync Linear release issue keys explicitly

skip Linear resync on release-bump deploys

run stack preview validation on opened PRs

define linear stack preview workflow

add graphite-app to advisory authors + bump-push retry on race (#662)

LC-296 release-notes header version + sharper feedback copy (#620)

LC-293 customer-facing release notes via LLM + honest pace stat (#618)

LC-291 use --notes-file to avoid YAML block-scalar escape (#616)

per-version release notes pulled from GitHub Release body (#605)

CI/CD redesign — Codex tier + version source-of-truth + Graphite policy

LC-516 execute_code Code Mode tool — CF Worker sandbox + pre-injected LeadCognitionClient

**LeadEnrichWorkflow** — 8-step weekly Inngest pipeline (Sundays 4am UTC) for enriching leads with AI-generated signals

**LC-196**: `lc196-generate-descriptions.ts` — ML.GENERATE_TEXT script to fill ~32K missing company descriptions via Vertex AI Llama 4 Scout (~$5/run, dry-run safe)

New `CompanyGitHubRelinkWorkflow` CF Durable Workflow scans companies missing a GitHub alias and probes GitHub's search API to link the correct org

Follow-up to v0.52.0 and PR #455.

This release ships the LC-213 and LC-218 data-script fixes after resolving the accepted Codex P1 review findings.

**CSP violations resolved** — Content-Security-Policy report-only was flagging ~1,300 weekly violations from three legitimately-used domains:

Rebased and released the useful local pipeline commits for multi-token GitHub App pooling.

Ship the LC-146/LC-147 ICP universe expansion and pgvector substrate.

Added FullEnrich people/company Search API pilot endpoints.

Upgraded AI playbook generation with Discover/customer context and LLMAPI-backed model selection.

align v0.46 release metadata

Hotfix release for the post-v0.46.0 automated review findings that blocked production deployment.

This release ships the R2 public API / MCP stack integration from PR #425.

Ships LC-93 via PR #424 as a clean replacement for stale/conflicted draft PR #291.

Ships the pending cleanup approval queue and apply pipeline from PR #382, including rollback/audit safeguards and review fixes for table dispatch and archived-state audit snapshots.

Fixes the outreach-ready company predicate to use a null-safe description/name comparison so companies with substantive descriptions and missing company names are not hidden by default.

**LC-165 M1.2 — Rule-based cleanup proposer** (#366): `lc165-propose-from-findings.ts` reads unresolved rows in `data_quality_findings` (39 findings across 8 check IDs) and stages proposals in `pen...

`icp-merge.ts` handles archived-domain collisions (Codex P1 fix applied)

Hotfix release.\n\n- Adjusted Intercom launcher placement so normal app pages use the standard lower-right position.\n- Raised the launcher only on the Leads list page so it stays clear of the pagi...

Small UX polish follow-up for onboarding repo inputs.

Ships the improved onboarding source setup and Lead Supply Check flow.

Fixed Discover Market reveal for over-limit free-plan workspaces so they receive a plan-limit upgrade message instead of a generic scan failure.

Completes LC-111 notification delivery cutover for email, Slack, webhook, and in-app delivery rows.

Closes Linear LC-94. Single-issue release; no other items bundled.

This minor release ships the new filtering system for Leads and Companies.

**FullEnrich waterfall contact enrichment end-to-end**: work email, personal email, mobile phone unlocked via FullEnrich's waterfall provider chain. Live at `/leads/:id`.

Hotfix — fixes misleading Slack deploy notifications from v0.38.5 and consolidates the deploy flow so every surface reports into one atomic message.

Hotfix release — closes the manual-deploy and visibility gap across the three deploy substrates.

Security patch — closes two open Codex **P1** findings from PR #233 plus one same-shape vulnerability Codex didn't see.

Patch release — CSP violations now stream to PostHog (Sheldon's Telegram digest picks them up), plus several landing / tools hotfixes merged alongside.

Patch release — Stripe webhook validation helper + CSP report-only observation.

Patch release closing out the v0.38.0 edge-case sweep follow-ups.

LC-80: Polish Discover workspace, Sources, and billing checkout by @wiselancer in https://github.com/wiselancer/leadcognition_v2/pull/220

monitored orgs card, Discoveries quota, Current Plan layout by @wiselancer in https://github.com/wiselancer/leadcognition_v2/pull/218

session index at /discover + human-readable status copy (closes #211 #207) by @wiselancer in https://github.com/wiselancer/leadcognition_v2/pull/214

When a user entered a popular repo (e.g. `stripe/stripe-js`), the backend would escalate BigQuery lookback to 6 months, exceed the 20 GB per-query cap, throw `BigQueryCostError`, and silently reset...

allow retry from batch_failed + surface start-batch errors in UI by @wiselancer in https://github.com/wiselancer/leadcognition_v2/pull/185

Ship public release notes and early adopter program by @wiselancer in https://github.com/wiselancer/leadcognition_v2/pull/157

**Full Changelog**: https://github.com/wiselancer/leadcognition_v2/compare/v0.31.2...v0.31.3

Empty content release — this tag contains a single `chore(release): bump version to v0.31.1` commit and exists because the release-tagging step fired twice. No code changes relative to v0.31.1; kep...

repo-intent-score legend centering + Network error UX by @wiselancer in https://github.com/wiselancer/leadcognition_v2/pull/151

Top-of-funnel SEO surfaces that work without login. Each uses shallow gating — show one primary number, CTA to signup.

**Free tier**: 25 one-time identity unlocks (LinkedIn URL + name + company) — was 5

**Pro $99/mo** ($79 annual) · 500 credits

Outcome-driven hero on `/discover` and `/onboarding` entry pages — replaces bland "enter your domain" with a sharper pitch about finding developers living in your ecosystem

enrich PostHog + Intercom identity with website, role, stage, unlocks

Full Intercom Messenger integration with PostHog as single source of truth for events.

**Skip inactive-repo leads in retry cron**: `retryDeanonJob` now filters to only leads with at least one signal from an *active* repo. 16,792 leads from deactivated/test repos (`openclaw/openclaw`,...

**Fix identify timing in onboarding**: `posthog.identify()` now fires synchronously before `onboarding_started`, ensuring the event is attributed to the identified person rather than an anonymous p...

New `retry-deanon` cron job runs every 10 minutes to drain the backlog of unenriched leads (~19k leads/day capacity)

Fixed `e2e` job: missing frontend dist stub caused all wrangler builds to fail (10/10 CI runs were red)

**Production migrated to Bun/Coolify** — `app.leadcognition.io` now serves from Hetzner (was CF Workers). Same runtime as staging, auto-deploys on push to `main`

Set up `t.leadcognition.io` as a managed reverse proxy via Cloudflare

improve UX — overflow states, accessibility, clarity

resolve TDZ error in leads refetchInterval

**Leads filter optimization:**

Identified via Chrome DevTools MCP performance audit (LCP 620ms, render delay 460ms driven by monolithic JS bundle).

6e378a9 release: deanon workflow hardening, company affiliations, and admin/web improvements

**API: Org-scoped lead detail verification**

Security fix: FullEnrich webhook endpoint now requires a shared secret token (`?secret=<token>`) since FullEnrich doesn't provide HMAC signatures. Secret set on both API and pipeline workers.

address review findings — NaN validation, dead code removal

Full audit and cleanup across pipeline, API, and packages.

**Shadow design tokens** — consistent elevation system via CSS variables

All 5 AI touchpoints now emit `$ai_generation` events to PostHog for full LLM observability:

FullEnrich results now delivered via webhook in ~30-90 seconds (was ~5 min polling)

PostHog analytics is now live across all 10 landing pages and the web app, providing comprehensive tracking of user behavior, conversion funnels, and engagement metrics.

Full go-to-market strategy at `docs/gtm-strategy.md`

Onboarding and dashboard UX improvements following v0.10.1:

**Match confidence badge**: was showing `80/5` (wrong scale) — now correctly shows `80% match`. matchScore is stored as 0-100 in DB; UI thresholds updated accordingly.

**Hardened validate_match**: score guard now strictly validates integer 1–5; hallucinated template strings fall back to 0 instead of crashing

**Unipile company detection**: auto-detects company from signup email via LinkedIn search

**Root cause**: `ENRICHMENT_QUEUE.send()` was called in a loop — one call per new GitHub actor. Each `.send()` consumes a Cloudflare subrequest. Popular repos (e.g. `meta-llama/llama-recipes`) can ...

**fix(pipeline): crash "events is not iterable" on dirty repo slugs**

Replaced sequential repo polling with `Promise.allSettled` batches of 20

**`migrate-haystax.ts`** — 4-phase migration from Haystax Heroku Postgres to LCv2 production

Full observability for LeadCognition: every external API call is now audited, costs and errors surface in the admin dashboard, and PostHog tracks product analytics.

**BigQuery job ID caching**: Cache the BQ job ID after submission so processing failures can retry for free (polling the existing job instead of re-submitting and re-scanning)

**Critical auth fix**: Missing Stripe secrets were crashing `createAuth()` on every request (`new Stripe(undefined)` throws). All 10 wrangler secrets now verified and set in production.

Plan-aware limits: Free (2 repos), Starter (5), Growth (15), Scale (50)

Full subscription management via Better Auth Stripe plugin

Lead detail page with locked/unlocked paywall. Tags and favorites system. Leads page v2 redesign (signal strength column, activity, enhanced filters). Companies page v2 (card grid, inline expansion...

BigQuery dry-run guards and byte limit enforcement (safeQuery()). UI normalization to shadcn default theme. BigQuery cost analysis and spending guardrails added to CLAUDE.md.

Companies page with table view and slide-over detail panel. Company-to-lead grouping from GitHub org signals.

Activity tab with day-grouped timeline and per-event-type icons. Signal strength scoring.

Leads list with filters (signal strength, event type, date). BigQuery intelligence endpoints: related-repos, tech-stacks, company-map. Materialized view org_lead_data as single read source.

Social accounts extraction from GitHub profiles (LinkedIn, Twitter, etc.). Commit email capture from PushEvents. Self-declared LinkedIn trusted without AI validation.

Org-scoped data model, lead_unlocks paywall (personal info redacted until unlocked), credits system with 4 plan tiers (Free/Starter/Growth/Scale).

GitHub Events polling cron (15-min), LinkedIn discovery via Unipile, AI validation via Claude Haiku, email enrichment via FullEnrich. Leads table unifying GitHub + LinkedIn profiles.

BigQuery integration for GitHub Archive data — backfill signals from starred/watched repos. Drizzle schema for signals, leads, and orgs.

Initial project scaffold: Hono API on Cloudflare Workers, React + TanStack Router frontend, Drizzle ORM + Postgres, Better Auth with Google OAuth, GitHub Events polling pipeline.