# Changelog of B2B Company Buying Signals — Hiring & Tech Radar (`nexascout/b2b-company-buying-signal-radar`) Actor

- **URL**: https://apify.com/nexascout/b2b-company-buying-signal-radar/changelog.md
- **Full Actor documentation**: https://apify.com/nexascout/b2b-company-buying-signal-radar.md

## Changelog

### 0.1.7

- Tightened Outreach technology detection so ordinary English uses of “outreach” do not create GTM-stack false positives.
- Added self-company technology filtering so a company name/product family is not counted as its own external tech-stack signal (for example Anthropic → Anthropic).
- Added regression tests for ambiguous outreach language, explicit Outreach sales-engagement usage, and self-company filtering.
- No collection, history, scoring-weight, pricing, or source-adapter behavior changed.

### 0.1.6

- Store-ready documentation and input copy.
- Production-tested default max jobs/company set to 100.
- Added explicit pricing, data-quality, public-source, agent-workflow and limitation documentation.
- Added automated Store metadata + Task setup scripts.
- No core collection or scoring behavior changed from V0.1.5.

### 0.1.5

- Set production default memory to 1024 MB and allowed max memory to 2048 MB.
- Canonicalized ATS source targets so equivalent Workday / Greenhouse / Lever URLs are fetched once per company.
- Generic career pages returning zero reliable job records are now treated as unresolved instead of a trustworthy zero-open-roles signal.
- Documented 1 GB production baseline and broad-scan timeout guidance.
- Preserved all buying-intelligence features; no feature cuts were made to target a lower RAM tier.

### 0.1.4

- Fixed the remaining 512 MB OOM cause on large Greenhouse boards.
- Greenhouse now fetches the lightweight board list with `content=false`, caps jobs, then enriches only a bounded recent-description sample.
- Lever now uses native `skip`/`limit` pagination instead of requesting an unbounded postings payload.
- Added `descriptionScanLimit` (default 15) to control bounded tech-signal detail enrichment per company.
- Source diagnostics now expose source totals and description-enrichment counts when available.
- Keeps V0.1.3 per-company process-and-release memory model.

### 0.1.3

- Refactored broad scans to process and release full job payloads one company at a time.
- Prevents memory growth from retaining every company's full job descriptions until run completion.
- Keeps only compact company intelligence rows and history snapshots in memory after each company is scored.
- Preserves Greenhouse / Lever / Workday / generic source behavior and recurring-history semantics.
- Intended to unlock reliable 10–50+ company scans at lower memory.

### 0.1.2

- Added high-confidence domain-to-ATS probes before generic career-page parsing.
- Domain input now tries the root company slug against public Greenhouse and Lever feeds and only accepts a match when the ATS returns real jobs.
- Fixes Cloudflare/Stripe-style cases where the company-owned careers page hides the underlying ATS.
- Keeps the strict generic job-link gate from 0.1.1 as the fallback.
- Engine version 0.1.2.

### 0.1.1

- Hardened generic career-page parsing to reject navigation/product links that look job-like.
- Added strong job-detail URL + role-title gates.
- Added one bounded follow-up level for company-owned `/careers/search`, `/careers/jobs`, `/jobs`, and similar indexes.
- Domain auto-discovery now prefers reliable ATS/job evidence and fails safe instead of fabricating vacancy counts.

### 0.1.0

- Initial private production candidate.
- HTTP-first Greenhouse, Lever and Workday adapters.
- Company-domain career-source discovery.
- Generic public career-page fallback.
- Company-level hiring, department, location, leadership and technology signals.
- Private recurring history with new/removed roles and new-tech/new-location detection.
- Explainable 0–100 buying-signal score and HOT/WARM/WATCH/LOW tiers.
- NEW/RISING/PERSISTENT/STABLE/FADING/RESURFACED states.
- Deterministic target buyer-role and relevant solution-category mapping.
- No LinkedIn dependency in V0.1.0.