# Changelog of BODACC · French Legal Announcements by Date or SIREN (`corent1robert/bodacc-announcements-scraper`) Actor

- **URL**: https://apify.com/corent1robert/bodacc-announcements-scraper/changelog.md
- **Full Actor documentation**: https://apify.com/corent1robert/bodacc-announcements-scraper.md

## Changelog

All notable changes to this Actor are documented in this file.

The format follows [Keep a Changelog](https://keepachangelog.com/en/1.0.0/). Actor versions use **MAJOR.MINOR** (Apify convention).

### \[1.11] - 2026-09-07

#### Added

- **Actor Standby:** HTTP lookup `GET`/`POST /lookup` — **SIREN mode only**, up to **5** SIRENs (announcements capped at 5 rows). Date-range and all-France catalog stay on **Start** and are rejected over HTTP.
- OpenAPI spec for the Standby tab (`web_server_openapi.json`)

### \[2026-09-01]

#### Changed

- Free plan cap: 20 rows per run. Anonymous Try off.

### \[1.10] - 2026-08-22

#### Changed

- Published-tasks kit aligned with Store examples (TEMPLATE, SEO strategy, 4/4 published).
- Store snippet names companion **French Companies**; README table labelled as the company-register suite.

### \[1.9] - 2026-08-20

#### Changed

- Store README: Also available at top, FAQ, Try cap / weekend retry, local `apify run`.

### \[1.8] - 2026-08-20

#### Changed

- Store Try prefill: 10 announcements (was 25).
- Empty yesterday window retries the last 7 days so weekend health checks still produce a dataset.

### \[1.7] - 2026-08-20

#### Changed

- Store categories: Lead generation / Business only.

### \[1.6] - 2026-08-19

#### Added

- Store **Examples** (published tasks): daily creations, insolvency watch, fonds de commerce, restaurant creations.

#### Changed

- Store categories: Lead generation / Business. SEO title and description tightened for French gazette buyers.

### \[1.5] - 2026-06-12

#### Added

- **Run presets** — Daily new businesses, Insolvency watch, M\&A sales, Full SIREN history (`runPreset` in Console).
- **Region presets** — One-click geographic filters (Île-de-France, PACA, etc.) with manual department override.
- **Company register enrichment** — Optional lookup via annuaire-entreprises (`enrichWithCompanyProfile`).
- **Dedupe by SIREN** — Keep the latest announcement per company for lead lists.
- **Export profiles** — `full`, `lead_gen`, `insolvency` column sets.
- **RUN\_LOG** — Plain-text run timeline in key-value store; linked from output schema.
- **Dataset views** — Lead gen and Insolvency tabs; schema keys aligned (`announcement_id`, `announcement_type`).

#### Fixed

- **Multi-type filters** — OR logic for announcement types (ODS `refine` repeated = AND).
- **Multi-department filters** — OR logic for region presets.
- **SIREN search** — `registre="…"` exact match.
- **Field coverage** — All 8 types labeled; modification, radiation, dpc fields extracted.
- **Console interface** — benefit-first copy, prefills, output view links, KV schema `collections` format, RUN\_LOG URL `/records/RUN_LOG`.

#### Changed

- Console form reorganized; default Try input = daily preset, 25 rows, dedupe ON.
- Tiered PPE pricing; actor-start + enrichment events.

### \[1.4] - 2026-06-12

#### Removed

- **Google Maps enrichment** — `enrichWithMaps`, Puppeteer, OpenAI matching, `maps-enrichment` billing event, and all `maps_*` fields. Actor is HTTP-only (~512 MB default memory).

#### Fixed

- **Global `maxResults` in SIREN mode** — cap now applies across the full SIREN list, not per company (e.g. 10 SIRENs × 500 no longer yields up to 5,000 billed rows).
- **Parallel fetch race** — monthly date-window workers share one mutex-backed counter so `maxResults` cannot be exceeded under concurrent pagination.

#### Added

- **CI** — unit tests + live API smoke (`npm run test:ci`) via GitHub Actions.
- **Tiered pay-per-event pricing** in `actor.json` — Bronze→Gold discounts on announcement rows.
- **README** — Who is this for?, pricing scenarios, legal disclaimer sections (Store SEO).

#### Changed

- `queryBuilder` unit tests aligned with ODS v1 `buildParams` / `buildApiUrl` API.
- Store listing copy (`title`, `description`, `seoDescription`) refreshed; Actor version **1.4**, `isPublic: true`.

### \[1.3] - 2026-06-05

#### Changed

- **README — Companion actors** — cross-links to French Accountants and Notaires.fr directories.

### \[1.2] - 2026-05-11

#### Added

- **Google Maps enrichment** (`enrichWithMaps`) — optional two-phase pipeline: Phase 1 fetches all BODACC records via HTTP; Phase 2 launches Puppeteer workers to search Google Maps for each announcement and merges the business profile when a confident match is found. Added fields: `maps_found`, `maps_name`, `maps_phone`, `maps_website`, `maps_category`, `maps_address`, `maps_rating`, `maps_reviews`, `maps_hours`, `maps_url`, `maps_score`, `maps_search_query`.
- **`mapsWorkers`** input — number of parallel browser tabs (1–10, default 3). Each tab runs an independent Maps search; cap to avoid OOM.
- **`mapsMinScore`** input — minimum match confidence score (0–10, default 7) to accept a Maps result. Scores below threshold are discarded.
- **`openaiApiKey`** input — when provided, GPT-4o-mini validates ambiguous matches and returns a numeric confidence score (0–10). Without it, a fast regex-based name matcher is used (score 8 on match, 0 on miss).
- **`trade_name`** extracted from BODACC `listepersonnes` (`nomCommercial` / `nomUsage`). Used as the primary search term for Maps enrichment.
- **Pay-per-event billing** on Apify Cloud — `Actor.chargeEvent('maps-enrichment')` is called for each accepted Maps match.
- **Progress logging** for Phase 2: every 25 records logs `searched/total`, match count, throughput (`rec/s`), and ETA.
- **`src/lib/mapsEnricher.js`** — self-contained Puppeteer module with `PagePool` (worker pool), `namesMatch` (token-based fuzzy matching), `buildSearchName` (trade name or first+last name), and `validateWithLLM` (GPT-4o-mini scoring with regex fallback). All logic is isolated for unit-testability.

#### Changed

- **`defaultRunOptions.memoryMbytes`** raised from 256 to 1024 — required when `enrichWithMaps: true` (each Puppeteer tab uses ~200MB).
- **`defaultRunOptions.timeout`** raised from 3600s to 7200s to accommodate large Maps enrichment runs.
- **`maxMemoryMbytes`** raised from 1024 to 4096.
- `actor.json` `version` updated to `1.2`.
- `actor.json` `generatedBy` updated to `Cursor with Claude`.
- Run logs: removed emoji from Maps match lines (`✓` → plain `Match:` label) to conform to premium/Store log style.

### \[1.1] - 2026-05-10

#### Added

- **Search URL mode** — paste any BODACC URL directly (`searchUrl` input field); the actor extracts keyword, types, date range, and departments automatically. Manual fields override URL values when both are set.
- **Keyword search** (`keyword`) — full-text filter passed directly to the API (e.g. `peinture`, `plomberie`). Matches company names, activity descriptions, and free-text fields.
- **Multi-keyword search** (`keywords` list) — one keyword per line, combined with OR logic (e.g. `peinture OR peintre OR paysagiste`). Takes priority over the single keyword field.
- **`creation` type** added to announcement types — the API's primary value for new registrations (distinct from `immatriculation`). Default types updated from `["immatriculation", "collective"]` to `["creation", "collective"]`.
- **`modification`, `radiation`, `dpc`, `divers`** added to available announcement types with correct labels.
- **`output.csv`** — results are now written progressively to `output.csv` at the project root during local runs (same behaviour as the other actors in this repo).
- **`csvWriter.js`** — dedicated CSV writer with BODACC-specific columns: `announcement_id`, `date`, `announcement_type`, `company`, `siren`, `person_type`, `city`, `postal_code`, `department`, `region`, `activity`, `legal_form`, `capital_eur`, `management`, `registration_date`, `activity_start_date`, `judgment_type`, `judgment_date`, `url`, etc.

#### Fixed

- **Switched from ODS v2.1 to v1 API** (`/api/records/1.0/search/`) — the v2.1 API's `q` parameter was treated as a relevance score (not a strict filter), returning inflated `total_count` figures. The v1 API matches the website behaviour exactly (e.g. `peinture + creation + 2026` → 1 877 results, not 203 506).
- **`creation` family** now correctly triggers registration field extraction (`legal_form`, `capital_eur`, `management`, `address`, `activity`, etc.) — previously only `immatriculation` and `vente` did.

### \[1.0] - 2026-04-28

#### Added

- **Initial release** — scrape BODACC legal announcements from the official ODS v2.1 API (`bodacc.fr/api/explore/v2.1`).
- **Two modes**: `dateRange` (filter by date, types, departments) and `sirens` (full announcement history per SIREN).
- **Five announcement types**: registrations (`immatriculation`), insolvency (`collective`), business sales (`vente`), accounts filings (`depot`), other notices (`avis`).
- **Smart defaults**: `dateFrom` / `dateTo` default to yesterday → today when omitted — ready for daily scheduled runs out of the box.
- **Rich output**: SIREN extracted from `registre` array, legal form, capital, directors, activity, address, judgment details (type, date, liquidator info), direct bodacc.fr URL.
- **Resilient HTTP layer**: exponential backoff on 429/5xx (up to 10 retries, 600ms inter-page delay + jitter).
- **Per-SIREN try/catch** in SIREN mode: a failing company does not abort the rest of the run.
- **`transformAnnouncement`** pure function: parses nested JSON strings from the API (`listepersonnes`, `acte`, `jugement`, `listeetablissements`) into flat, clean records with empty fields omitted.
- **`buildWhereClause`** pure helper: constructs ODS SQL `where` clauses from filters — unit-testable without network.
- **Unit tests** covering: `buildWhereClause` (type filter, date range, department, SIREN, combined), `extractSiren` (various formats), `transformAnnouncement` (registration, insolvency, minimal/empty input), HTTP retry logic (`computeBackoffMs`, `parseRetryAfterMs`, `fetchWithRetry`).
- **`dataset_schema.json`** with `overview` view showing key columns.
- **README** with examples for daily monitoring, regional search, and SIREN history.