docs(changelog): catch up with per-version entries for 1.4.0 -> 1.5.1

CHANGELOG.md

@@ -4,6 +4,79 @@ All notable changes to PyScrappy are documented here. The format is based on
 [Keep a Changelog](https://keepachangelog.com/), and this project adheres to
 [Semantic Versioning](https://semver.org/).
 
+## [1.5.1] - 2026-08-10
+
+### Added
+- **Adaptive (self-healing) selectors.** `Selector.css` gains `auto_save=True` to fingerprint the matched element and `adaptive=True` to relocate it by similarity when the selector later matches nothing, so scrapers survive site redesigns. Relocation uses weighted signals (a stable `id`/`data-*` hook outweighs weaker ones), anchor-relative fingerprints (nearest stable ancestor + depth), volatility-aware text scoring (prices/dates/counts down-weighted), and tag-pruning before scoring. Returns a confidence score and runner-up gap (`SelectorList.adaptive_confidence`); fingerprints persist in a JSON store namespaced by site.
+- README logo header and an updated title/tagline.
+
+## [1.5.0] - 2026-08-10
+
+### Added
+- **Chainable `Selector` parser.** Navigate HTML directly (Scrapy/BeautifulSoup-style): `css()`/`xpath()` with `::text` / `::attr(name)` pseudo-elements, `find_all()`, `find_by_text()`, and `find_similar()`. Returns a `SelectorList` with `.get()`/`.getall()`/`.text()`. Usable standalone on an HTML string.
+- **`pyscrappy extract` CLI.** Scrape a URL straight to a file with the format inferred from the extension (`.md`/`.json`/`.txt`/`.html`), plus `--css-selector` and `--render-js`.
+- **TLS-fingerprint impersonation.** `ScraperConfig(impersonate="chrome")` routes the sync HTTP client through a `curl_cffi` session that mimics a real browser's fingerprint, getting past anti-bot filters that block plain clients. Optional `pyscrappy[stealth]` extra; the async path raises a clear error when set.
+
+### Fixed
+- `extract` moved to an MCP-free `pyscrappy.cli` module so the CLI works on Python 3.9 (the MCP stack requires 3.10+).
+
+### CI
+- The weekly live-scraper integration job is now non-blocking and writes a per-scraper pass/fail summary, so a site blocking a CI IP no longer shows as a red failure.
+
+## [1.4.7] - 2026-08-09
+
+### Fixed
+- `to_markdown()` escapes pipes and newlines in table cells, so a cell containing `|` or a line break no longer corrupts the rendered table.
+- Cache key no longer produces a double `?` when the URL already has a query string, closing a second collision case.
+
+## [1.4.6] - 2026-08-07
+
+### Fixed
+- `TableExtractor` keeps ragged rows (aligning cells by position, padding short rows and preserving extra cells) instead of silently dropping any row whose cell count differs from the header.
+- Hacker News docstrings now say `num_comments`, matching the key the scraper actually emits.
+
+## [1.4.5] - 2026-08-06
+
+### Added
+- **Async concurrency helpers** `scrape_many_async` / `scrape_all_async` for running scrapes concurrently with a configurable concurrency cap.
+- MCP Toplist rank badge in the README.
+
+### Fixed
+- `image_search` returns one consistent result schema across engines (Bing and Google produce the same keys).
+- `convert_currency` and `get_weather` docstrings corrected (`wind` → `wind_speed`, right default base/target).
+
+## [1.4.4] - 2026-08-05
+
+### Fixed
+- `NewsScraper` returns a `ScrapeError` on a network/transport failure instead of raising.
+- `image_search` validates the `engine` argument and raises `ValueError` for an unsupported value instead of silently falling back to Bing.
+- MCP `scrape_stock` forwards the `interval` argument (history was locked to daily) and `search_hackernews` forwards the `tags` filter; `scrape_stock`'s mode-default docstring corrected.
+
+## [1.4.3] - 2026-08-04
+
+### Fixed
+- Pagination extracts the page number from `offset=` / `start=` / `/p/`-style URLs instead of silently stopping (one regex now drives both detection and extraction).
+- The HTTP client is closed before it's rebuilt on retry (no connection leak), and retries re-pick a different proxy from a rotating list.
+
+## [1.4.2] - 2026-08-03
+
+### Fixed
+- `GitHubScraper` default `max_results` is now 20, matching the `search_github` MCP tool, so both surface the same number of repositories.
+- Multi-valued image attributes (e.g. `srcset`) are coerced to strings, honoring `ImageExtractor`'s `dict[str, str]` contract.
+
+## [1.4.1] - 2026-08-02
+
+### Added
+- **Opt-in robots.txt politeness** (`obey_robots`) honoring `Disallow` and `Crawl-delay`, with a per-client, per-User-Agent parser cache.
+- **`ScrapeResult.to_csv()` and `.save(path)`** alongside `to_json` / `to_markdown` / `to_dataframe`, with a stdlib fallback that matches the pandas output (incl. empty-data and line-ending handling).
+
+## [1.4.0] - 2026-08-01
+
+### Added
+- **Native async scraping across every scraper** (`scrape_async`), built on a real `AsyncHttpClient` rather than a thread pool.
+- **Configurable retry/backoff** (`retry_delay`, `backoff_factor`, `backoff_max`), a **response cache TTL** (`cache_ttl`, also settable for the MCP server via `PYSCRAPPY_MCP_CACHE_TTL`), **custom headers / User-Agent override**, and **proxy rotation**.
+- **`--json` output** for the `pyscrappy chat` CLI agent, and scheduled (weekly) live integration tests.
+
 ## [1.3.1] - 2026-07-28
 
 ### Added