mdfetch 0.5.2__tar.gz → 0.6.0__tar.gz

mdfetch-0.6.0/.specify/feature.json

@@ -0,0 +1,3 @@
+{
+  "feature_directory": "specs/010-boomi-blog-provider"
+}

{mdfetch-0.5.2 → mdfetch-0.6.0}/CLAUDE.md

@@ -17,11 +17,12 @@ src/mdfetch/
     ├── devto.py      # DevToExtractor (dev.to)
     ├── substack.py   # SubstackExtractor (substack.com + *.substack.com)
     ├── thenewstack.py  # TheNewStackExtractor (thenewstack.io)
-    └── dzone.py      # DZoneExtractor (dzone.com)
+    ├── dzone.py      # DZoneExtractor (dzone.com)
+    └── boomi.py      # BoomiExtractor (boomi.com/blog)
 
 tests/
 ├── unit/             # pytest unit tests (no network)
-└── integration/      # real network tests (Medium + dev.to + Substack + TheNewStack URLs + snapshots)
+└── integration/      # real network tests (Medium + dev.to + Substack + TheNewStack + DZone + Boomi URLs + snapshots)
 
 .github/workflows/
 ├── ci.yml            # lint + unit tests on push/PR (Python 3.12–3.14)
@@ -76,4 +77,9 @@ make typecheck    # type check
 **Issue:** `brew audit --strict --new Formula/md-fetch.rb` fails with missing system library declarations when `lxml` is a resource block.
 **Root Cause:** `lxml` requires `libxml2` and `libxslt`, which are macOS system libraries. Homebrew requires these to be declared explicitly via `uses_from_macos`.
 **Prevention Rule:** Any Homebrew formula that includes `lxml` as a resource MUST declare `uses_from_macos "libxml2"` and `uses_from_macos "libxslt"` after the `depends_on` lines. Discovered via `brew audit --strict --new` during implementation.
+
+### ⚠️ boomi.com article detection relies on `div.post-content`, NOT `section.wysiwyg-section`
+**Issue:** The Boomi blog index (`/blog/`) renders a `section.wysiwyg-section` intro but NO `div.post-content`. Selecting `wysiwyg-section` as the body container would fail to raise `UnsupportedContentTypeError` for the index and other non-article pages.
+**Root Cause:** Confirmed via live DOM inspection: only article pages render `div.post-content` (containing `section.wysiwyg-section` + a `div.blog-nav` prev/next block); the index renders `wysiwyg-section` only.
+**Prevention Rule:** `BoomiExtractor.clean_html()` MUST select `div.post-content` as the body container (its presence is the article discriminator) and strip the inner `div.blog-nav`. Do not switch to `wysiwyg-section`. The title `<h1>` lives in the page hero outside the body and is prepended.
 <!-- SPECKIT END -->

{mdfetch-0.5.2 → mdfetch-0.6.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mdfetch
-Version: 0.5.2
+Version: 0.6.0
 Summary: Extract article content from web platforms and return it as clean Markdown.
 Project-URL: Homepage, https://github.com/stn1slv/md-fetch
 Project-URL: Source, https://github.com/stn1slv/md-fetch
@@ -74,6 +74,7 @@ markdown = extract("https://dev.to/username/article-slug")
 markdown = extract("https://example.substack.com/p/article-slug")
 markdown = extract("https://thenewstack.io/article-slug")
 markdown = extract("https://dzone.com/articles/article-slug")
+markdown = extract("https://boomi.com/blog/article-slug")
 print(markdown)
 ```
 
@@ -117,6 +118,7 @@ except EmptyContentError as e:
 | Substack | `substack.com`, `*.substack.com` |
 | The New Stack | `thenewstack.io` |
 | DZone | `dzone.com` |
+| Boomi | `boomi.com` |
 
 ## Development
 

{mdfetch-0.5.2 → mdfetch-0.6.0}/README.md

@@ -41,6 +41,7 @@ markdown = extract("https://dev.to/username/article-slug")
 markdown = extract("https://example.substack.com/p/article-slug")
 markdown = extract("https://thenewstack.io/article-slug")
 markdown = extract("https://dzone.com/articles/article-slug")
+markdown = extract("https://boomi.com/blog/article-slug")
 print(markdown)
 ```
 
@@ -84,6 +85,7 @@ except EmptyContentError as e:
 | Substack | `substack.com`, `*.substack.com` |
 | The New Stack | `thenewstack.io` |
 | DZone | `dzone.com` |
+| Boomi | `boomi.com` |
 
 ## Development
 

{mdfetch-0.5.2 → mdfetch-0.6.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "mdfetch"
-version = "0.5.2"
+version = "0.6.0"
 description = "Extract article content from web platforms and return it as clean Markdown."
 readme = "README.md"
 license = { text = "MIT" }

mdfetch-0.6.0/specs/010-boomi-blog-provider/checklists/requirements.md

@@ -0,0 +1,37 @@
+# Specification Quality Checklist: Boomi Blog Provider
+
+**Purpose**: Validate specification completeness and quality before proceeding to planning
+**Created**: 2026-06-02
+**Feature**: [spec.md](../spec.md)
+
+## Content Quality
+
+- [x] No implementation details (languages, frameworks, APIs)
+- [x] Focused on user value and business needs
+- [x] Written for non-technical stakeholders
+- [x] All mandatory sections completed
+
+## Requirement Completeness
+
+- [x] No [NEEDS CLARIFICATION] markers remain
+- [x] Requirements are testable and unambiguous
+- [x] Success criteria are measurable
+- [x] Success criteria are technology-agnostic (no implementation details)
+- [x] All acceptance scenarios are defined
+- [x] Edge cases are identified
+- [x] Scope is clearly bounded
+- [x] Dependencies and assumptions identified
+
+## Feature Readiness
+
+- [x] All functional requirements have clear acceptance criteria
+- [x] User scenarios cover primary flows
+- [x] Feature meets measurable outcomes defined in Success Criteria
+- [x] No implementation details leak into specification
+
+## Notes
+
+- Items marked incomplete require spec updates before `/speckit-clarify` or `/speckit-plan`
+- Note: FR-001 and the Key Entities section reference the existing provider/`@register` mechanism. This is an
+  intentional, established convention in this repository's spec template (see prior provider specs 002–007),
+  describing the routing contract rather than implementation internals. Retained for consistency.

mdfetch-0.6.0/specs/010-boomi-blog-provider/contracts/extractor-contract.md

@@ -0,0 +1,57 @@
+# Provider Contract: BoomiExtractor
+
+This is a library; its external contracts are (1) the public `extract()` API and
+(2) the `BaseExtractor` subclass contract the new provider must satisfy.
+
+## 1. Public API (unchanged)
+
+```python
+from mdfetch import extract
+
+markdown: str = extract("https://boomi.com/blog/<slug>/", retries=3, retry_delay=2.0)
+```
+
+- **Input**: a `boomi.com` blog article URL (`str`).
+- **Output**: clean Markdown (`str`), title-first.
+- **Raises**: `UnsupportedPlatformError` (domain not registered — N/A once Boomi is registered),
+  `UnsupportedContentTypeError`, `EmptyContentError`, `HTTPStatusError`, `FetchError`
+  (all from `mdfetch.exceptions`). No new exception types.
+
+## 2. Subclass contract
+
+```python
+@register
+class BoomiExtractor(BaseExtractor):
+    DOMAINS: frozenset[str] = frozenset({"boomi.com"})
+    # MATCH_SUBDOMAINS stays False (default)
+
+    def clean_html(self, soup: BeautifulSoup) -> Tag: ...
+```
+
+### `clean_html(soup)` obligations
+
+| # | Obligation |
+|---|------------|
+| C1 | Return the `div.post-content` `Tag` with chrome stripped and the title prepended. |
+| C2 | Raise `UnsupportedContentTypeError` when `div.post-content` is absent. |
+| C3 | Decompose every `div.blog-nav` descendant before returning. |
+| C4 | Prepend a copy of the page `<h1>` as the first child of the returned body. |
+| C5 | MUST NOT modify the base class or shared utilities. |
+| C6 | MUST NOT add image-specific logic (body images flow through default conversion). |
+
+`extract()` and `convert_to_markdown()` are inherited unchanged; the latter enforces the
+ATX-heading, blank-line-collapsing, and `EmptyContentError` behavior.
+
+## 3. Acceptance contract (maps to spec SC-00x)
+
+| Check | Maps to |
+|-------|---------|
+| Each reference article → Markdown begins with `# <title>` and contains body section headings + paragraph text; no chrome | SC-001, FR-002/003/004/005 |
+| `extract("https://boomi.com/blog/")` (index) raises `UnsupportedContentTypeError` | SC-003, FR-006, FR-009 |
+| Output has no 3+ consecutive blank lines | SC-004, FR-008 |
+| ≥1 integration test hits a real reference URL with snapshot containment | SC-005 |
+
+## 4. Routing contract
+
+- After `@register`, `route("https://boomi.com/blog/...")` resolves to `BoomiExtractor`.
+- `test_router.py` unsupported-domain fixture (`wordpress.com`) remains valid — no change needed.

mdfetch-0.6.0/specs/010-boomi-blog-provider/data-model.md

@@ -0,0 +1,45 @@
+# Phase 1 Data Model: Boomi Blog Provider
+
+This library is stateless; there are no persisted entities. The "data model" describes
+the DOM structures the extractor reads and the in-memory shapes it produces.
+
+## Entity: Boomi Blog Post (input DOM)
+
+A single article page at `https://boomi.com/blog/<slug>/`.
+
+| Field | Source selector | Notes |
+|-------|-----------------|-------|
+| title | `h1` (within `section.post-detail-hero`) | Exactly one per page; outside the body container |
+| body | `div.post-content` | Body container; presence = "is an article" |
+| content | `section.wysiwyg-section.bullet-styled` | Real article content (child of body) |
+| nav (chrome) | `div.blog-nav` | Prev/next post links (child of body) — stripped |
+| images | `img` inside `div.post-content` | Preserved (body-only, per clarification) |
+| blockquotes | `blockquote` inside content | Preserved natively |
+| subtitle/deck | — | Not present on any reference article |
+
+**Validation rules**:
+- `div.post-content` MUST exist → else `UnsupportedContentTypeError`.
+- After stripping and conversion, Markdown MUST be non-empty → else `EmptyContentError`.
+
+## Entity: Extraction Result (output)
+
+A single Markdown `str` (the public `extract()` return value):
+- Line 1: `# <title>` (top-level ATX heading).
+- Followed by the converted body: headings, paragraphs, lists, blockquotes, links, and
+  any in-body images, in document order.
+- No runs of 3+ consecutive blank lines (collapsed by base `convert_to_markdown`).
+- Contains no site chrome (nav, language selector, CTAs, TOC, share buttons, blog-nav,
+  sidebar promos, footer).
+
+## DOM Element Taxonomy (selector → action)
+
+| Selector | Action | Reason |
+|----------|--------|--------|
+| `div.post-content` | **keep** (body root) | Article body; absence ⇒ non-article |
+| `div.blog-nav` (inside body) | **strip** (`decompose`) | Prev/next post chrome |
+| `h1` (hero) | **prepend** (copy into body) | Article title heading |
+| everything outside `div.post-content` | **excluded** implicitly | Nav, TOC, share, sidebar promos, footer, hero image |
+
+## State Transitions
+
+None. Each `extract(url)` call is independent: `fetch → parse → clean → convert → return | raise`.

mdfetch-0.6.0/specs/010-boomi-blog-provider/plan.md

@@ -0,0 +1,115 @@
+# Implementation Plan: Boomi Blog Provider
+
+**Branch**: `010-boomi-blog-provider` | **Date**: 2026-06-02 | **Spec**: [spec.md](./spec.md)
+
+**Input**: Feature specification from `/specs/010-boomi-blog-provider/spec.md`
+
+## Summary
+
+Add a new provider that extracts public Boomi blog articles (`boomi.com/blog/<slug>/`) and returns them as clean Markdown. The site runs WordPress; all three reference articles share an identical, stable structure: the article body is `div.post-content`, which contains the real content (`section.wysiwyg-section`) plus a previous/next post navigation block (`div.blog-nav`) to strip. The article title is an `<h1>` rendered in the page hero (`section.post-detail-hero`), outside the body. The blog index, non-article pages, and other `boomi.com` pages do **not** contain `div.post-content`, so its absence cleanly signals "not an article" → `UnsupportedContentTypeError`. The implementation is a single new provider file subclassing `BaseExtractor`, reusing all shared fetch/convert logic.
+
+## Technical Context
+
+**Language/Version**: Python 3.12+ (matches CI matrix: 3.12–3.14)
+
+**Primary Dependencies**: `httpx` (HTTP fetch), `BeautifulSoup` / `lxml` (HTML parsing), `markdownify` (Markdown conversion), `pytest` (testing)
+
+**Storage**: N/A — stateless extraction library
+
+**Testing**: `pytest` via `uv run pytest` — unit tests (no network) + integration tests (`-m integration`, real URLs + snapshots)
+
+**Target Platform**: PyPI library (cross-platform)
+
+**Project Type**: Library
+
+**Performance Goals**: Inherits base class 30-second fetch timeout; no additional targets
+
+**Constraints**: One new provider file; no changes to shared infrastructure
+
+**Scale/Scope**: Single-article extraction per call
+
+## Constitution Check
+
+*GATE: Must pass before implementation. Re-check after design phase.*
+
+- [x] Validates Provider Pattern Architecture (new `BoomiExtractor(BaseExtractor)`; no base-class changes; no code duplication — reuses `fetch_html`/`convert_to_markdown`)
+- [x] Confirms Technology Stack (`httpx`, `BeautifulSoup`, `Markdownify`, `pytest` — all inherited)
+- [x] Adheres to Coding Standards (PEP 8, strict type hints, clear vocabulary)
+- [x] Incorporates Integration Testing (real reference URLs + snapshot containment, matching existing providers)
+- [x] Respects Packaging and Distribution standards (`pyproject.toml`, `src/` layout, `uv` for all dev commands)
+
+**Result**: PASS — no violations. Complexity Tracking not required.
+
+## Project Structure
+
+### Documentation (this feature)
+
+```text
+specs/010-boomi-blog-provider/
+├── plan.md              # This file
+├── research.md          # Phase 0 output
+├── data-model.md        # Phase 1 output
+├── quickstart.md        # Phase 1 output
+├── contracts/
+│   └── extractor-contract.md   # Phase 1 output (provider contract)
+└── tasks.md             # Phase 2 output (/speckit-tasks — NOT created here)
+```
+
+### Source Code
+
+```text
+src/mdfetch/providers/
+└── boomi.py                          # NEW: BoomiExtractor
+
+tests/unit/
+└── test_boomi_extractor.py           # NEW: unit tests (no network)
+
+tests/integration/
+├── snapshots/
+│   ├── boomi-gartner-magic-quadrant-ipaas-2026.md          # NEW
+│   ├── boomi-real-time-vs-batch-data-integration.md        # NEW
+│   └── boomi-data-consistency-saas-on-prem.md              # NEW
+└── test_boomi_integration.py         # NEW: integration tests (real URLs)
+```
+
+**Non-runtime changes**: `README.md` (add Boomi row to Supported platforms table + usage example); `pyproject.toml` (version bump `0.5.2` → `0.6.0`, new provider = minor per dev.to/Substack precedent); `CLAUDE.md` (provider tree + integration-test description + Boomi discriminator gotcha). No `test_router.py` fixture change required — the unsupported-domain fixture uses `wordpress.com`, which Boomi does not register.
+
+### Revision: Implementation Sync 2026-06-02
+- Reason: Reconciled release/docs drift found after implementation — version was not bumped and `CLAUDE.md` (project structure, integration-test description, gotchas) did not reflect the new Boomi provider. No behavioral drift; extraction output verified clean against all 3 reference articles.
+
+## Extraction Algorithm
+
+```
+extract(url):                                  # inherited from BaseExtractor
+  html ← fetch_html(url)                        # 30s timeout, 3 retries on transient errors
+  soup ← BeautifulSoup(html, "lxml")
+  body_tag ← clean_html(soup)
+  return convert_to_markdown(body_tag)          # ATX headings, collapse 3+ blank lines, EmptyContentError if blank
+
+clean_html(soup):                              # BoomiExtractor implementation
+  1. body ← soup.find("div", class_="post-content")
+        → if not a Tag: raise UnsupportedContentTypeError   # index / non-article / non-blog page
+  2. for nav in body.find_all("div", class_="blog-nav"): nav.decompose()   # strip prev/next chrome
+  3. title ← soup.find("h1")                    # hero title, outside body
+        → if Tag: body.insert(0, copy.copy(title))          # prepend as top-level heading
+  4. return body
+```
+
+**Design notes**:
+- `div.post-content` is the *discriminating* selector: present on every article, absent on the blog index and non-article pages (verified live across 3 articles + the `/blog/` index). This is why no URL-path filtering is needed (FR-009 satisfied via body-container presence, consistent with the existing route-by-domain pattern).
+- Body images live inside `post-content` and are preserved by default markdownify (Option A clarification). The hero/banner image sits in `section.post-detail-hero` — outside the body — and is therefore excluded automatically. No image-specific code is needed.
+- Blockquotes (e.g., the Gartner pull-quote) are inside `wysiwyg-section` and convert natively. No embed/iframe handling observed in references; if present, base `_replace_iframes_with_links` is available but not wired in unless a future article requires it.
+- `MATCH_SUBDOMAINS` stays `False` (default): only `boomi.com` is in scope.
+
+## Error Mapping
+
+| Condition | Exception |
+|-----------|-----------|
+| `div.post-content` not found (index / non-article / non-blog page) | `UnsupportedContentTypeError` |
+| Body found but no extractable text after conversion | `EmptyContentError` |
+| HTTP error (any non-2xx after retries) | `HTTPStatusError` |
+| Network / timeout failure | `FetchError` |
+
+## Complexity Tracking
+
+> No Constitution Check violations — section intentionally empty.

mdfetch-0.6.0/specs/010-boomi-blog-provider/quickstart.md

@@ -0,0 +1,47 @@
+# Quickstart: Boomi Blog Provider
+
+## Use it
+
+```python
+from mdfetch import extract
+
+md = extract("https://boomi.com/blog/real-time-vs-batch-data-integration-choosing-the-right-approach/")
+print(md)  # "# Real-Time vs Batch Data Integration: ..." followed by the article body
+```
+
+## Develop it (TDD-friendly order)
+
+```bash
+make setup            # uv sync --all-extras
+
+# 1. Add src/mdfetch/providers/boomi.py (BoomiExtractor) — see plan.md "Extraction Algorithm"
+# 2. Write tests/unit/test_boomi_extractor.py (no network) using sample HTML fixtures
+make test             # unit tests only — should pass without network
+make typecheck        # mypy src/ — zero errors
+make lint             # ruff check
+
+# 3. Add integration test + snapshots
+make integration      # network required; hits real boomi.com URLs
+```
+
+## Reference URLs
+
+| Slug | Use |
+|------|-----|
+| `gartner-magic-quadrant-ipaas-2026` | has 1 in-body image + blockquote |
+| `real-time-vs-batch-data-integration-choosing-the-right-approach` | headings + lists |
+| `how-to-maintain-data-consistency-across-saas-and-on-prem-systems` | headings + blockquote |
+| `https://boomi.com/blog/` (index) | non-article → expect `UnsupportedContentTypeError` |
+
+## Generate a snapshot (first 30 lines, blank lines preserved)
+
+```bash
+uv run python -c "from mdfetch import extract; c=extract('<url>'); \
+open('tests/integration/snapshots/boomi-<slug>.md','w',encoding='utf-8').write('\n'.join(c.split('\n')[:30]).rstrip())"
+```
+
+## Done when
+
+- `make test`, `make typecheck`, `make lint` all green.
+- `make integration` passes (snapshot containment for the 3 articles; index raises `UnsupportedContentTypeError`).
+- `README.md` Supported platforms table includes a `Boomi | boomi.com` row.

mdfetch-0.6.0/specs/010-boomi-blog-provider/research.md

@@ -0,0 +1,76 @@
+# Phase 0 Research: Boomi Blog Provider
+
+All findings below come from live DOM inspection (2026-06-02) of the three reference
+articles plus the `/blog/` index, using the project's browser-like User-Agent.
+
+## Decision: Article body container = `div.post-content`
+
+- **Decision**: Isolate the article body by selecting `soup.find("div", class_="post-content")`.
+- **Rationale**: Present on every article page; its direct children are exactly
+  `section.wysiwyg-section.bullet-styled` (the real content) and `div.blog-nav`
+  (prev/next chrome). Crucially, the `/blog/` **index page does NOT contain
+  `div.post-content`** (it has a `wysiwyg-section` intro but no `post-content`),
+  so the selector doubles as the article-vs-non-article discriminator.
+- **Alternatives considered**:
+  - `section.wysiwyg-section` — **rejected**: also present on the blog index, so it
+    would fail to raise `UnsupportedContentTypeError` for non-articles.
+  - `<article>` / `<main>` landmarks — **rejected**: not emitted with usable classes
+    in this WordPress theme.
+
+## Decision: Strip `div.blog-nav`
+
+- **Decision**: `decompose()` every `div.blog-nav` inside the body before conversion.
+- **Rationale**: It is the only chrome element living *inside* `post-content`; it holds
+  the "Previous / Next" post links. All other chrome (top nav, language selector,
+  login/demo CTAs, "On this page" TOC, social share, sidebar report promos, footer)
+  lives **outside** `post-content` and is excluded automatically by isolating the body.
+- **Alternatives considered**: Stripping each chrome class individually — **rejected** as
+  unnecessary; isolating `post-content` already removes everything except `blog-nav`.
+
+## Decision: Title from hero `<h1>`, prepended to body
+
+- **Decision**: `soup.find("h1")` and prepend a copy to the body as the top-level heading.
+- **Rationale**: The title `<h1>` is rendered in `section.post-detail-hero` (the page hero),
+  structurally outside `post-content`, so an unconditional prepend includes it exactly once
+  with no body-scan deduplication (same approach as `SubstackExtractor`). There is exactly
+  one `<h1>` per page. No subtitle/deck element is present on any reference article.
+- **Alternatives considered**: `og:title` meta — **rejected**: prefer the on-page `<h1>` for
+  consistency with other providers and to avoid meta/visible-title drift.
+
+## Decision: Image handling — body-only (Option A)
+
+- **Decision**: Convert only images inside `post-content`; do not hoist the hero image.
+- **Rationale**: Matches the spec clarification (2026-06-02). The hero/banner image lives in
+  `section.post-detail-hero` (outside the body) and is excluded automatically; genuine inline
+  content images (e.g., 1 image inside reference #1's body) are preserved by default markdownify.
+  No image-specific code required.
+- **Evidence**: img count inside `post-content` — ref#1 (gartner): 1; ref#2 (real-time-vs-batch): 0;
+  ref#3 (data-consistency): 0. All have 1 blockquote (Gartner / pull-quotes), converted natively.
+
+## Decision: No URL path filtering; route by domain only
+
+- **Decision**: Register `DOMAINS = frozenset({"boomi.com"})`, `MATCH_SUBDOMAINS = False`.
+  Rely on `post-content` presence to reject non-`/blog/` pages.
+- **Rationale**: Consistent with the codebase's route-by-domain + validate-body pattern. Non-blog
+  `boomi.com` pages and the blog index lack `post-content` → `UnsupportedContentTypeError`,
+  satisfying FR-009 without bespoke path logic.
+
+## Decision: No new infrastructure / markdownify overrides
+
+- **Decision**: Inherit `fetch_html`, `convert_to_markdown`, and default markdownify kwargs.
+- **Rationale**: No code blocks observed in references (so no `code_language_callback` like DZone),
+  no embeds requiring link conversion. Keeps the provider minimal per the Provider Pattern principle.
+
+## Platform facts summary
+
+| Fact | Value |
+|------|-------|
+| CMS | WordPress 6.9.4 |
+| Domain | `boomi.com` (articles under `/blog/<slug>/`) |
+| Body container | `div.post-content` |
+| Real content | `section.wysiwyg-section.bullet-styled` |
+| In-body chrome | `div.blog-nav` (prev/next) |
+| Title | single `<h1>` in `section.post-detail-hero` (outside body) |
+| Subtitle/deck | none |
+| Paywall | none (freely readable) |
+| Index page has `post-content`? | No (→ clean non-article detection) |

mdfetch-0.6.0/specs/010-boomi-blog-provider/spec.md

@@ -0,0 +1,99 @@
+# Feature Specification: Boomi Blog Provider
+
+**Feature Branch**: `010-boomi-blog-provider`
+
+**Created**: 2026-06-02
+
+**Status**: Draft
+
+**Input**: User description: "I would like to add support of the articles from https://boomi.com/blog/. Please use the following ones as references: 1. https://boomi.com/blog/gartner-magic-quadrant-ipaas-2026/ 2. https://boomi.com/blog/real-time-vs-batch-data-integration-choosing-the-right-approach/ 3. https://boomi.com/blog/how-to-maintain-data-consistency-across-saas-and-on-prem-systems/"
+
+## User Scenarios & Testing *(mandatory)*
+
+### User Story 1 - Extract a Boomi blog article as clean Markdown (Priority: P1)
+
+A user passes the URL of a public Boomi blog article to the library and receives the article's title and full body rendered as clean Markdown, with all site chrome (navigation, CTAs, share buttons, related-post links, footer) removed.
+
+**Why this priority**: This is the core value of the feature — without it, Boomi blog articles cannot be consumed at all. It is the minimum viable slice and delivers the complete user benefit on its own.
+
+**Independent Test**: Can be fully tested by calling `extract()` with one of the three reference URLs and verifying the returned Markdown begins with the article title as a top-level heading and contains the body's section headings and paragraph text, with no navigation, subscription/demo CTAs, or footer text.
+
+**Acceptance Scenarios**:
+
+1. **Given** a public Boomi blog article URL, **When** `extract()` is called, **Then** the result is Markdown whose first line is `# <article title>` followed by the body content.
+2. **Given** a Boomi blog article containing H2 sections and bullet lists, **When** extracted, **Then** the headings and lists are preserved as Markdown headings and lists in document order.
+3. **Given** a Boomi blog article, **When** extracted, **Then** the output contains none of: top navigation, language selector, login/demo CTAs, "on this page" jump links, social share buttons, related/previous/next post links, sidebar report promotions, or footer links.
+
+---
+
+### User Story 2 - Reject non-article Boomi URLs cleanly (Priority: P2)
+
+A user passes a Boomi URL that is not a readable blog article (the blog index, a category listing, or a marketing page on `boomi.com`). The library raises a typed error rather than returning garbage or partial chrome.
+
+**Why this priority**: Protects users from silently receiving meaningless output and keeps behavior consistent with existing providers. Valuable but secondary to the happy path.
+
+**Independent Test**: Can be tested by calling `extract()` with a non-article Boomi URL (e.g., the blog index `https://boomi.com/blog/`) and asserting that a typed extraction error is raised.
+
+**Acceptance Scenarios**:
+
+1. **Given** the Boomi blog index URL, **When** `extract()` is called, **Then** an `UnsupportedContentTypeError` is raised because no single article body is present.
+2. **Given** a Boomi article whose body element is present but yields no text after stripping, **When** `extract()` is called, **Then** an `EmptyContentError` is raised.
+
+---
+
+### Edge Cases
+
+- **Non-article pages**: The Boomi blog index (`/blog/`), category/tag listings, and non-blog marketing pages on `boomi.com` do not contain a single recognizable article body → `UnsupportedContentTypeError`.
+- **Hero/inline images**: Reference articles include a featured hero image and may include inline images. Only images inside the article body container are preserved as Markdown images; the featured hero/banner image (rendered outside the body) and other chrome images are stripped with their containers.
+- **Blockquotes**: Some articles embed pull quotes (e.g., the Gartner quote in reference #1). These are preserved as Markdown blockquotes.
+- **Legal disclaimers**: Trailing legal/attribution text that is part of the article body (e.g., the Gartner disclaimer) is preserved as body content; site-wide footer legal links are not.
+- **HTTP errors**: A 404/410 for a removed article, or a 429/503 transient error, is surfaced via the base class's existing fetch/error behavior — this feature introduces no new network handling.
+- **No embedded media in references**: The reference articles contain no code blocks, videos, tweets, or iframes; should such embeds appear in other Boomi articles, they are handled by the inherited base conversion behavior, not by feature-specific logic.
+
+## Requirements *(mandatory)*
+
+### Functional Requirements
+
+- **FR-001**: The system MUST route `boomi.com` blog URLs to the new provider using the existing domain-registration mechanism (`@register` decorator + `DOMAINS` frozenset).
+- **FR-002**: The system MUST extract the main article body from a Boomi blog post page and return it as clean Markdown.
+- **FR-003**: The system MUST strip all non-content elements before conversion: top navigation, language selector, login/demo CTAs, "on this page" jump links, social share buttons, related/previous/next post navigation, sidebar report/promo callouts, and the site footer.
+- **FR-004**: The system MUST prepend the article title as a top-level Markdown heading (`# Title`).
+- **FR-005**: The system MUST preserve structural body content: section headings, paragraphs, ordered and unordered lists, blockquotes, hyperlinks, and images located inside the article body container. Images outside the body container (e.g., a featured hero/banner image rendered in a page header) MUST be treated as page chrome and excluded.
+- **FR-006**: The system MUST raise `UnsupportedContentTypeError` when the page does not contain a recognizable single article body element (e.g., the blog index or a non-article page).
+- **FR-007**: The system MUST raise `EmptyContentError` when the article body yields no extractable text after stripping.
+- **FR-008**: The system MUST collapse runs of three or more consecutive blank lines to a single blank line.
+- **FR-009**: The system MUST yield article output only for pages that contain a recognizable article body container. Pages on `boomi.com` without one — the blog index, category/tag listings, and non-blog pages — MUST NOT yield article output and MUST raise `UnsupportedContentTypeError`. (In practice all article pages live under `boomi.com/blog/<slug>/`; scoping is enforced by body-container presence rather than URL path matching.)
+
+### Key Entities *(include if feature involves data)*
+
+- **Boomi Blog Post**: A single article page under `https://boomi.com/blog/<slug>/`. Key attributes: title, body content, author (e.g., "Ed Macosky", "Boomi"), publication date, category, optional hero image. Subtitle/deck is generally absent.
+- **DOM Element Taxonomy**: Mapping of CSS selectors for the Boomi blog layout to actions (keep the article body container; strip navigation, language selector, CTAs, jump links, share buttons, related-post navigation, sidebar promos, and footer).
+
+## Success Criteria *(mandatory)*
+
+### Measurable Outcomes
+
+- **SC-001**: Each of the three reference articles returns Markdown containing the full title and all body section headings and paragraph text, with zero non-content elements (navigation, CTAs, share buttons, related-post links, footer).
+- **SC-002**: Extraction completes within the base class 30-second fetch timeout on a stable connection.
+- **SC-003**: A non-article Boomi URL (the blog index) raises `UnsupportedContentTypeError` within the normal fetch timeout.
+- **SC-004**: The extracted Markdown contains no consecutive blank-line runs of three or more lines.
+- **SC-005**: The new provider is exercised by at least one integration test using a real network request against a reference URL, matching the pattern established by existing providers.
+
+## Clarifications
+
+### Session 2026-06-02
+
+- Q: How should the featured/hero image be handled? → A: Body-only images — include only images inside the article body container; the featured hero/banner image (outside the body) is treated as page chrome and excluded.
+
+## Assumptions
+
+- The Boomi blog's public HTML structure is stable enough for CSS-selector-based extraction; a major redesign would require an extractor update.
+- All target content is freely readable; no authentication or paywall handling is required (confirmed across the three reference articles).
+- Only the primary `boomi.com` domain is in scope. Extraction is limited to article pages under the `/blog/` path; the blog index and other `boomi.com` pages are treated as unsupported content rather than routed to a different provider.
+- Subtitle/deck handling is best-effort: when no deck is present (as in the references), only the title heading is prepended.
+- The implementation follows the existing provider pattern: one new file under `src/mdfetch/providers/` subclassing `BaseExtractor` and reusing its shared conversion methods, with no changes to shared infrastructure.
+- The base class retry/timeout behavior (existing defaults) is inherited without modification.
+- The `test_router.py` unsupported-domain fixture remains valid (Boomi registers `boomi.com`, which is not currently used as the unsupported-domain example — `wordpress.com` is); no fixture update is required unless that changes.
+
+### Revision: Implementation Sync 2026-06-02
+- Reason: Post-implementation audit confirmed the shipped extraction matches all functional requirements with no behavioral drift. Reconciled release/docs drift only (version bump + `CLAUDE.md` provider listing); no spec requirements changed.