Automated SEO Blog Pipeline: Gemini + CRAAP Test Source Verification + WordPress REST API

1from datetime import datetime, timedelta
2from urllib.parse import urlparse
3import re
4
5class SourceVerifier:
6    TRUSTED_DOMAINS = {
7        ".gov", ".edu", "reuters.com", "bbc.com", "ft.com",
8        "cloud.google.com", "arxiv.org", "nature.com", "techcrunch.com"
9    }
10    BIAS_PATTERNS = [
11        r"buy now", r"limited offer", r"click here", r"sponsored"
12    ]
13
14    def score(self, source: dict, topic: str) -> float:
15        """Return 0–100 CRAAP score. Below 60 = reject."""
16        score = 0.0
17
18        # Currency (25 points): how recent?
19        pub_date = source.get("published_date")
20        if pub_date:
21            age_months = (datetime.now() - pub_date).days / 30
22            if age_months <= 6: score += 25
23            elif age_months <= 18: score += 15
24            elif age_months <= 36: score += 5
25
26        # Relevance (20 points): semantic similarity to topic
27        similarity = self._cosine_similarity(
28            self._embed(source["title"] + " " + source["snippet"]),
29            self._embed(topic)
30        )
31        score += similarity * 20
32
33        # Authority (25 points): domain credibility
34        domain = urlparse(source["url"]).netloc.lower()
35        if any(d in domain for d in self.TRUSTED_DOMAINS):
36            score += 25
37        elif domain.endswith((".com", ".org", ".net")):
38            score += 12
39
40        # Accuracy (15 points): cross-referenced?
41        if source.get("corroborated_by_n_sources", 0) >= 2:
42            score += 15
43
44        # Purpose (15 points): informational, not commercial
45        content = source.get("snippet", "").lower()
46        has_bias = any(re.search(p, content) for p in self.BIAS_PATTERNS)
47        if not has_bias:
48            score += 15
49
50        return score
51
52    def filter_sources(self, sources: list, topic: str) -> list:
53        scored = [(s, self.score(s, topic)) for s in sources]
54        verified = [s for s, score in scored if score >= 60]
55        if len(verified) < 3:
56            raise InsufficientSourcesError(
57                f"Only {len(verified)} verified sources — needs ≥3 to proceed"
58            )
59        return verified

1def build_generation_prompt(
2    topic: str,
3    sources: list[dict],
4    target_keyword: str,
5    lsi_keywords: list[str],
6) -> str:
7    source_context = "\n".join([
8        f"- [{s['title']}]({s['url']}) — {s['snippet']}"
9        for s in sources
10    ])
11
12    return f"""Write a 1,500–2,500 word blog article about: {topic}
13
14PRIMARY KEYWORD: {target_keyword} (use naturally at 1.5–2.5% density)
15LSI KEYWORDS: {', '.join(lsi_keywords)} (vary these throughout)
16
17SOURCES (cite these, do not invent others):
18{source_context}
19
20STRUCTURE REQUIREMENTS:
211. Title containing the primary keyword
222. Opening paragraph with keyword in first 100 words
233. 4–6 H2 sections with keyword variations in headings
244. Inline citations: [Source Name](URL) — only from the provided sources
255. Conclusion with keyword and clear takeaway
266. Meta description: exactly 150–155 characters
27
28STYLE: Informative, authoritative. Avoid: "delve into", "In conclusion", filler phrases.
29AI DISCLOSURE: Add one sentence: "This article was researched and drafted with AI assistance."
30"""

1import requests
2from base64 import b64encode
3
4class WordPressPublisher:
5    def __init__(self, site_url: str, username: str, app_password: str):
6        credentials = b64encode(f"{username}:{app_password}".encode()).decode()
7        self.headers = {
8            "Authorization": f"Basic {credentials}",
9            "Content-Type": "application/json",
10        }
11        self.api_url = f"{site_url}/wp-json/wp/v2"
12
13    def publish_post(self, article: dict) -> dict:
14        payload = {
15            "title": article["title"],
16            "content": article["html_content"],
17            "status": "publish",  # or "draft" for review
18            "categories": [article["category_id"]],
19            "tags": article["tag_ids"],
20            # Yoast SEO meta (requires Yoast plugin)
21            "yoast_head_json": {
22                "title": article["seo_title"],
23                "description": article["meta_description"],
24                "og_title": article["seo_title"],
25                "og_description": article["meta_description"],
26            },
27        }
28        response = requests.post(
29            f"{self.api_url}/posts",
30            json=payload,
31            headers=self.headers,
32        )
33        response.raise_for_status()
34        return response.json()