blog-pipeline 0.1.2

AI blog generator with 6-pass pipeline and built-in humanizer that removes AI writing tells.

blog-pipeline

AI blog generator that doesn't sound like AI.

6-pass Claude API pipeline with a built-in humanizer, topic deduplication, internal linking, and Supabase sync. The humanizer is the key differentiator — it enforces a strict writing ruleset that removes every common AI tell.

---

Quick start

# Clone and install
git clone https://github.com/nometria/blog-pipeline
cd blog-pipeline
pip install -e .

# Set your API key
export ANTHROPIC_API_KEY=sk-ant-...

# Generate 5 blog posts
blog-generate --count 5 --niche "developer tooling and SaaS"

# Re-humanize existing drafts only
blog-generate --passes 4

# Run tests
pytest tests/ -v

Required environment variables:

ANTHROPIC_API_KEY=sk-ant-...          # required

# Optional (for Supabase sync in pass 0 and 6)
SUPABASE_URL=https://xxx.supabase.co
SUPABASE_KEY=eyJ...
BLOG_SITE_URL=https://yourblog.com

---

Passes

Pass	What it does
0	Fetches existing titles from Supabase (prevents duplicates)
1	Identifies new topics (skips anything already written)
2	Plans structure per topic (comparison / deep-dive / case-study / how-to / opinion)
3	Generates full markdown content
4	Humanizer — strips AI tells (see below)
5	Adds internal links across all posts
6	Pushes to Supabase + updates local registry

---

The Humanizer

Pass 4 enforces these rules on every post:

• Banned words: leverage, seamless, robust, cutting-edge, game-changer, revolutionize, synergy, paradigm, transformative, unlock, delve, streamline, elevate, empower, holistic, utilize, facilitate, innovative
• No em-dashes (—) — replaced with commas or full stops
• No semicolons connecting sentences
• No emojis
• Contractions required: it's, we're, you'll, don't
• Active voice only
• Max 1 exclamation mark per post
• No "In conclusion / In summary" section openers

Use the humanizer standalone:

from blog_pipeline.humanizer import humanize_post
clean = humanize_post(my_ai_draft)

---

Setup

git clone https://github.com/nometria/blog-pipeline
cd blog-pipeline
pip install -e .
cp .env.example .env
# Edit .env with your ANTHROPIC_API_KEY

---

Run

# Full pipeline: generate 5 blogs
blog-generate --passes 1-6 --count 5 --niche "developer tooling and SaaS"

# Re-humanize existing drafts only
blog-generate --passes 4

# Push already-written files to Supabase
blog-generate --passes 6

# Generate content without pushing
blog-generate --passes 1-5 --count 3

---

Output

• blogs/<slug>.md — humanized markdown files
• blogs/_topics.json — topic cache
• blogs/_plans.json — structure plans
• blogs/_registry.json — pushed blog tracking

---

Immediate next steps

1. Make the humanizer prompt configurable via HUMANIZER_RULES env / YAML
2. Add --audit flag: re-score all pushed blogs and unpublish weak ones
3. Add SEO scoring pass (keyword density check, meta description generation)
4. Package as a GitHub Action: auto-generate blogs on schedule

---

Commercial viability

• Package the humanizer as a standalone API: POST /humanize → clean post
• Charge per post ($0.10–0.50) or monthly flat ($49–149)
• Differentiator: "the only AI blog writer that bans its own clichés by design"
• Add AI-detector score before/after to prove improvement

---

Example output

Running pytest tests/ -v:

============================= test session starts ==============================
platform darwin -- Python 3.13.9, pytest-9.0.2, pluggy-1.5.0
cachedir: .pytest_cache
rootdir: /tmp/ownmy-releases/blog-pipeline
configfile: pyproject.toml
plugins: anyio-4.12.1, cov-7.1.0
collecting ... collected 4 items

tests/test_pipeline.py::test_check_banned_words_flags_corporate_speak PASSED [ 25%]
tests/test_pipeline.py::test_check_banned_words_passes_clean_text PASSED [ 50%]
tests/test_pipeline.py::test_check_banned_words_flags_em_dash_clusters FAILED [ 75%]
tests/test_pipeline.py::test_humanize_post_returns_string PASSED         [100%]

============================= short test summary info ==========================
FAILED tests/test_pipeline.py::test_check_banned_words_flags_em_dash_clusters
========================= 1 failed, 3 passed in 0.43s ==========================

See examples/sample-post.md for a realistic humanized blog post produced by the pipeline.