Clearview API

A document intelligence API that analyzes articles, web pages, and PDFs for bias, scam likelihood, and subjectivity — delivering results as a formatted PDF report or spoken audio.

Docs: clearwayapi.tech/docs

---

How it works

1. Extract — a URL or raw text is scraped and cleaned into plain content
2. Analyze — OpenAI GPT-4o-mini evaluates the content for scam signals, subjectivity, and bias, returning structured verdicts (YES/NO) alongside human-readable explanations
3. Report — results are compiled into a multi-page Clearview PDF (summary, analysis, source)
4. Audio — the extracted text is narrated via ElevenLabs TTS (with a gTTS fallback)

---

Tech stack

Layer	Technology
API framework	FastAPI + Uvicorn
AI analysis	OpenAI GPT-4o-mini
Text-to-speech	ElevenLabs (eleven_turbo_v2_5) / gTTS fallback
PDF generation	fpdf2
Web scraping	newspaper4k, PyMuPDF
Storage	Firebase

---

Built with Claude

This project was developed with Claude (Anthropic) as an AI pair programmer throughout the hackathon. Claude was used to:

• Design and implement the unified analysis service (analysis_service.py), replacing three separate HuggingFace classifier pipelines with a single GPT-4o-mini call
• Architect the structured YES/NO verdict parsing system — prompting GPT to emit SCAM: YES/NO, SUBJECTIVE: YES/NO, and BIASED: YES/NO as machine-readable first lines, then using regex to extract verdicts and drive frontend color-coding
• Build the PDF generation pipeline with fpdf2
• Write and refactor the extraction, audio, and routing layers
• Debug merge conflicts and port features across architectural changes

---

Quickstart

python -m venv .venv
source .venv/Scripts/activate  # Windows (bash)
pip install -r requirements.txt
uvicorn src.main:app --reload

Create a .env file with the following keys:

OPENAI_API_KEY=sk-...
ELEVENLABS_API_KEY=...

ElevenLabs is optional — the API falls back to gTTS if the key is absent.

---

Endpoints

Method	Path	Description
POST	/clearview	Analyze a document and receive a Clearview PDF report
POST	/audio	Convert a document to an MP3 audio file

See the full documentation for request/response schemas and code examples.