README.md

metadata

title: Amanah AI
emoji: 🕌
colorFrom: green
colorTo: yellow
sdk: static
app_file: index.html
pinned: false
license: mit
short_description: Quran and Hadith Quote Verification

Amanah AI

AI-Powered Verification and Correction of Quranic and Prophetic Quotations

Amanah AI is an evidence-grounded tool for detecting, verifying, and correcting Quranic verses and Prophetic Hadith quotations in AI-generated Islamic content.

It identifies religious quotations, retrieves matching source texts, compares wording against the available sources, and presents verification evidence. When reliable evidence is insufficient, the system abstains from automatic correction and recommends human review.

Features

• Ask & Verify: Ask a question, receive an answer, and check the religious quotations it contains.
• Verify Directly: Paste any text to detect and verify Quranic and Hadith quotations.
• Evidence-Based Verification: Compare detected quotations against the bundled source texts.
• Source-Backed Correction: Correct Quranic wording using the original source text when sufficient evidence is available.
• Human Review: Refer uncertain cases for human verification rather than inventing corrections.
• Try an Example: Use the built-in examples to explore different verification scenarios, including correct, incorrect, mixed, and uncertain quotations.

How It Works

The system follows an integrated workflow:

1. Detect potential Quranic and Hadith quotations.
2. Retrieve relevant source passages.
3. Compare the detected wording with the retrieved sources.
4. Present evidence and the verification decision.
5. Provide a source-backed correction when justified, or recommend human review.

The verification pipeline runs in the browser using Python through Pyodide. The optional question-answering feature communicates with a server-side proxy.

Privacy: Text submitted to the verification pipeline is processed in the browser. The optional question-answering feature sends its prompt to the configured server-side proxy.

For the Arabic documentation, see README_AR.md.

Data Sources and Attribution

Quran text: The Quran text is sourced from the Tanzil Project and is distributed unchanged under the stated CC BY 3.0 terms. Please consult the original source for attribution requirements.

Hadith and development data: The project uses the six Hadith collections and development data distributed through the IslamicEval 2025 Subtask 1 repository, identified as Apache-2.0 in the project documentation. The Hadith texts originate from OmarShafie/hadith, with attribution to Shafie (2021).

Please consult the original repositories and the project's REFERENCES.md for detailed source attribution and licensing information.

Technology

• Python
• Pyodide
• Browser-based text processing
• Information retrieval and text similarity
• Cloudflare Worker for the optional question-answering proxy

License

The project code is licensed under the MIT License, as indicated by the repository. Data files may have separate licensing and attribution requirements; consult the relevant original sources and REFERENCES.md.