VerifIA is an open-source AI testing framework for domain‑aware verification of AI models during the staging phase—before deployment.
Project description
Domain‑Aware Verification Framework for AI Models
Docs • Website
VerifIA is an open‑source Python library that automates domain‑aware verification of machine‑learning models during the staging phase—before deployment. It generates novel, in‑domain inputs and checks your model against expert‑defined rules, constraints, and specifications, helping you:
- ✅ Validate behavioral consistency with domain knowledge
- 🔍 Detect edge‑case failures beyond your labeled data
- 📊 Generate comprehensive HTML reports for decision‑making and debugging
📖 Try in Colab
🚀 Install
# Core framework
pip install verifia
# Include AI‑Based Domain Generation
pip install verifia[genflow]
Supports Python 3.10+.
🤸♀️ Quickstart
from verifia.verification import RuleConsistencyVerifier
# 1. Load your domain spec
verifier = RuleConsistencyVerifier("domain_rules.yaml")
# 2. Attach model and data
report = (
verifier
.verify(model_card_fpath_or_dict="model_card.yaml")
.on(data_fpath="test_data.csv") # .csv, .json, .xlsx, .parquet, .feather, .pkl
.using("GA") # RS, FFA, MFO, GWO, MVO, PSO, WOA, GA, SSA
.run(pop_size=50, max_iters=100) # search budget
)
# 3. Save your report
report.save_as_html("verification_report.html")
Quickstart Steps
- Install: install
- Prepare Your Components (Domain, Model, Data): prepare-your-components
- Run Verification: run-a-verification
- Inspect Results: inspecting-results
👉 Full Quickstart guide: https://docs.verifia.ca/quickstart
📚 Feature Spotlight: AI‑Based Domain Generation
Automatically build your domain specification from CSVs, DataFrames, and PDFs using LLM‑powered agents. No manual rule‑writing required—point VerifIA at your data and let it generate variables, constraints, and rules for you.
🎬 Demo: AI‑Based Domain Generation UI
Fig. Interactive animated demo—click to open full resolution.
📖 Learn More
- Prerequisites & Setup: #environment-setup
- Prepare Domain Spec: #prepare-your-domain
- Run Generation: #run-domain-generation
🧰 Ecosystem & Integrations
VerifIA works with any model, in any environment and integrates seamlessly with your favorite tools ⤵️
📖 More Resources
- Documentation: https://docs.verifia.ca
- Website: https://verifia.ca
- Source Code: https://github.com/VerifIA/verifia
- Contact: contact@verifia.ca
🤝 Contributing
We welcome all contributions! Please read our CONTRIBUTING.md to get started.
⚖️ License
VerifIA is released under the AGPL‑3.0 License. See LICENSE for details.
Made with ❤️ by the VerifIA contributors.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file verifia-0.1.1.tar.gz.
File metadata
- Download URL: verifia-0.1.1.tar.gz
- Upload date:
- Size: 120.4 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.1.0 CPython/3.10.10
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
476198829dbd2b40d9a03745817462995c1e4e0c6e6b0a043bbef5cb1454cbbb
|
|
| MD5 |
1cd17124eb5f02234158107ca04436aa
|
|
| BLAKE2b-256 |
ef46e3b0e16961ff15d06fed5e3eef6c9a92e0b476b284e3040398fa518ed343
|
File details
Details for the file verifia-0.1.1-py3-none-any.whl.
File metadata
- Download URL: verifia-0.1.1-py3-none-any.whl
- Upload date:
- Size: 125.3 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.1.0 CPython/3.10.10
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
31af66b9f388934f94d26f0421f04eb624224e3c12aa2b38297c623fc728c8f0
|
|
| MD5 |
f3fececfde31eca840ff571a2031d927
|
|
| BLAKE2b-256 |
29c0dd43994604048d60e5ea2152bb72f2d8bd6d846c1bb274dfbd77b846d62c
|