ag-pii-vault

PII tokenization for LLM pipelines.

Detects personal data (names, emails, phone numbers and other Presidio entities), replaces it with deterministic HMAC tokens before text is sent to an LLM, and restores it in the response. Token mappings live in a local SQLite vault or in the hosted vault.

Install

pip install "ag-pii-vault[all]"
npm install ag-pii-vault

Extras: openai, anthropic, all. Detection uses the spaCy model en_core_web_lg: python -m spacy download en_core_web_lg.

Usage

from ag_pii_vault import Detector, LocalVault, Tokenizer

detector = Detector()
tokenizer = Tokenizer(vault=LocalVault(path=":memory:"), key="my-hmac-secret")

text = "Contact John Smith at john@acme.com"
safe = tokenizer.tokenize(text, detector.analyze(text))
print(safe)                        # PII replaced by tokens
print(tokenizer.dehydrate(safe))   # original text restored

Output:

Contact PERSON_353fcd26 at EMAIL_9abc0eb1
Contact John Smith at john@acme.com

SafeOpenAI and SafeAnthropic wrap the official clients: messages are tokenized on the way out and restored on the way back.

The TypeScript package ag-pii-vault is a port of the core classes.

Hosted API

HostedVault stores the token mappings in the hosted API instead of local SQLite. Pass hosted_api_key="ag_live_..." to SafeOpenAI or SafeAnthropic. API keys are issued at https://ai.artifact-network.org. The default endpoint is https://ai-api.artifact-network.org; pass a different base URL to use your own deployment.

Development

cd packages/python && pip install -e ".[all]" && pytest
cd packages/typescript && npm ci && npm test

License

MIT, Copyright (c) 2026 Adnan Ghribi.