Skip to content

Quickstart

Get lexvault masking your proprietary terms in under 2 minutes.

1. Install

pip install lexvault

lexvault needs Python 3.10+ and runs as a LiteLLM proxy guardrail.

2. Create a dictionary

Define the proprietary terms you want masked. These are YOUR terms — codenames, products, customers, schema names — not classic PII.

# dictionary.yaml
terms:
  - term: Project Titan
    type: codename
  - term: customer_database
    type: schema
regex_terms:
  - name: Employee ID
    pattern: 'EMP-\d{4,6}'
    type: id

3. Configure LiteLLM

Mount the shim (LiteLLM's guardrail loader does a naive split(".") and loads <file>.py from the config dir, so you need the one-line shim next to your config):

# config.yaml
model_list:
  - model_name: gpt-4o
    litellm_params:
      model: openai/gpt-4o
      api_key: os.environ/OPENAI_API_KEY

guardrails:
  - guardrail_name: lexvault
    litellm_params:
      guardrail: lexvault_shim.LexVaultGuardrail
      mode: [pre_call, post_call]
      default_on: true
      dictionary_path: dictionary.yaml
      org_key: os.environ/LEXVAULT_ORG_KEY
      scope: default

# lexvault_shim.py (next to config.yaml)
from lexvault import LexVaultGuardrail as LexVaultGuardrail  # noqa: F401

4. Run

export OPENAI_API_KEY=sk-...
export LEXVAULT_ORG_KEY=$(openssl rand -hex 32)   # keep this secret + stable
litellm --config config.yaml

5. Verify

curl http://localhost:4000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [{"role": "user", "content": "What is Project Titan?"}]
  }'
  • The upstream LLM receives [LEX-NZNH3BZX], not Project Titan.
  • The client response has Project Titan restored — you never see the placeholder.

Docker

The fastest way to try it without a real API key is the Docker quickstart, which boots LiteLLM + a fake backend that echoes requests so you can see the mask/restore round-trip directly.

git clone https://github.com/nitishagar/lexvault.git
cd lexvault
docker compose -f examples/docker-compose.yml up --build

Next steps

  • Concepts — how masking, scopes, and deterministic placeholders work.
  • Configuration — every config key.
  • Integration — LiteLLM proxy setup, per-request overrides, guardrail ordering.