anontex 0.2.4

PII anonymizer as a LLM proxy

AnonTex

AnonTex is a privacy-first experimental LLM proxy that anonymizes Personally Identifiable Information (PII) before forwarding requests to the OpenAI Completion API. It is designed to be compatible with the /v1/chat/completions endpoint, making it a drop-in proxy with minimal integration effort.

⚠️ Note: This is an experimental project. Use with caution in production environments.

---

✨ Features

• Acts as a transparent proxy for OpenAI's chat completion endpoint.
• Automatically anonymizes user input using PII detection.
• Compatible with OpenAI clients, such as the OpenAI Python SDK and LangChain.
• Redis-backed for entity management and fast caching.

---

📦 Installation

Install via pip:

pip install anontex

✅ Note: Redis is a required external dependency for caching and PII management. Make sure you have Redis running locally or remotely.

---

🚀 Usage

Once installed and configured, AnonTex runs a proxy server compatible with OpenAI’s Chat Completion API.

🔁 Example with curl

curl --request POST \
  --url http://localhost:8000/v1/chat/completions \
  --header 'Authorization: Bearer YOUR-OPENAI-API-KEY' \
  --header 'Content-Type: application/json' \
  --data '{
    "model": "gpt-4o-mini",
    "messages": [
      {
        "role": "system",
        "content": "You are a helpful assistant."
      },
      {
        "role": "user",
        "content": "Hello! My name is John Smith"
      }
    ]
  }'

🐍 Example with OpenAI Python SDK

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:8000/v1",
    api_key="YOUR-OPENAI-API-KEY"
)

response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "My email is [e-mail]."}
    ]
)

print(response.choices[0].message.content)

🔗 Example with LangChain

from langchain.chat_models import ChatOpenAI
from langchain.schema import HumanMessage

chat = ChatOpenAI(
    openai_api_key="YOUR-OPENAI-API-KEY",
    openai_api_base="http://localhost:8000/v1"
)

messages = [HumanMessage(content="My phone number is 123-456-7890.")]
response = chat(messages)
print(response.content)

---

⚙️ Configuration

Running Locally

Start the proxy via CLI:

anontex run

CLI Options

• --host: Server host (default: 0.0.0.0)
• --port: Server port (default: 8000)
• --config: Path to configuration file (default: spacy engine configs)
• --log-level: Logging level (default: info)

Config File (Optional)

You can pass settings via a YAML config file. Read the following documentation to customize the config file.

This project uses the presidio-analyzer Python package as an entity detector. You can use the default config file without specifying a custom file or point to a presidio-analyzer supported config file.

.env File (Optional)

Additional configurations can be done via environment variables in a .env file. If .env is not set, default values will be used. Read the following documentation to customize the .env file.

---

🐳 Docker Deployment

You can deploy AnonTex with Docker using Docker Compose.

Clone repo:

git clone https://github.com/ChamathKB/AnonTex

Run:

docker compose up -d

---

🚧 Limitations & Future Improvements

• ❌ No support for multi-turn PII tracking (PII memory is per-message only).
• 🔗 Only supports OpenAI API compatible endpoints.
• 🌐 Limited language support (primarily English).
• 📈 Planned support for:
  • Multi-turn entity memory
  • Custom anonymization rules
  • Model switching and vendor abstraction
  • Analytics & tracing integration

---

🤝 Contributing

Pull requests are welcome! For major changes, please open an issue first to discuss what you’d like to change.

---

📄 License

This project is licensed under the Apache 2.0 License.

---