RelevAI Python SDK

Helper SDK to simplify access and integration with RelevAI's AI services.

This library offers a lightweight, developer-friendly interface for authenticating and interacting with RelevAI’s API endpoints.
It includes automatic token management (ClientKey / ServiceKey), sync/async client support, and flexible serialization strategies.

While not required to use RelevAI’s APIs, this SDK is designed to reduce boilerplate and speed up development.

---

Features

• Secure authentication with RelevAI services using OAuth2 tokens
• Automatic token renewal in the background
• Synchronous and asynchronous clients
• Built-in support for RelevAI AI-Lang services (chat, embed)
• Pluggable serialization system (JSON, LZ4, MsgPack)

---

Installation

pip install relevai

Requires Python 3.9+

---

Authentication

You can authenticate with RelevAI using either a ClientKey (API key) or ServiceKey (client credentials).

Example using ClientKey:

from relevai.key import ClientKey
from relevai.ai import AILangClient

key = ClientKey(
    api_key="your_refresh_token",
    client_id="your_client_id",
    auth_url="https://auth.relev.ai/token"
)

client = AILangClient(key)
response = client.chat(model="llm-model", messages=[{"role": "user", "content": "Hello!"}])

---

Async Client

For asynchronous use:

import asyncio
from relevai.key import ClientKey
from relevai.ai import AILangAsyncClient

async def main():
    key = ClientKey(
        api_key="your_refresh_token",
        client_id="your_client_id",
        auth_url="https://auth.relev.ai/token"
    )

    client = AILangAsyncClient(key)
    response = await client.chat(model="relevai/ai-lang", messages=[{"role": "user", "content": "Hi!"}])

asyncio.run(main())

---

Serialization

This SDK includes a simple serialization interface with multiple backends:

• JSONSerializer (default)
• LZ4Serializer (requires joblib)
• MsgPackSerializer (requires msgpack)

Each serializer supports:

• dumps() / loads() — base64 string encoding
• dump() / load() — raw bytes

Example:

from relevai.serializers import JSONSerializer

serializer = JSONSerializer()
data = {"key": "value"}

encoded = await serializer.dumps(data)
decoded = await serializer.loads(encoded)

---

License

© 2025 RelevAI S.L. — Licensed under the Apache License 2.0

---

Links

• RelevAI official webpage