Agentic Examples for SAP Generative AI Hub

The examples showcase how to easily use different agentic frameworks with the "SAP provider" via LiteLLM.

SAP customers can access LLMs across all vendors (OpenAI, Google, Amazon, Mistral, SAP, ...) available in the SAP Generative AI Hub.

📚 View Online documentation 📚

🎥 Watch the Devtoberfest Talk on LiteLLM & SAP GenAI Hub 🎥

Sap Generative AI Hub

The SAP Generative AI Hub enables customers to access large language models (LLMs) from various providers e.g., Gemini, ChatGPT, Claude, Mistral in a centralized manner.

LiteLMM

LiteLLM is an open-source library which supports 100+ LLMs from various providers. LiteLLM acts as the bridge to use standard open-source frameworks via the OpenAI api. LiteLLM integrates with various LLM providers. The "SAP provider" is recently added to LiteLLM to support the SAP Generative AI Hub.

Prerequisites

• SAP AI Core Gen AI subscription via SAP BTP tenant.
• Install latest Litellm (includes SAP provider).

Examples Overview

The following examples include python code and jupyter notebooks to demonstrate the integration step by step.

1. Langgraph

• langgraph_agent notebook

2. CrewAI

• crewai_litellm_lib notebook
• crewai_litellm_proxy notebook

3. PydanticAI

• pydanticai_litellm_proxy notebook

4. Google ADK

• google_adk notebook

5. OpenAI ADK

• openai_adk notebook

6. AWS

• aws_strands notebook

7. LlamaIndex

• Llamaindex_litellm notebook

8. smolagents

• smolagents_litellm notebook

9. Microsoft Agent Framework

• microsoft_agent_litellm_proxy notebook

10. Agentscope

• agentscope_litellm notebook

11. AG2

• ag2_litellm notebook

Note for Maintainers of these examples

Cleanup Notebooks before Commit

Clear cell outputs and metadata using the ".pre-commit-config.yaml". Installation procedure:

python3 -m venv env
source env/bin/activate
python3 -m pip install pre-commit nbstripout

Manual run:

pre-commit run --all-files

Skip Hooks temporarily:

git commit -m "Message" --no-verify

Update Documentation via Sphinx

The documentation is automatically built and deployed via GitHub Actions on each push to the main branch. To build the documentation locally, follow the instructions in DOCUMENTATION_SETUP.md.