Six authentication errors between your first curl command and a working SAP AI Core bearer token — PowerShell alias conflicts, curly quote corruption, single quote failures, silent empty responses, missing /oauth/token path, and RBAC denial — with th...
An LLM that only generates text is like a consultant who can advise but never execute. In this blog you give it hands, tools to fetch data, structured output to return typed results your code can use, memory to hold a conversation, and streaming for ...
Microsoft just leveled up VS Code and GitHub with real AI horsepower. So I explored how this could boost prompt generation in Fiori projects—particularly when Joule isn’t available.
Sales Standard Operating Procedures (SOPs) are goldmines of organizational knowledge — they define who does what, which products are sold, how deals are closed, and which tools the team uses. But this knowledge is buried in unstructured prose, spread...
Can AI agents write production-ready ABAP code? Most developers experiment with AI assistants for modern languages, but ABAP remains largely unexplored territory. When I decided to use Claude Code to generate a real feature from my production backlog...
In this blog, we set up a production-grade Python project from scratch using uv and hatchling, securely load SAP AI Core credentials with pydantic-settings, and build an app that lets you chat with any LLM deployed on SAP AI Core. This is Blog 1 of N...
The final part of this series surfaces the AI stack from Parts 1–5 through Joule — building custom Joule skills backed by the CAP service from Part 5, assembling a multi-skill agent for knowledge base Q&A and exception log triage, covering prerequisi...
This blog walks through how to deploy a custom AI model to SAP AI Core, register it as a callable service via SAP AI Launchpad and consume it directly from a CAP backend
We have officially moved beyond the era of the simple "chatbot." While a single prompt to a Large Language Model (LLM) can produce impressive results, developers building production-grade enterprise applications quickly realize that single-inference ...
Hi Experts/Gurus,I have recently enabled Joule Copilot specifically joule base within S/4HANA Private Cloud edition 2023 version. It works fine for informational & navigational scenarios. However, when i am testing it for transactional capabilities l...
End-to-end architecture for AI-powered SAP extension applications using CAP — project setup, AI Core service binding, HANA Cloud persistence, Generative AI Hub integration in service handlers, SAP Integration Suite connectivity, and Cloud Foundry dep...
If your langchain RAG pipeline agent works, users are happy but the budget team is not sharing that happiness, this package may be of help.langchain-hana-cache stores LLM responses in HANA Cloud and returns them instantly for similar prompts using CO...
Without any carefully crafted skills or MCPs, Claude Code exceeded my expectations and started analyzing ABAP internals, I have no idea about, to port a slightly outdated, abapGit managed, git repository to new APIs.
End-to-end RAG pipeline on SAP data — REAL_VECTOR schema setup, SAP embedding model via AI Core, document chunking, COSINE_SIMILARITY search, and four working use cases: knowledge base search, data dictionary retrieval, transport log search, and mast...
In November 2025, I started using Claude Code differently. Instead of just throwing prompts at it and hoping for the best, I began building skills — little workflow playbooks written in markdown that tell the AI how to approach a problem, not just wh...
SAP RPT-1 · The next big thing in enterprise AIYour data is smarter than you think.RPT-1 just proved it.Six reasons this changes everything.What if your ERP could predict outcomes without a single data scientist — just by looking at a few example row...
Four multi-step prompt pipelines for real SAP workloads — exception log triage, HANA code review, document processing, and master data enrichment — using sequential chaining and chain-of-thought reasoning with the generative-ai-hub-sdk and LangChain.
Taming the Dependency Update Avalanche If you've ever managed dependency updates across multiple repositories, you know the pain. Every morning, there's a fresh batch of version bumps waiting for you. Click into GitHub, check the build status, review...
AI code generation helped the hana-cli project add more features in two months than I'd been able to deliver in the last two years; but quality still came from discipline, architecture, and a ton of testing.
Five working Python code examples for real SAP workloads — exception log summarization, HANA query analysis, ABAP/SQL code generation, master data classification, and structured data extraction — using the generative-ai-hub-sdk across SAP AI Launchpa...
Step-by-step configuration from BTP entitlements to your first foundation model API call — including the missing AI Launchpad entitlement, zero default quota, resource group API setup, OAuth authentication pattern, and regional model availability gap...
Because brain in a jar is so 2024.
Guess who else speaks MCP: Codex does.
On one rainy morning I mentioned to my Google Antigravity Agent: "I hear people are experimenting with connecting to SAP Systems via MCP. They use ABAP Development Tools API as a way in. Could you help me set up something like that for my ABAP Trial ...
Agentic memory enables AI to remember past interactions, but its real value lies in knowing precisely which information to store and when. This article defines the concept of 'agentic memory,' outlines key use cases such as personalization and agenti...
Your RAG pipeline retrieves chunks. But can it connect the dots between them? This blog walks through building a Knowledge Graph on SAP HANA Triple Store and combining it with vector search, with full working code using SAP AI Core, HANA ML, and Fast...
BM25 keyword ranking and hybrid retrieval for RAG pipelines on SAP HANA Cloud. Combine vector search with keyword matching using LangChain-compatible retrievers.
RAG (Retrieval-Augmented Generation) has quickly become the backbone of modern enterprise AI. In the SAP ecosystem—where business processes rely on massive, complex, and varied documents—RAG isn't just a buzzword. It's a critical layer of contextual ...
LangGraph checkpointer for SAP HANA Cloud — persist AI agent state where your data already lives.
IntroductionI’m an SAP functional consultant by role, but I’ve always had the soul of a developer. This is the story of how that instinct — along with a lot of late nights, AI assistance, and a fair share of dead ends — led me to build something I ne...
Hello,I have build the smallest OData MCP bridge server (in the world )!You can find the source code (publicly available) here:https://github.com/danmcool/odata_mcpsrv_xsDISCLAIMER: current authentication method used is only for test purposes, for a ...
There is a moment in every AI project when the model finally starts working.Not perfectly, but well enough that people stop asking whether it works and start asking something far more dangerous:Can we show this to customers?And more importantly, can ...
When a client opens a ticket for SAP software support into AMS services, time to resolve remain a relevant KPI. S.P.A.R.K. (SAP Process Augmented Response Knowledge) is a Retrieval‑Augmented Generation (RAG) platform built entirely on SAP technology...
In my recent Hello World with SAP’s Harmonized API, we discovered how SAP’s Orchestration Service lets us talk to models from OpenAI, Anthropic, Google, and others through a single unified interface. The killer feature: swap out a model name string, ...
SAP Joule for Consultants provides SAP-grounded, AI-driven guidance to accelerate SAP delivery, ensure clean-core alignment, and scale professional expertise efficiently.
I'll walk you through the steps to build a "Hello, World" agent. Along the way, you'll get familiar with the basic workflow in SAP Build, explore the UI, avoid at least one common pitfall, and have your agent up and running in just a few minutes. By ...
In this post, I show how to build a private, local, AI-powered ABAP development environment using VS Code, Ollama, and open-source LLMs — no BTP, no cloud, and no code leaving your machine.It’s about privacy, performance, and putting AI directly in t...
AI agents don’t fail like traditional software and that changes everything. When identical inputs can produce different decisions, traditional logs and breakpoints can no longer explain what went wrong. In the agent era, understanding why something h...
A step-by-step tutorial on calling SAP GenAI Hub's large language models from self-hosted n8n to build enterprise AI automation workflows.SAP Build is the primary, recommended solution for enterprise automation and orchestration within the SAP ecosys...
JSON is the default input format for LLM prompts, mostly out of popularity.In this experiment, I compared JSON and TOON using the same data, the same tasks, and measurable metrics: accuracy, token count, and latency.The takeaway isn’t “replace JSON,”...
Stuck behind corporate firewalls without access to the latest AI tools? I was too. In my latest post, I share how to bridge the gap using Model Context Protocol (MCP).I show you how to build and configure two powerful open-source tools—vibing-steampu...
The Harmonized API of the Orchestration Service is probably one of SAP's best kept secrets. It allows you to talk to all the language models on AI core in a harmonized format across model families. This means that you can simply swap out the model na...
I hope everyone found their way into the new year and had some peaceful time over the holidays! Being back at work, I started to try out all the cool announcements from last year. One of them is SAP contributing to the third-party open-source Python ...
SAP Joule for Developers, ABAP AI Capabilities is your intelligent coding companion, bringing deep expertise in ABAP Cloud and AI-driven development. When we talk about ABAP AI use cases, they revolve around three key pillars — Accelerate, Empower, a...
SAP’s All‑in AI Vision: A Bold and Timely MoveSAP’s All‑in AI strategy is not just ambitious—it’s absolutely the right direction. As an Enterprise Architect, I’m genuinely excited by how SAP’s AI vision is evolving from being a text generator to bec...
In traditional web development, we design for the "average" user. We build fixed dashboards, hard-coded forms, and rigid menus. We try to predict every possible path a user might take. The result? Friction:The Overload: A power user and a beginner se...
How do you test SAP's RPT-1 model beyond a simple Hello World example? Its claim to fame is that it can replace traditional machine learning models for classification and regression tasks. No model training, no hyperparameter tuning, just show it you...
MCP scales agent capabilities, but only a centralized control plane makes it enterprise-safe. This post outlines the architecture required to enforce identity, authorization, auditability, and containment across MCP-based systems.
MCP is quickly becoming the USB port for AI agents, but in the enterprise, it’s also a security nightmare waiting to happen.As teams rush from demos to production, direct MCP connections are quietly creating blind spots in attribution, access control...
As enterprise GenAI moves from proof-of-concept to production, standard Vector RAG is hitting a ceiling, often struggling with complex reasoning and the rich structure of business documents. In my latest blog post, I explore the next evolution: Hybri...
| User | Count |
|---|---|
| 28 | |
| 12 | |
| 9 | |
| 9 | |
| 9 | |
| 9 | |
| 9 | |
| 9 | |
| 7 | |
| 7 |
emf
:
I Passed the SAP Joule Studio Developer Exam. Here...
ChristophMorgen
:
Zero-shot time series forecasting with SAP-RPT-TS ...
ulka_rathi359f4
denismih
:
CAP Agents that collaborates with humans - Announc...
| Subject | Likes |
|---|---|
| 21 | |
| 9 | |
| 9 | |
| 9 | |
| 7 | |
| 7 | |
| 5 | |
| 5 |
