Christophe Crémon

Today at IBM TechXchange, I delivered my session 𝐃𝐛𝟐 𝐀𝐈 𝐕𝐞𝐜𝐭𝐨𝐫𝐬: 𝐏𝐨𝐰𝐞𝐫𝐢𝐧𝐠 𝐌𝐨𝐝𝐞𝐫𝐧 𝐀𝐈 𝐰𝐢𝐭𝐡 𝐃𝐛𝟐 I presented how Db2 is evolving for the AI era — where relational data and semantic intelligence meet inside one database. 𝐊𝐞𝐲 𝐭𝐚𝐤𝐞𝐚𝐰𝐚𝐲𝐬 𝐟𝐫𝐨𝐦 𝐭𝐡𝐞 𝐭𝐚𝐥𝐤: → Vectors represent meaning — king is closer to queen than to cat — enabling similarity search based on semantics, not keywords. → Db2 introduced a native VECTOR type (FLOAT32 or INT8) starting with Db2 12 Mod Pack 2, allowing embeddings to be stored alongside relational data. → Built-in SQL functions like VECTOR_DISTANCE() compute similarity using COSINE, EUCLIDEAN, DOT, and other supported metrics. → Developers can combine relational filters (e.g., brand, category, region) and semantic filters/ranking in a single SQL query. → Db2's LangChain and LlamaIndex integrations let Db2 serve as a Vector Store for RAG and other LLM-powered application workflows. → SQL-native embedding generation (early preview) enables Db2 to call embedding models such as watsonx.ai or OpenAI-compatible APIs directly from SQL. → Vector Index (early preview) introduces graph-based approximate nearest neighbor (ANN) search for faster similarity queries at larger scale. → These capabilities bring AI workloads closer to enterprise data — enabling applications that search, match, and reason by meaning as well as structure. 𝐄𝐱𝐩𝐥𝐨𝐫𝐞 𝐦𝐨𝐫𝐞 𝐫𝐞𝐬𝐨𝐮𝐫𝐜𝐞𝐬 𝐚𝐧𝐝 𝐞𝐱𝐚𝐦𝐩𝐥𝐞𝐬: ibm.biz/learndb2ai #ibm #db2 #ai #IBMTechXChange2025

32

5 commentaires

🚀 Claude Sonnet 4.6 is Now Available in Azure AI Foundry — Here’s My 4-Step Upgrade Framework As enterprises move from AI experimentation to production-scale systems, model selection becomes an architectural decision — not just a technical one. Claude Sonnet 4.6, now available in Azure AI Foundry, delivers frontier-level reasoning, adaptive thinking, expanded agentic capabilities, and up to a 1M token context window (beta). I’ve put together a practical 4-step framework to help teams migrate seamlessly: 1️⃣ Technical Baseline & Breaking Changes Before swapping model IDs, treat this as an API contract upgrade: ✅ Update deployment target to claude-sonnet-4-6 ✅ Remove assistant prefilling (transition to structured outputs / output_config.format) ✅ Provide either temperature OR top_p (not both) ✅ Update deprecated tool versions (e.g., text_editor_20250728) if migrating from 3.5 Small oversights here can cause silent failures in production. 2️⃣ Reasoning & Performance Tuning Sonnet 4.6 introduces adaptive thinking, reducing the need for manual chain-of-thought prompting. 🧠 Enable adaptive thinking for dynamic reasoning depth 🎯 Use budget_tokens for high-stakes reasoning control ✂️ Remove long CoT scaffolding — native extended reasoning is more efficient This improves cost governance, latency, and output reliability at scale. 3️⃣ Scaling the Context Window With up to 1M tokens (beta): 📄 Consolidate entire codebases or long legal/financial documents 🔁 Reduce complex RAG chunking for many workloads ⚡ Implement prompt caching in Azure Foundry to control cost and latency This is a major architectural simplification opportunity. 4️⃣ Agentic & Computer Use Validation If you're building AI agents or automation workflows: 🖥️ Sonnet 4.6 shows stronger computer-use performance (OSWorld benchmarks) 🔄 Re-test multi-step UI workflows and legacy system automations ⚡ Prepare backend services for parallel tool execution (not just sequential calls) Your orchestration layer must now handle concurrency properly. 📌 Bottom line: Migrating to Claude Sonnet 4.6 in Azure AI Foundry is not just a model upgrade — it’s a strategic shift toward scalable, governed, production-grade AI architecture. Excited to help teams design secure, enterprise-ready AI systems on Azure. #Claude #ClaudeAI #ClaudeSonnet #ClaudeSonnet46 #Anthropic #AzureAI #MicrosoftAIFoundry #EnterpriseAI #AIArchitecture #GenerativeAI #CloudTransformation

8

While working on the “Building with the Claude API” course from Anthropic. I created a jupyter notebook on Anitgravity ran the following code : message = client.messages.create( model=model, max_tokens=1000, system="You are a helpful and precise assistant.", messages=[ { "role": "user", "content": "What is quantum computing? Answer in one sentence" } ] ) And I got the following “Insufficient Balance” error as below: APIStatusError: Error code: 402 - {'error': {'message': 'Insufficient Balance', Troubleshooting steps: Used curl cli to verify my Athropic API (around $25 fund in my Anthropic account) that have sufficient funds in it. curl https://lnkd.in/eYVfdsi9 \ --header "x-api-key: YOUR_API_KEY_HERE" \ --header "anthropic-version: 2023-06-01" \ --header "content-type: application/json" \ --data '{ "model": "claude-haiku-4-5-20251001", "max_tokens": 10, "messages": [{"role": "user", "content": "Respond with the word OK."}] }' It returned “OK” as response message below: {"model":"claude-haiku-4-5-20251001","id":"msg_XXX","type":"message","role":"assistant","content":[{"type":"text","text":"OK"}],... This means that my API key is active and there are sufficient fund in my Anthropic account. Consulted “Gemini” in “Antigravity” with the following prompt : How to troubleshoot load_dotenv() where using curl cli is able to call claude api successfully but getting error 'insufficient fund" when running the python code in jupyther notebook? Root cause is there is a global environment variable set: ANTHROPIC_BASE_URL=https://lnkd.in/gaeA3ysC which was set previous for my previous learning/testing that causing this error. The resolution is to explicitly specify “ANTHROPIC_BASE_URL” in the code to point it to “https://api.anthropic.com”” as updated code below: import os from dotenv import load_dotenv import anthropic # 1. Load the environment variables from the .env file load_dotenv() # 2. Explicitly bypass the system ANTHROPIC_BASE_URL environment variable client = anthropic.Anthropic(base_url="https://api.anthropic.com") # 3. Specify the model you verified with curl model = "claude-haiku-4-5-20251001" # 4. Send the message request message = client.messages.create( model=model, max_tokens=500, messages=[ { "role": "user", "content": "What is quantum computing?" } ] ) # Print the response text print(message.content[0].text) And the code above successfully returned the response from Anthropic : Quantum computing uses quantum bits (qubits) that can exist in multiple states simultaneously, allowing quantum computers to process certain types of problems much faster than classical computers. Please let me if you also facing similar error and how did you resolve it. https://lnkd.in/eQmaVawh

4

🔎 Is your data architecture holding back BI and AI ROI? Rigid, batch-centric architectures struggle with flexibility, real-time needs, and scale — and that’s exactly what Active Data Architecture™ fixes. It sits between physical data stores and points of consumption, enabling virtualized, distributed access, governance, and security so organizations can deliver timely, trusted insights. We’re proud that Denodo was named a top Active Data Architecture vendor in the 2025 Dresner Advisory Services, LLC Active Data Architecture® Report. The study maps market requirements for data orchestration, integration, and #analytics — and explains why organizations are moving from bulk #ETL to real-time and virtualization-based approaches. What you’ll gain from the report: • Why legacy ETL limits #ROI and agility. • How #ActiveDataArchitecture supports real-time insights and governed access. • Market priorities, competitive signals, and practical guidance for BI producers & consumers. 📥 Download the full 2025 Active Data Architecture Report (Dresner Wisdom of Crowds®) to see how leading vendors compare and how your architecture can evolve into an engine for #BI & #AI success 👉 https://okt.to/ck9ChR #DataVirtualization #DataArchitecture #Dresner #DataForALL #NoDataLeftBehind #NoAIWithoutData

12

[Tech Community] Spotlight On... AI Transformation & the Frontier Firm Playbook. Introduction  AI transformation is not a single launch – it’s a multi‑phase journey that blends strategy, change management, and... #azure #techcommunity https://lnkd.in/epUNinsx

1

Greengage Database (based on Greenplum®, open source alternative)(https://lnkd.in/ehaCaYHt) is now available right inside DBeaver. Starting with the latest release on October 6, you can connect to and work with it directly from the DBeaver interface - just like you do with PostgreSQL, Greenplum, or any other database.

27

🔥 A must-watch for anyone shaping the future of cybersecurity… Sopra Steria Ventures is proud to share an exclusive cross-perspective interview between Jean-Luc Gibernon (Sopra Steria) and 🛡️ Yassir KAZAR (Yogosha). You’ll hear about: ✅ How Sopra Steria and Yogosha create joint value for their customers ✅ Why cybersecurity is gaining crucial momentum in companies’ roadmaps ✅ What lies ahead in terms of top priorities for the future of European cybersecurity 🎥 Watch the full interview: https://lnkd.in/ec-NTKr6

123

4 commentaires