Practical, production-ready Retrieval-Augmented Generation (RAG) examples built using Haystack, IBM Db2 Vector Store, IBM Granite Embedding Models, IBM Granite LLMs, and Ollama.
This repository demonstrates how to build enterprise Retrieval-Augmented Generation (RAG) applications using the official IBM Db2 integration for Haystack.
Each example is intentionally designed to be self-contained and focuses on a real-world enterprise use case. Every application shows how IBM Db2 can be used as a native vector database for semantic search while Haystack orchestrates the complete RAG pipeline.
The examples use:
- IBM Db2 Vector Store
- Haystack Pipelines
- IBM Granite Embedding Models
- IBM Granite LLMs
- Ollama
- Native Vector Similarity Search
Whether you're learning Haystack, evaluating IBM Db2 Vector Search, or building enterprise AI assistants, these practical examples provide a solid starting point.
| Example | Description |
|---|---|
| 01 Banking & Financial Assistant | Answers customer questions using banking policies, FAQs, product documentation, and internal knowledge bases. |
| 02 Healthcare Clinical Knowledge Assistant | Retrieves clinical guidelines and healthcare documentation while generating safe, grounded responses. |
| 03 Legal Document Assistant | Performs semantic search across legal contracts, agreements, policies, and compliance documentation. |
| 04 Insurance Claims Assistant | Retrieves insurance policy coverage, exclusions, claim conditions, and benefits. |
| 05 HR Policy Assistant | Answers employee questions from HR policies, employee handbooks, leave policies, and internal documentation. |
| 06 Enterprise IT Support Assistant | Provides troubleshooting assistance using enterprise IT knowledge base articles and internal documentation. |
haystack-db2-practical-usecases/
│── cover.png
│── README.md
│
├── 01_banking_assistant.py
├── 02_healthcare_assistant.py
├── 03_legal_assistant.py
├── 04_insurance_assistant.py
├── 05_hr_assistant.py
└── 06_it_support_assistant.py
Every example follows the same Retrieval-Augmented Generation workflow.
flowchart LR
U([User Question])
E["IBM Granite<br/>Embedding Model"]
DB[("IBM Db2<br/>Vector Store")]
R["IBMDb2<br/>Embedding Retriever"]
P["Prompt Builder"]
L["IBM Granite<br/>LLM"]
A([Grounded Response])
U --> E
E --> R
R --> DB
DB --> R
R --> P
P --> L
L --> A
style DB fill:#0F62FE,color:#ffffff,stroke:#0F62FE,stroke-width:3px
style L fill:#6F42C1,color:#ffffff
style E fill:#24A148,color:#ffffff
style A fill:#198038,color:#ffffff
| Component | Technology |
|---|---|
| AI Framework | Haystack |
| Vector Database | IBM Db2 |
| Embedding Model | IBM Granite Embedding 278M |
| Language Model | IBM Granite 3.3 |
| Local Inference | Ollama |
| Programming Language | Python |
Before running these examples, ensure you have:
- Python 3.11+
- IBM Db2 with Vector Search enabled
- Ollama installed locally
- IBM Granite Embedding Model
- IBM Granite 3.3 Model
Clone the repository.
git clone https://github.com/<your-username>/haystack-db2-practical-usecases.git
cd haystack-db2-practical-usecasesInstall the dependencies.
pip install -r requirements.txtollama pull granite-embedding:278m
ollama pull granite3.3:8bSet your Db2 credentials before running any example.
export DB2_USERNAME=db2inst1
export DB2_PASSWORD=password
export DB2_HOST=localhost
export OLLAMA_URL=http://localhost:11434Run any assistant directly.
python 01_banking_assistant.pyor
python 02_healthcare_assistant.pyor
python 03_legal_assistant.pyor
python 04_insurance_assistant.pyor
python 05_hr_assistant.pyor
python 06_it_support_assistant.pyThis repository demonstrates how to:
- Build Retrieval-Augmented Generation (RAG) pipelines with Haystack
- Store vector embeddings in IBM Db2
- Perform semantic similarity search using IBM Db2 Vector Store
- Generate embeddings with IBM Granite Embedding Models
- Generate grounded responses with IBM Granite LLMs
- Build production-ready enterprise AI assistants
- Develop modular Haystack pipelines that are easy to customize
Build a customer-facing banking assistant capable of answering questions from policies, product documentation, and FAQs using semantic retrieval.
Retrieve clinical guidelines and medical documentation while ensuring responses remain grounded and safe.
Search legal contracts, agreements, compliance documentation, and internal policies using semantic search.
Answer questions related to insurance coverage, claim policies, exclusions, and policy conditions.
Provide employees with accurate answers from company handbooks, leave policies, onboarding guides, and HR documentation.
Retrieve internal knowledge base articles and provide step-by-step troubleshooting instructions for enterprise IT support.
- Official IBM Db2 Integration for Haystack
- IBM Db2 Native Vector Store
- IBM Granite Embedding Models
- IBM Granite Large Language Models
- Local inference with Ollama
- Production-ready RAG Pipelines
- Enterprise AI Assistant examples
- Easy to customize for your own documents and knowledge bases
Want to learn more about the official IBM Db2 integration for Haystack? Explore the announcement, technical guide, and complete walkthrough below.
Build Grounded AI Applications with the new IBM Db2 Integration for Haystack
IBM Db2 Database
https://www.ibm.com/products/db2-database
Haystack Documentation
https://docs.haystack.deepset.ai/
IBM Granite Models
Ollama
Contributions are always welcome.
If you'd like to improve an example, add a new enterprise use case, or enhance the IBM Db2 + Haystack ecosystem, feel free to open an Issue or submit a Pull Request.
Built using:
- Haystack by deepset
- IBM Db2
- IBM Db2 Vector Store Integration
- IBM Granite Models
- Ollama
- Python
Special thanks to everyone contributing to the Haystack and IBM Db2 open-source ecosystem.
If you found this repository helpful, please consider giving it a Star on GitHub. It helps others discover the project and supports continued development of enterprise AI examples using Haystack and IBM Db2.