Skip to content

Welcome to Vector Search Hands-on

In this hands-on workshop, you will combine Building Blocks and IBM Bob to experience AI-driven development for building a "semantic search" feature (Vector Search).

Prerequisites

IBM Bob is already installed and available for use. This hands-on uses IBM Bob 1.0.3.

What You'll Experience in This Hands-on

Value of Building Blocks + IBM Bob

This hands-on workshop demonstrates how combining Building Blocks (pre-built technical components) with IBM Bob (an AI development assistant) can complete development that would typically take days to weeks in approximately 90 minutes.

Without Building Blocks (Time required: days to weeks):

Development flow without Building Blocks

Without Building Blocks, the following work is required:

  • Vector database selection and learning
  • Embedding model selection and integration
  • API design and implementation
  • Error handling
  • Performance tuning

With Building Blocks + IBM Bob (This hands-on, Time required: approximately 90 minutes):

Development flow with Building Blocks + IBM Bob

Responsibilities for each process:

  • Building Blocks:
    • Technology selection (Milvus, embedding models)
    • Environment setup support (Bob mode, API samples)
  • IBM Bob:
    • Requirements definition
    • Coding
    • Testing
    • Debugging
About IBM Bob's Coverage

IBM Bob can support the entire Software Development Lifecycle (SDLC) as an AI SDLC partner, from requirements definition to debugging. In this hands-on, Building Blocks provides technology selection (Milvus, embedding models) and environment setup support (Milvus setup, Bob mode), and the instructor prepares the Milvus environment in advance with Docker Compose, so IBM Bob focuses mainly on coding, testing, and debugging. However, if you use Plan mode, you can also utilize it in the requirements definition and design stages.

What are Building Blocks?

Building Blocks are pre-built technical components leveraging IBM's technology stack. Using Building Blocks accelerates solution development.

Features of Building Blocks

  • Ready to use: Start using immediately without complex configuration or learning
  • Best practices: Optimal implementation patterns designed by IBM's engineering team
  • Domain-specific: Provides Vector Search-specific guidance and implementation patterns
  • Customizable: Flexibly extend to meet business requirements using IBM Bob

Building Block Used in This Hands-on

Vector Search Builder (Milvus-based)

What it provides: Vector database (Milvus) construction and management capabilities

Included features:

  • Milvus database setup
  • Collection (data container) creation
  • Local embedding model integration with Hugging Face Transformers
  • Sample product data ingestion workflow
  • Vector search optimization

Integration with IBM Bob: Using Vector Search Builder mode, IBM Bob provides specialized support for Vector Search

Value of Building Blocks

Without Building Blocks: Read Milvus documentation, learn Python SDK, select and integrate embedding models (days)

With Building Blocks: Install Vector Search Builder and instruct IBM Bob (minutes)

Unique Innovations in This Hands-on

File and directory paths in this section are relative to the following GitHub repository.

What Building Blocks Provide

Building Blocks provide the following technical components:

  • Vector Search Builder Mode
    • Participant package: vector-search-builder-en.zip
    • Contents:
      • IBM Bob custom mode configuration
      • 3 Vector Search Builder rule files
      • AI assistant functionality specialized for Vector Search
      • Milvus operation best practices
      • Participant scripts and connection configuration template
    • Excluded:
      • Instructor files
      • Documentation files
      • Local .env files and generated caches

What This Hands-on Adds

In addition to the Building Blocks foundation, the following have been added for educational purposes:

  • setup/instructor/: Instructor Milvus environment (Docker Compose)
  • setup/participant/: Participant connection test scripts
  • docs/: Hands-on documentation (MkDocs)

1. Instructor-Participant Separation Architecture

Building Blocks alone:

  • Each person builds their own Milvus environment (Docker/Podman/Colima)
  • Individually download embedding models (approximately 460 MB)
  • Environment setup takes about 30 minutes

This hands-on's innovation:

  • Instructor: Centrally manages Milvus environment (setup/instructor/docker-compose.yml)
  • Participants: Participate with IBM Bob, .bob/custom_modes.yaml, .bob/rules-vector-search-builder/, participant scripts, and connection information only

2. Hybrid Delivery Support

Building Blocks alone:

  • Assumes local environment execution

This hands-on's innovation:

  • On-site: Local network sharing (http://instructor IP:8001)
  • Remote: Document delivery via GitHub Pages or ngrok

3. API Key-Free Design

Building Blocks alone:

  • Cloud-based embedding options often require API keys
  • Participants configure credentials individually

This hands-on's innovation:

  • Hugging Face Transformers used (no API key required)
  • Local execution: Works with internet connection only

4. Progressive Learning Path

Building Blocks alone:

  • Focuses on technical implementation

This hands-on's innovation:

  • Part 1: Experience Vector Search (understanding)
  • Part 2: Add features with IBM Bob (practice)
  • Part 3: Code review and improvement (application)

Summary of Role Division

Provider What's Provided Purpose
Building Blocks Vector Search Builder mode
FastAPI sample
Milvus setup guide		 Technology foundation provision
Development acceleration
This Hands-on Instructor environment (Docker Compose)
Participant scripts
Educational documentation		 Educational design
Learning experience optimization

Benefits of This Hands-on

Building Blocks (technology foundation) + Hands-on unique innovations (educational design) = High learning effectiveness in a short time

  • Setup time reduction: 30 minutes → 5 minutes (instructor centrally manages environment)
  • No API key required: Using Hugging Face reduces participant preparation burden
  • Flexible delivery format: Supports on-site/remote/hybrid delivery
  • Progressive learning: Even beginners can progress from understanding → practice → application

What is IBM Bob?

IBM Bob is a development tool where AI assists with coding.

What IBM Bob Can Do

  • Natural language instructions: Communicate what you want to do in words
  • Automatic code generation: Automatically writes high-quality code
  • Code review: Points out code issues
  • Integration with Building Blocks: Provides technology-specific support through custom modes

Synergy with Building Blocks

Building Blocks alone:

  • Basic functionality is provided, but customization requires technical knowledge

IBM Bob alone:

  • Code generation is possible, but building from scratch takes time

Building Blocks + IBM Bob:

  • Building Blocks instantly builds the foundation
  • IBM Bob customizes with natural language instructions only
  • Result: Achieve production-level quality in the shortest time

Comparison of Development Methods

Development Method Time Required Required Skills Code Quality
Without Building Blocks Days to weeks Programming, DB design, API design Depends on developer skills
IBM Bob only Hours to days Basic technical understanding High quality but time-consuming to build
Building Blocks + IBM Bob Minutes to hours Just need to instruct in natural language Production-level high quality

Vector Search is a technology that searches by understanding the "meaning" of words.

Traditional keyword search:

  • "red sneakers" → Searches for products containing the characters "red" and "sneakers"
  • "red running shoes" won't be found (different characters)

Vector Search (semantic search):

  • "red sneakers" → Understands the meaning of "red" and "sneakers"
  • "red running shoes" will be found (similar meaning)
  • "beginner camera" → "entry-level digital camera" will be found

Real-world Use Cases

  • E-commerce sites: "Find similar products" feature
  • Internal search: "Find documents similar to this document"
  • Customer support: "Find similar questions"

Hands-on Flow

Total: Approximately 90 minutes

Part Content Time Required
Preparation Vector Search Builder setup 15 minutes
Part 1 Experience Vector Search 20 minutes
Part 2 Add features with IBM Bob 30 minutes
Part 3 Verification 15 minutes
Summary Review and Q&A 10 minutes
About This Hands-on's Documentation Design

Why Manual Methods Differ Between First and Second Half

In this hands-on, the first half (preparation, Part 1) describes both IBM Bob delegation and manual execution methods, but the second half (Part 2-3) describes only IBM Bob delegation methods. This is for the following reasons:

1. Complexity and Length of Manual Work

  • First half work: Simple command execution (pip install -r requirements.txt, python test_connection.py), can be completed in one line manually
  • Second half work: Editing multiple files such as app.py, schema.py, data insertion scripts, and sample product data; changing data models, response structures, error handling, etc.; requiring dozens to hundreds of lines of code changes. Manual description would be very long and complex, making the documentation enormous

2. Educational Intent

  • First half: Show options that "can be done with IBM Bob or manually"
  • Second half: Let users experience the value that "what's difficult manually is easy with IBM Bob"
Experience the Value of Building Blocks + IBM Bob

In particular, the experience of completing complex code changes with a short instruction to add an image_url field to the /search API JSON response is designed to most effectively convey the value of Building Blocks + IBM Bob.

Why This Instruction is Most Effective:

Building Blocks Effect:

  • Vector Search knowledge: IBM Bob understands Milvus, embedding models, and vector search best practices through Vector Search Builder mode
  • Existing foundation: Sample data, API structure, shared schema definitions, and data models are already prepared, and IBM Bob can add features using them
  • No technology selection needed: Technology selection for Milvus, embedding models, API design, etc. is complete, and IBM Bob can focus on implementation

IBM Bob Effect:

  • Natural language instructions: Just one line in natural language, without any technical details
  • Automatic code generation: Automatically executes editing of multiple files, schema/data model changes, response structure changes
  • Immediate results: Can verify operation immediately after instruction, getting the feeling that "it really worked"

Synergy of IBM Bob and Building Blocks:

  • First experience in Part 2: The moment participants "add a feature themselves" for the first time, making it memorable
  • Contrast with other instructions: Price filters and recommendation reasons are similarly easy, but this first experience is most impactful
  • Gap with complexity: Work that would take days without Building Blocks is completed with one IBM Bob instruction

3. Building Blocks Value Proposition

  • Let users experience the time reduction effect of "days to weeks → approximately 90 minutes"
  • Emphasize this effect by omitting manual methods in the second half

4. Consideration for Time Constraints

  • Designed for approximately 90 minutes total
  • Just reading detailed manual methods would run out of time
  • Focusing on IBM Bob delegation secures time for actual hands-on work

5. Complexity of Error Handling

When manually changing code, troubleshooting for syntax errors, indentation errors, type errors, logic errors, etc. is necessary. Describing all of these would make the documentation several times longer

6. Progressive Learning Design

  • First half: Get familiar with using IBM Bob through simple tasks
  • Second half: Experience IBM Bob's true value through complex tasks

This design allows participants to naturally understand IBM Bob's value and acquire practical skills.

Requirements

  • Computer (Mac, Windows) and internet connection
  • IBM Bob (already installed)
  • Web browser (Chrome, Firefox, Safari, Edge, etc.)

Distributed by instructor:

  • Hands-on procedure URL
  • Minimal Vector Search Builder participant package (vector-search-builder-en.zip)
  • Connection information (Milvus connection information)

Next Steps

Let's proceed to the Preparation page!