How to Build a RAG-Based AI Chatbot on Drupal

Semantic search represents a shift in how content is discovered on websites. Unlike traditional keyword-based search, which relies on exact matches or simple string comparisons, semantic search leverages artificial intelligence (AI), machine learning (ML), and vector embeddings to understand the context and intent of a user query, even when the query includes synonyms or typos.

In Drupal, semantic search is implemented through contributed modules and third-party integrations with AI models or vector database services. It builds on Drupal's modular architecture, often extending the Search API module to bring AI capabilities. In a nutshell, content is converted into high-dimensional vectors that capture semantic essence, then stored in specialized databases, and queried using similarity algorithms such as K-Nearest Neighbors (KNN) or cosine similarity algorithms.

Semantic search enhances content-heavy websites drastically, such as e-commerce platforms, knowledge bases, or education portals.

Benefits of Semantic Search in Drupal

Implementing semantic search in Drupal brings numerous advantages over traditional methods:

• Improve relevance and accuracy: By understanding the intent behind the query, much better results can be delivered, with no "no results found" page. This also benefits from scenarios where users are diverse, and phrase works differently.
• Better user experience: Faster results lead to higher engagement and user satisfaction.
• Better content discoverability: In large Drupal sites, content can get buried easily. With semantic search, buried content becomes accessible again through contextual matching.
• Minimize model hallucinations with Retrieval-Augmented Generation (RAG), where AI models generate responses based on retrieved content.
• Privacy: modules like Semantic Search - AI Vectored Search offer an alternative to proprietary APIs to address GDPR concerns.
• Hybrid capabilities: Combining semantic search with keyword search provides the best of both worlds in achieving precision for complex queries.

All of these benefits can increase site traffic, reduce bounce rate, and improve SEO.

Transform Your Drupal Search Experience with Drupal AI Chatbot

Imagine replacing your Drupal site's traditional search bar with an intelligent assistant always available, understanding user questions in natural language, and delivering precise answers in seconds.

That's the power of integrating an RAG-based AI chatbot. In this blog, we'll show you how to implement it for a Drupal-powered knowledge base, but this approach can enhance any Drupal website, from documentation portals to customer support and beyond.

With a Drupal AI chatbot, you can:

• Enable conversational search: Users ask questions naturally, no exact keywords needed.
• Deliver instant, context-aware responses: Powered by semantic search and vector databases.
• Continuously improve: The chatbot learns from interactions, refining its accuracy over time.

Static search can't match modern expectations. By leveraging Drupal AI, you turn content into smart, interactive experiences no matter how users phrase their questions.

Key Concepts in Drupal AI: Indexing Explained

Think of your AI chatbot as a librarian, not only knowing where every piece of information is stored but also understanding the meaning behind every question asked. To make that possible, indexing is at the core of how Drupal AI delivers fast, accurate answers.

Structured vs. Unstructured Indexing

• Structured Indexing: Perfect for well-defined data like titles, metadata, or tags.
• Unstructured Indexing: Leverages NLP to interpret long-form content articles, guides, and FAQs where meaning goes beyond simple keywords.

Dual Index Approach

• Keyword Index: Handles traditional, fast keyword-based searches.
• Vector Index (RAG): Enables AI-powered, semantic responses by understanding context and intent.

This dual-layered indexing ensures your chatbot delivers both speed and intelligence, finding the right answer, whether it's a simple lookup or a complex query.

Step-by-Step: Building Your Drupal AI Chatbot

Note: This guide is based on a typical Drupal installation using recommended modules. If you're working within a customized knowledge base or Drupal environment, be sure to adjust configurations accordingly to fit your specific setup.

Step 1: Install Drupal AI Modules

When starting any AI integration in Drupal, it's crucial to build on solid, flexible foundations. That's why we use Drupal's AI modules combined with the powerful Search API. These tools are designed to fit seamlessly into Drupal's architecture, saving developers hours of custom work. By connecting to OpenAI's API, we tap into cutting-edge AI capabilities while keeping full control within the Drupal ecosystem. This approach ensures scalability, security, and alignment with Drupal best practices no need to reinvent the wheel.

composer require 'drupal/ai:^1.0'
composer require 'drupal/ai_provider_openai:^1'
composer require 'drupal/search_api:^1'

Step 2: Enable Core AI Modules

Activating the right modules is like unlocking the core engine of your AI chatbot. Each module, whether it's AI Search, AI Chatbot, or AI Assistant API, plays a key role in delivering intelligent, context-aware responses. By enabling them together, you create a modular and extensible AI framework within Drupal. This flexibility allows developers to customize features as projects evolve, making it ideal for both small sites and enterprise-level applications.

This command enables the AI Search, Search API module, AI Chatbot, and AI Assistant API modules.

drush en ai search_api ai_search ai_chatbot ai_assistant_api

Step 3: Add Milvus as a Vector Database Provider

Traditional databases aren't built for understanding context or semantics; that's where Milvus comes in. As a vector database, Milvus allows your AI chatbot to "think" beyond keywords by enabling semantic search. We chose Milvus because it's open-source, developer-friendly, and optimized for handling large-scale AI queries efficiently. This ensures that your Drupal chatbot delivers smarter, more relevant answers, even as your knowledge base grows.
Add Milvus, which is a powerful vector database for semantic search for smarter results.

composer require 'drupal/ai_vdb_provider_milvus:~1.0'
drush en ai_vdb_provider_milvus

Step 4: Create a Zilliz Cloud Account (Milvus)

While Milvus is powerful, managing infrastructure can be a headache. That's why we leverage Zilliz Cloud, the managed service for Milvus. It removes the operational burden, letting developers focus on building features instead of maintaining servers. With just a few clicks, you get a scalable, secure vector database ready to integrate with Drupal, perfect for projects that need rapid deployment without compromising performance.

• Sign up at Zilliz Cloud
• Create a cluster and generate your API key

Step 5: Add the Zilliz API Key to Drupal

Security and seamless integration go hand in hand. By storing your Zilliz API key within Drupal's key management system, you ensure sensitive credentials are handled properly. This step keeps your AI chatbot connected to the vector database while following Drupal's best practices for secure configuration. It's a simple action that sets the stage for stable, long-term performance.

1. Go to /admin/config/system/keys
2. Click Add key
3. Enter:
   a. Key Name: Zilliz Milvus
   b. Key Type: Authentication
   c. Key Value: Paste your API key

Image

Step 6: Configure Milvus in Drupal

Now that your API key is in place, configuring Milvus within Drupal bridges the gap between your CMS and your AI's "memory." This connection allows Drupal to send and retrieve vector-based queries effortlessly. The beauty of this setup is that it abstracts the complexity developers get advanced AI search capabilities without needing deep expertise in vector databases.

1. Navigate to /admin/config/ai/vdb_providers/milvus

   Fill in:
   a. Public Endpoint (Port 443)
   b. API Key: Zilliz Milvus

2. Save and clear cache:

Image

Step 7: Create a Search Server for AI

Every smart chatbot needs a brain, and in this case, the Search Server is exactly that. By setting up a dedicated AI search server, you define how your chatbot will process and retrieve knowledge. Choosing tools like OpenAI for embeddings and Milvus for vector storage ensures your Drupal site is equipped with enterprise-grade AI search power. This modular setup also means you can fine-tune performance as your data or use cases evolve.

1. Go to /admin/config/search/search-api
2. Click Add server
3. Configure:
   a. Server name: "KB AI Search Server."
   b. Backend: "AI Search." 
4. Save and clear the cache 

Then configure:

• Embeddings Engine: OpenAI | text-embedding-3-large
• Tokenizer Model: OpenAI - gpt-4o
• Vector DB: Milvus
• Database Name: KBDatabase1
• Collection: KBCollection1
• Similarity Metric: Cosine Similarity
• Embedding Strategy: "Enriched Embedding Strategy." 

Image

Image

Step 8: Set Up Your RAG Search Index

This is where the magic of Retrieval-Augmented Generation (RAG) comes alive. By creating a tailored RAG index, you enable your chatbot to pull precise, context-rich answers from selected content types like articles, FAQs, and documentation. This approach ensures your AI doesn't just "guess," but retrieves verified information directly from your Drupal Knowledge Base. It's a developer-friendly way to combine AI with content control.

1. Go to Search API > Add Index
2. Name: KB RAG Index
3. Datasource: Select content types (Articles, Books, FAQs, Docs)
4. Server: KB AI Search Server
5. Queue items for reindexing or run the index manually

Image

Image

Image

Field Mapping Tips:

• Main Content: Body/content field
• Context: Titles, summaries, tags
• Filters: Taxonomies, dates, metadata

Step 9: Configure the AI Assistant

An AI assistant is only as good as its instructions. That's why configuring it with a clear pre-prompt is critical. This step defines how your chatbot behaves, ensuring it sticks to your KB content, avoids generating misleading answers, and responds consistently. For developers, this means peace of mind: your AI will always act within boundaries you set, maintaining both accuracy and user trust.

• Go to /admin/config/ai/assistants
• Click Add Assistant
• Name: KB Chatbot
• Pre-prompt: To keep your Drupal AI assistant accurate and reliable, it's important to set clear instructions using a pre-prompt. This pre-prompt tells the assistant exactly how to behave when answering user questions.

You are a Drupal AI assistant. You must strictly retrieve content from the (RAG index) and never generate outside or AI-based content.

Here's how the pre-prompt helps:

1. The assistant only retrieves information from the Drupal Knowledge Base (KB) using RAG search; it never invents or generates answers on its own.
2. If there's no matching content, it responds with a simple message letting the user know that no information was found.
3. When a user says "Thanks" or "Thank you," the assistant replies politely without running a search.
4. All answers include the exact text from the KB along with a clear source link, no changes, no extra content.
5. The assistant will never provide fallback answers or guesswork beyond what's in the KB.

Pre-Action Prompt:
- Immediately analyze the user's input. If it is gratitude, respond with a gratitude text.
- If results exist, provide them verbatim or lightly reformatted, ensuring the correct reference link is appended.

By setting these rules in the pre-prompt, you make sure your AI assistant stays focused, trustworthy, and delivers consistent answers directly from your Drupal Knowledge Base.

System Prompt:
You are an AI assistant that retrieves knowledge only from the KB only, If the user says "Thanks," respond with "You're welcome! Is there anything else I can help you with?".

• RAG Index: KB RAG Index
• RAG Threshold: 0.3–0.4

Step 10: Deploy the DeepChat Chatbot Block

Finally, it's time to bring your AI chatbot to life on your Drupal site. By placing the DeepChat block, you seamlessly integrate the assistant into your website's interface. This block-based approach aligns perfectly with Drupal's flexible layout system, allowing developers to control where and how users interact with the chatbot. With just a few clicks and a cache clear, your intelligent assistant is live, ready to enhance user experience and support.

1. Go to Structure > Block Layout
2. Place the AI DeepChat Chatbot block
3. Select the KB Chatbot assistant
4. Save & clear cache

Image

Key Benefits of Drupal AI Chatbots

Implementing an AI-powered chatbot in your Drupal website isn't just about adding a cool feature; it's about transforming how users access information, improving efficiency, and future-proofing your digital experience. Here's what makes Drupal AI Chatbots a game-changer:

Conclusion: Embrace the Future of User Interaction with Drupal AI

Building a RAG-based AI chatbot isn't just about automation; it's about empowering your website visitors with instant, intelligent, and human-like interactions whenever they need them. By combining Drupal AI, OpenAI embeddings, and vector-powered search, you're not just enhancing your Drupal website; you're future-proofing your entire digital experience with scalable, intelligent technology.

Whether you're a developer crafting smarter solutions or an IT manager optimizing user support and engagement, now is the perfect time to lead with AI-driven innovation.