Connect Local AI to
Enterprise Databases

🎯 What You'll Build

By the end of this module, you'll have a working API endpoint that:

• Accepts a record identifier as input (customer ID, order number, ticket ID, etc.)
• Retrieves data from your database via a secure stored procedure
• Sends the data to a locally-hosted LLM (like Llama 3 via Ollama)
• Returns an AI-generated summary, analysis, or recommendation

The key insight: your AI never touches the database directly. DreamFactory's AI Data Gateway acts as the secure intermediary, enforcing authentication, rate limiting, and audit logging on every request.

💡 Example Use Cases

This pattern applies to countless business scenarios:

🏗️ Architecture Overview

This integration uses three DreamFactory services working together:

📋 Prerequisites

• DreamFactory 7.x instance (self-hosted or cloud)
• Local LLM running via Ollama, LM Studio, or similar
• Database with data to analyze (SQL Server, MySQL, PostgreSQL, Oracle, etc.)
• API Key configured in DreamFactory with appropriate role
• Basic familiarity with Python and REST APIs

🔧 Implementation Steps

💻 Key Code: Python Orchestration

Here's the core logic of the scripted service that orchestrates the entire flow:

import json, urllib.request

def get_ai_summary(record_id, api_key, base_url):
    # Step 1: Fetch data from database via stored procedure
    url = f"{base_url}/api/v2/db/_proc/GetRecordData"
    url += f"?RecordId={record_id}"

    req = urllib.request.Request(url, method='GET')
    req.add_header('X-DreamFactory-Api-Key', api_key)

    with urllib.request.urlopen(req, timeout=120) as response:
        record_data = json.loads(response.read().decode())

    # Step 2: Build the AI prompt (customize for your use case)
    prompt = f"""Analyze this data and provide a concise summary
with key insights and recommendations:

{json.dumps(record_data, indent=2)}"""

    # Step 3: Call the local LLM
    llm_url = f"{base_url}/api/v2/ollama/api/chat"
    payload = {
        "model": "llama3:70b",
        "messages": [{"role": "user", "content": prompt}]
    }

    req = urllib.request.Request(llm_url,
        data=json.dumps(payload).encode(),
        headers={'Content-Type': 'application/json',
                 'X-DreamFactory-Api-Key': api_key})

    with urllib.request.urlopen(req, timeout=300) as response:
        result = json.loads(response.read().decode())

    return result['message']['content']

⚠️ Common Pitfalls

📊 Real-World Performance

In production environments, enterprise teams are seeing these response times:

• 4-16 seconds: Cloud model (GPT-4o mini) end-to-end response
• 25-32 seconds: Local 14B parameter model on NVIDIA DGX
• Under 3 minutes: Hot-swapping between completely different models

🚀 Next Steps

• Read the full blog post for deeper technical details and troubleshooting
• Explore Module 04 to learn why stored procedures are essential for AI data access
• Try connecting your own database and LLM to build a custom AI-powered API

🎓 Course Complete!

Congratulations! You've completed the DreamFactory AI Academy! You now have the knowledge to connect any AI agent to your enterprise data with full security, RBAC, identity passthrough, and audit logging.