Krochman Solutions

LLM-Solution Documentation

A comprehensive Large Language Model (LLM) integration solution for Salesforce, enabling AI-powered chatbots with Retrieval Augmented Generation (RAG) capabilities.

Named Credential Configuration

Important: Named Credentials are pre-configured for both providers. You only need to update the Authentication Parameter value in the Named Credential’s External Credentials.

OpenAI Configuration

  1. Go to Setup → Named Credentials → OpenAI External Credential
  2. Set Authentication Parameter to: Bearer YOUR_API_KEY_HERE
    • Replace YOUR_API_KEY_HERE with your OpenAI API key
    • Make sure to keep the “Bearer ” prefix (with a space)
    • Get your API key from: OpenAI API Keys
  3. For more details, see the OpenAI Authentication Documentation

Ollama Configuration

  1. Go to Setup → Named Credentials → Ollama External Credential
  2. Set Authentication Parameter to: None
    • Ollama typically doesn’t require authentication for local deployments
    • If you’ve configured authentication, use the appropriate token format
  3. For more details, see the Ollama API Documentation

Core Functionality

  • Dual LLM Provider Support:
    • OpenAI integration for cloud-based LLM capabilities
    • Ollama integration for local/self-hosted LLM deployment
  • RAG (Retrieval Augmented Generation):
    • Knowledge base management with embedding support
    • Multiple source types: Salesforce Records and File Uploads
    • Semantic search capabilities
    • Configurable embedding models
  • Chat Management:
    • Real-time chat interface
    • Chat session persistence
    • Configurable chat transcript management
    • Debug console for development and troubleshooting

Quick Start Guide

  1. Initial Setup:
    • Select your preferred LLM provider (OpenAI or Ollama)
    • Configure the Named Credential for your chosen provider
    • Test the connection using the “Test Connection” button
  2. Knowledge Base Setup:
    • Create a new Knowledge Base with a name and description
    • Select knowledge type (Internal/External)
    • Choose an embedding model
    • Add knowledge sources (Salesforce Records or File Uploads)
  3. Chat Configuration:
    • Configure chat transcript saving options
    • Set up chatbot agents with specific knowledge bases
    • Configure debug console settings if needed

Troubleshooting Tips

  • Named Credential Issues:
    • Verify credential configuration
    • Check API key validity
    • Use “None Selected” as emergency off switch
  • Embedding Generation:
    • Verify knowledge base configuration
    • Check embedding model availability
    • Monitor debug console for errors