Skip to main content
This page will help you migrate your Zylon instance to version v1.43.0
Following this migration guide is mandatory when you are updating to versions above to 1.43.0
The guide contains the following steps:
  1. Preparation
  2. Update
  3. Configuration

​
1. Preparation

1

Locate Your Current Configuration

Find your existing Zylon config in /etc/zylon/zylon-conf.yaml and open it.
2

Identify Your Configuration Pattern

Look at your current ai: section and determine which pattern matches:
  • Only has preset: → Scenario 1 (No action needed)
  • Has config.embedding with multilingual settings → Scenario 2
  • Has config.llm or custom LLM settings → Scenario 3a
  • Has config.embedding with custom settings → Scenario 3b
3

Apply the Corresponding Migration

Follow the exact replacement shown in your matching scenario below.
4

Save Your Changes

Update your configuration file with the new format

​
Scenario 1: Base Preset Only (No Action Required)

How to detect: Your configuration only contains a preset line without any config: section. Current Configuration (No Changes Needed): 
ai:
  preset: "baseline-48g"
This configuration remains unchanged and will continue to work as expected.

​
Scenario 2: Multilingual Embedding Configuration

How to detect: Your configuration has config.embedding with multilingual-related settings. Old Configuration: 
ai:
  preset: "baseline-48g"
  config:
    embedding:
      name: multilingual-e5-large
      contextWindow: 512
      vectorDim: 1024
      gpuMemoryUtilization: 0.10
      prefixQuery: "query:"
      prefixText: "passage:"
New Configuration: 
ai:
  preset: "baseline-48g,capabilities.multilingual"
The new preset system automatically handles the multilingual embedding configuration with optimal default settings, providing the same functionality with a much simpler syntax.

​
Scenario 3: Custom Model Configuration

How to detect: Your configuration has config.llm or config.embedding with custom model settings.
If your custom configuration includes non-model related settings, those configurations may remain the same in the new system. Only model-specific configurations need to be migrated to the new models structure.

​
Scenario 3a: Custom LLM Configuration

Look for: config.llm or config section with LLM-specific settings. Old Configuration: This configuration is modifying the default context window. 
ai:
  preset: "baseline-48g"
  config:
    llm:
      name: qwen-3-14b-awq
      contextWindow: 16384
      # ... other LLM settings
New Configuration: 
ai:
  preset: "baseline-48g"
  config:
    models:
      - id: llm
        contextWindow: 16384

​
Scenario 3b: Custom Embedding Configuration

Look for: config.embedding with custom (non-multilingual) settings. Old Configuration: This configuration is modifying the default context window. 
ai:
  preset: "baseline-96g"
  config:
    embedding:
      name: mxbai-embed-large
      contextWindow: 512
New Configuration: 
ai:
  preset: "baseline-96g"
  config:
    models:
      - id: embed
        contextWindow: 512

​
Troubleshooting

Ensure that the YAML syntax is correct and all required fields are properly indented. The models array should be properly formatted with the correct id values.
The new system should guarantee the same performance and configuration as before. Please be sure that you have followed correctly the guide. If you remains having issues, please contact with Zylon team to check it.
Verify that you’re using the correct preset combination format: "baseline-48g,capabilities.multilingual".

​
2. Update

Run sudo zylon-cli update to update Zylon

​
3. Configuration

Now we are going to enable Omnisearch for existing Agent Flows.
1

Update Zylon CLI to at least version 1.16.0

Run sudo zylon-cli and check the version.Run sudo zylon-cli self-update if needed.
2

Backfill content using Zylon CLI

Run sudo zylon-cli backfill-agent-flow-content.Wait for the script to finish. This command can be runned multiple times and stopped at any time.
You are all set to start searching for content inside Agent Flows with Omnisearch. Don’t forget to enable your preferred languages (default is English) to make Omnisearch as efficient as possible following the Search configuration guide