Sync Engine Tool

Overview

Bidirectional sync engine for managing agent definitions between local markdown files and the Langfuse prompt registry. Supports push, pull, diff, status, and list operations with content hashing for change detection and metadata validation.

Available Operations

Push

• Push - Push local agent files to Langfuse
• Validates metadata (required fields: agent_type, cost_tier, output_format, version, status, provider)
• Compares content hashes to skip unchanged files
• Supports dry-run mode

Pull

• Pull - Pull agent definitions from Langfuse to local files
• Creates new files or updates existing ones
• Rebuilds markdown from Langfuse config + prompt body
• Supports dry-run mode

Diff

• Diff - Show differences between local and remote
• Compares content hashes for each agent
• Identifies local-only, remote-only, and modified items

Status

• Status - Show sync status for all agents
• Table view of local/remote existence and sync state

List

• List - List all agents with metadata
• Shows model, description, and sync state

Parameters

Parameter	Type	Required	Description
operation	string	Yes	One of: push, pull, diff, status, list
dryRun	boolean	No	Preview changes without applying (default: false)
validate	boolean	No	Validate metadata before push (default: true)

Metadata Validation

Required fields for agent files:

• agent_type: gene, utility, soul, worker, orchestrator, critic
• cost_tier: low, medium, high
• output_format: json, markdown, text
• version: semantic version (e.g., 1.0.0)
• status: production, deprecated, staging
• provider: openai, anthropic, google

Use Cases

• Keeping local agent files in sync with Langfuse
• Deploying agent changes to production
• Pulling remote changes to local development
• Auditing sync state across environments