Skip to main content

sypho build

Build an agent Docker image with auto-discovery of entrypoints and tools.

Usage​

sypho build --version <VERSION> [OPTIONS]

Options​

FlagDescription
--version <VERSION>Image version tag (required)

What It Does​

  1. Auto-Discovery: Runs auto_package.py to scan code for @entrypoint and @tool decorators
  2. Generates Manifest: Creates manifest.json with agent configuration
  3. Builds Image: Runs docker build with generated Dockerfile
  4. Tags Image: Names image as {agent-name}:{version}

Example​

cd my-agent
sypho build --version 0.1.0

Output:

🔍 Discovering entrypoints and tools...
✓ Found 1 entrypoint: main
✓ Found 2 tools: example_tool, fetch_data

📦 Building Docker image...
[+] Building 12.3s (10/10) FINISHED
✓ Built: my-agent:0.1.0

Generated Manifest​

builds/my-agent/0.1.0/manifest.json:

{
  "schema_version": "1.0",
  "agent_name": "my-agent",
  "version": "0.1.0",
  "execution_mode": "agent",
  "agent_config": {
    "model": "anthropic/claude-3-5-sonnet-20241022",
    "max_tokens": 4096,
    "max_iterations": 10,
    "system_prompt": "You are a helpful assistant."
  },
  "entrypoints": {
    "main": {
      "tools": ["example_tool", "fetch_data"],
      "local_tools": []
    }
  }
}

Docker Image​

Uses multi-stage build:

  1. Install Python dependencies
  2. Copy agent code
  3. Set entrypoint to SDK runtime

Local Testing​

# Verify image
docker images | grep my-agent

# Run locally
docker run my-agent:0.1.0

Troubleshooting​

Auto-discovery fails​

Ensure decorators are used correctly:

from sypho_sdk import entrypoint, tool

@entrypoint  # Required
async def main(input: dict, context: AgentContext):
    ...

Docker build fails​

Check Dockerfile and requirements.txt are present.

Next Steps​