Design

How PyNydus works: system architecture, pipeline internals, the Egg interchange format, and how platform connectors map agent state to and from Eggs.

• Architecture
  • Data flow
  • Spawn pipeline
  • Hatch pipeline
  • Module responsibilities
  • Key design decisions
• Spawn Pipeline
  • Pipeline context
  • Step 1: Resolve base egg
  • Step 2: Read source files
  • Step 3: Redaction
  • Step 4: Parse sources
  • Step 5: Build structured records
  • Step 6: Merge with base egg
  • Step 7: LLM refinement (optional)
  • Step 8: Post-processing
  • Step 9: Package egg
  • Step 10: Generate standard artifacts
• Hatch Pipeline
  • Hatch modes
  • Step 1: Version check
  • Step 2: Build file dict
  • Step 3: LLM polish
  • Step 4: Secrets IN
  • Step 5: Write connector (agent) files
  • Step 6: Write standard artifacts
  • Step 7: Hatch log
• The Egg Format
  • Data model
  • The .egg archive format
  • Manifest fields
  • Memory record schema
  • Secret record schema
  • raw/ directory
  • API reference
• Connectors
  • Pipeline overview
  • Egg contents
  • OpenClaw
  • Letta
  • ZeroClaw