Skip to main content

​
Built-in Scaffolds

Egregore includes three production-ready scaffolds for common agent capabilities.

​
InternalNotesScaffold

Automatic note-taking and memory system.

​
Features

  • Automatic capture: Remembers user preferences, facts, and instructions
  • Persistent storage: Notes survive across sessions
  • Search and retrieval: Query notes by content or category
  • Organization: Categorize notes by topic

​
Usage

agent = Agent(provider="openai:gpt-4", enable_scaffolds=True)

# Agent automatically captures notes
agent.call("My email is alice@example.com")
agent.call("I prefer dark mode")
agent.call("Remember: I work in Python")

# Access notes
notes_scaffold = agent.scaffolds["notes"]
print(notes_scaffold.state.notes)
# ["Email: alice@example.com", "Preference: dark mode", "Language: Python"]

# Agent recalls automatically
agent.call("What's my email?")
# "Your email is alice@example.com"

​
Operations

# Add note manually
notes_scaffold.add_note("Important: project deadline March 15")

# Search notes
results = notes_scaffold.search_notes("email")
# ["Email: alice@example.com"]

# Clear notes
notes_scaffold.clear_notes()

​
State Structure

class NotesState(BaseModel):
    notes: list[str] = []
    categories: dict[str, list[str]] = {}
    last_updated: datetime = None

​
FileManager

Track file operations and maintain file system context.

​
Features

  • Operation tracking: Monitors read, write, delete operations
  • Working directory: Tracks current directory
  • File history: Maintains recent file access log
  • Relationship mapping: Understands file dependencies

​
Usage

agent = Agent(provider="openai:gpt-4", enable_scaffolds=True)

# Scaffold tracks file operations automatically
agent.call("Read config.json")
agent.call("Write output to results.txt")

# Access file history
files = agent.scaffolds["files"]
print(files.state.recent_files)
# ["config.json", "results.txt"]

# Working directory awareness
print(files.state.working_directory)
# "/home/user/project"

​
Operations

# Track file operation
files.track_file_operation("read", "data.csv")

# Get recent files
recent = files.get_recent_files(limit=10)

# Clear history
files.clear_history()

​
State Structure

class FileManagerState(BaseModel):
    recent_files: list[str] = []
    working_directory: str = "."
    file_operations: list[dict] = []

​
ShellScaffold

Command execution history and environment tracking.

​
Features

  • Command history: Tracks executed commands
  • Exit codes: Monitors success/failure
  • Environment tracking: Current directory, environment variables
  • Output capture: Stores command output

​
Usage

agent = Agent(provider="openai:gpt-4", enable_scaffolds=True)

# Scaffold tracks commands
agent.call("Run: npm install")
agent.call("Execute: pytest tests/")

# Access history
shell = agent.scaffolds["shell"]
print(shell.state.command_history)
# [("npm install", 0), ("pytest tests/", 0)]

# Check last command
last_cmd, exit_code = shell.state.command_history[-1]
print(f"Last: {last_cmd}, Exit: {exit_code}")

​
Operations

# Track command
shell.track_command("git status", exit_code=0)

# Get command history
history = shell.get_command_history(limit=20)

# Clear history
shell.clear_history()

​
State Structure

class ShellState(BaseModel):
    command_history: list[tuple[str, int]] = []
    working_directory: str = "."
    environment: dict[str, str] = {}

​
Disabling Built-in Scaffolds

# Disable all built-in scaffolds
agent = Agent(provider="openai:gpt-4", enable_scaffolds=False)

# Selective enabling with custom scaffolds
agent = Agent(
    provider="openai:gpt-4",
    scaffolds=[MyCustomScaffold]  # Only custom scaffolds
)

​
What’s Next?

Creating Scaffolds

Build your own custom scaffolds

Scaffold IPC

Learn scaffold communication