- Introduce HeropromptChild to unify file and dir items
- Replace nested Dir/File structs with a flat `children` list
- Generate prompt content by traversing the filesystem on-demand
- Add `workspace.add_file` for direct file selection
- Simplify `workspace.add_dir` to only add the directory path
- Add `list()` method to generate a full workspace file tree
- Introduce `WorkspaceItem` and `WorkspaceList` structs
- Remove `HeropromptSession` to simplify the public API
- Rename Heroscript action to `heropromptworkspace.configure`
- Enable full heroscript encoding/decoding for workspaces
- Add `select_all` option to recursively add directory contents
- Implement `select_all_files_and_dirs` for file traversal
- Rework prompt building with file tree and content formatters
- Improve `get_file_extension` to handle dotfiles and special files
- Update prompt template to use new structured data model
- Introduce sessions and workspaces for managing context
- Allow adding directories and selecting files
- Generate structured prompts with file maps and content
- Add example script and a prompt template
- Define core data models like `HeropromptWorkspace`
- Replace manual script concatenation with playbook include handling
- Preserve site configuration (imports, menu) during generation
- Add support for copying static files from imported content
- Handle static assets from sibling `ebooksall` directories
- Fix import copy logic to not delete destination before copying
- Add central `process_site_from_path` function
- Recursively process heroscript files in `cfg` directory
- Remove duplicated site processing from `run` and `add` commands
- Respect `play` parameter from heroscript `define` block
- Change site.new to always create/overwrite a site
- Add early exit to site.play if no config exists
- Use explicit sitename from docusaurus.config
- Enable openai.play action processing
- Remove debug code and improve struct initialization
- Rework `hero docusaurus` command to use local `cfg` files
- Scan and export doctree collections during site generation
- Fix `baseUrl` redirect path handling in `index.tsx`
- Add cycle detection for `play.include` in playbooks
- Improve site config processing to prevent duplicate items
- Start two local mycelium nodes without TUN interfaces
- Add pre-test cleanup for processes and configurations
- Implement readiness check to wait for servers to start
- Test bidirectional messaging between the two nodes
- Verify message payload and source public key
