heroagent/aiprompts/instructions/instructions_doctree.md

there is a module called doctree

the metadata info of doctree is stored in a redis server

there is the concept of collection

a collection has markdown pages each page has a unique name in the collection

a collection has files which can be a sstd file or image each file has a unique name

the names are lower_cased and all non ascci chars are removed, use the namefix function as used in internal/tools/name_fix_test.go

its a struct called DocTree which has as argument a path and a name which is also namefixed

the init walks over the path and finds all files and .md files

we remember the relative position of each file and markdown page in a hset

hset is:

• key: collections:$name
• hkey: pagename.md (always namefixed)
• hkey: imagename.png ... or any other extension for files (always namefixed)
• val: the relative position in the doctree location

use redisclient to internal redis to store this

create following methods on doctree

• Scan (scan the collection again, remove hset and repopulate)
• PageGet get page from a name (do namefix inside method) return the markdown
• PageGetHtml same as PageGet but make html
• FileGetUrl the url which can then be used in static webserver for downloading this content
• PageGetPath relative path in the collection
• Info (name & path)

in PageGet implement a simple include function which is done as !!include name:'pagename' this needs to include page as mentioned in this collection if !!include name:'othercollection:pagename' then pagename comes from other collection do namefix to find

Objects

DocTree

• has add, get, delete, list functions in relation to underlying Collection

Collection

• has get/set/delete/list for pages
• has get/set/delete/list for files

namefix used everywhere to make sure

• in get for page we do the include which an get other pages

special functions

Include

!!include collectionname:'pagename'
!!include collectionname:'pagename.md'
!!include 'pagename'
!!include collectionname:pagename
!!include collectionname:pagename.md

the include needs to parse the following

note:

• pages can have .md or not, check if given if not add
• all is namefixed on collection and page level
• there can be '' around the name but this is optional