Obsidian format parser: aliases, wikilinks, and tags extraction

[AlexMikhalev]

Summary

Extend parse_markdown_directives_content() in terraphim_automata/src/markdown_directives.rs to parse Obsidian-native patterns alongside the existing Logseq synonyms:: syntax.

Current Parser

The parser handles 4 directives:

• synonyms:: -> synonyms (Logseq property syntax)
• type::: -> document type
• route:: -> LLM routing config
• priority:: -> priority score

New Patterns to Parse

1. YAML Frontmatter

---
aliases: [context management, semantic scope]
tags: [engineering, knowledge-graph]
---

• aliases: -> treated as synonyms (same semantics as synonyms::)
• tags: -> stored as concept categories (new field in MarkdownDirectives)
• Supports both YAML list syntax ([a, b]) and multi-line (- a\n- b)

2. Wikilinks

Related to [[knowledge graph]] and [[ontology]] concepts.

• Extract all [[target]] and [[target|display text]] patterns
• Store as related concepts (new Vec<String> field in MarkdownDirectives)
• These become edges in the rolegraph, not synonyms

3. Inline Tags

This concept is important for #engineering and #search.

• Extract #tag patterns (excluding headings ## heading)
• Store as categories alongside YAML frontmatter tags
• Deduplicate with YAML tags

Parser Mode Configuration

Add a format field to haystack config:

[[roles.engineer.haystacks]]
location = "~/.config/terraphim/kg"
format = "terraphim"  # synonyms:: + type::: + route:: + priority:: (default)

[[roles.engineer.haystacks]]
location = "~/synced/pages"
format = "logseq"     # synonyms:: only

[[roles.engineer.haystacks]]
location = "~/synced/ObsidianVault"
format = "obsidian"   # aliases: + [[wikilinks]] + #tags + synonyms::

All formats output the same MarkdownDirectives struct. The obsidian format is a superset that also accepts synonyms:: (for Dataview users).

Extended Struct

pub struct MarkdownDirectives {
    pub doc_type: DocumentType,
    pub synonyms: Vec<String>,           // existing
    pub route: Option<RouteDirective>,    // existing
    pub priority: Option<u8>,            // existing
    pub categories: Vec<String>,         // NEW: from tags:/\#tag
    pub related_concepts: Vec<String>,   // NEW: from [[wikilinks]]
}

Test Plan

• Fixture: real Obsidian vault files with YAML frontmatter, wikilinks, and tags
• Test: aliases parsed as synonyms alongside existing synonyms:: support
• Test: wikilinks extracted as related concepts
• Test: tags extracted from both YAML frontmatter and inline #tags, deduplicated
• Test: mixed format (file has both synonyms:: and aliases:) merges correctly
• Test: obsidian format config flag routes to extended parser

Affected Crates

• terraphim_automata (primary -- extend markdown_directives.rs)
• terraphim_types (add categories + related_concepts to MarkdownDirectives)
• terraphim_config (add format field to haystack config)

Estimated Effort

~4 hours

Part of

Epic #603