Markdown & TOML Parsing Support

[josephismikhail]

Summary

Adds TOML and Markdown file parsing to the ingestion pipeline.

.toml files are parsed into the graph as config_entry entities — table headers ([package], [profile.release]) and key=value pairs — with CONTAINS relationships linking them.

.md/.markdown files are parsed into the graph as section entities derived from headings (ATX #, setext underline, and HTML <h1>–<h6>), with hierarchical CONTAINS relationships, YAML frontmatter extraction, and fenced code block boundaries respected. Also fixes ix contains disambiguation when the same name appears across multiple files.

Closes #145

Type

• Bug fix
• Feature
• Refactor
• Docs
• Test
• CI

Changes

TOML

• feat(core-ingestion): Add TOML parser — extracts [table] headers and key = value pairs as config_entry entities with CONTAINS relationships, mirroring the YAML parser pattern
• fix(toml): Emit intermediate nodes for dotted table headers — [profile.release] now materialises a profile node in the graph, not just release; a per-file seen-set deduplicates shared prefixes
• fix: Add TOML to the isGrammarSupported early-return guard — .toml files were being silently dropped before reaching parseFile
• feat(cli): Add --path filter to ix contains — allows disambiguating by file path when multiple entities share the same name across workspaces or files

Markdown

• feat(core-ingestion): Add Markdown parser — extracts headings as section entities with hierarchical CONTAINS relationships; supports ATX headings, setext headings, and HTML headings (<h1>–<h6>); parses YAML frontmatter as a body chunk; respects fenced code block boundaries so heading-like lines inside code aren't parsed as headings
• fix(ingest): Add .md and .markdown to SUPPORTED_EXTENSIONS
• fix(core-ingestion): Store section chunks with kind 'section'; bump extractor to 1.21
• fix(cli): Wire up ix config set workspace so it actually takes effect; scope CLI search to active workspace
• fix(cli): Resolve absolute paths, short IDs, and section kind display
• fix(md-parse): Hierarchical section scoping and root-file disambiguation
• fix(md-parse): Clean VitePress heading syntax from entity names (strip {#anchor}, badges, <sup>, escaped angle brackets)
• fix(md-parse): Setext heading detection, fence delimiter matching, frontmatter body chunk

Validation

TOML

• Tested against ripgrep (real-world Rust workspace with multiple Cargo.toml files)
• ix text --language toml returns results from Cargo.toml
• name, version, edition resolve as config_entry with language: toml
• [package], [dependencies] resolve as table entities
• ix contains package returns correct child keys (name, version, edition, authors, etc.)
• [profile.release] → profile intermediate node present; opt-level resolves as child
• ix contains package --path crates/regex resolves without ambiguity prompt
• 263-line test suite in queries.toml.test.ts covers the parser

Markdown

• Tested on 3 repos
• No regressions on existing parsers (ix map on a known TS/JS repo, verify counts unchanged)
• Unit + smoke tests pass (npm test in ix-cli)
• Heading hierarchy chains correctly: file → h1 → h2 → h3
• Both ATX and setext heading syntax ingested
• ix contains returns children for a known markdown file and heading
• Frontmatter entities present where expected
• Section chunks created alongside heading entities
• ix text returns results with language: markdown
• Both .md and .markdown extensions ingested

Checklist

• Tests pass
• Smoke tests pass
• No raw errors introduced
• CLI output follows Ix format