various: markdown improvements

Signed-off-by: NotAShelf <raf@notashelf.dev>
Change-Id: I81fda8247814da19eed1e76dbe97bd5b6a6a6964

crates/pinakes-core/src/import.rs

@@ -6,7 +6,8 @@ use tracing::info;
 use crate::audit;
 use crate::error::{PinakesError, Result};
 use crate::hash::compute_file_hash;
-use crate::media_type::MediaType;
+use crate::links;
+use crate::media_type::{BuiltinMediaType, MediaType};
 use crate::metadata;
 use crate::model::*;
 use crate::storage::DynStorageBackend;
@@ -168,6 +169,9 @@ pub async fn import_file_with_options(
         None
     };
 
+    // Check if this is a markdown file for link extraction
+    let is_markdown = media_type == MediaType::Builtin(BuiltinMediaType::Markdown);
+
     let item = MediaItem {
         id: media_id,
         path: path.clone(),
@@ -206,10 +210,25 @@ pub async fn import_file_with_options(
 
         // New items are not deleted
         deleted_at: None,
+
+        // Links will be extracted separately
+        links_extracted_at: None,
     };
 
     storage.insert_media(&item).await?;
 
+    // Extract and store markdown links for markdown files
+    if is_markdown {
+        if let Err(e) = extract_and_store_links(storage, media_id, &path).await {
+            tracing::warn!(
+                media_id = %media_id,
+                path = %path.display(),
+                error = %e,
+                "failed to extract markdown links"
+            );
+        }
+    }
+
     // Store extracted extra metadata as custom fields
     for (key, value) in &extracted.extra {
         let field = CustomField {
@@ -372,3 +391,44 @@ pub async fn import_directory_with_options(
 
     Ok(results)
 }
+
+/// Extract markdown links from a file and store them in the database.
+async fn extract_and_store_links(
+    storage: &DynStorageBackend,
+    media_id: MediaId,
+    path: &Path,
+) -> Result<()> {
+    // Read file content
+    let content = tokio::fs::read_to_string(path).await.map_err(|e| {
+        PinakesError::Io(std::io::Error::new(
+            std::io::ErrorKind::Other,
+            format!("failed to read markdown file for link extraction: {e}"),
+        ))
+    })?;
+
+    // Extract links
+    let extracted_links = links::extract_links(media_id, &content);
+
+    if extracted_links.is_empty() {
+        // No links found, just mark as extracted
+        storage.mark_links_extracted(media_id).await?;
+        return Ok(());
+    }
+
+    // Clear any existing links for this media (in case of re-import)
+    storage.clear_links_for_media(media_id).await?;
+
+    // Save extracted links
+    storage.save_markdown_links(media_id, &extracted_links).await?;
+
+    // Mark links as extracted
+    storage.mark_links_extracted(media_id).await?;
+
+    tracing::debug!(
+        media_id = %media_id,
+        link_count = extracted_links.len(),
+        "extracted markdown links"
+    );
+
+    Ok(())
+}

crates/pinakes-core/src/lib.rs

@@ -12,6 +12,7 @@ pub mod hash;
 pub mod import;
 pub mod integrity;
 pub mod jobs;
+pub mod links;
 pub mod managed_storage;
 pub mod media_type;
 pub mod metadata;

crates/pinakes-core/src/links.rs

@@ -0,0 +1,456 @@
+//! Markdown link extraction and management for Obsidian-style bidirectional links.
+//!
+//! This module provides:
+//! - Wikilink extraction (`[[target]]` and `[[target|display]]`)
+//! - Embed extraction (`![[target]]`)
+//! - Markdown link extraction (`[text](path)` for internal links)
+//! - Link resolution strategies
+//! - Context extraction for backlink previews
+
+use std::path::Path;
+
+use regex::Regex;
+use uuid::Uuid;
+
+use crate::error::Result;
+use crate::model::{LinkType, MarkdownLink, MediaId};
+
+/// Configuration for context extraction around links
+const CONTEXT_CHARS_BEFORE: usize = 50;
+const CONTEXT_CHARS_AFTER: usize = 50;
+
+/// Extract all markdown links from file content.
+///
+/// This extracts:
+/// - Wikilinks: `[[target]]` and `[[target|display text]]`
+/// - Embeds: `![[target]]`
+/// - Markdown links: `[text](path)` (internal paths only, no http/https)
+pub fn extract_links(source_media_id: MediaId, content: &str) -> Vec<MarkdownLink> {
+    let mut links = Vec::new();
+
+    // Extract wikilinks: [[target]] or [[target|display]]
+    links.extend(extract_wikilinks(source_media_id, content));
+
+    // Extract embeds: ![[target]]
+    links.extend(extract_embeds(source_media_id, content));
+
+    // Extract markdown links: [text](path)
+    links.extend(extract_markdown_links(source_media_id, content));
+
+    links
+}
+
+/// Extract wikilinks from content.
+/// Matches: `[[target]]` or `[[target|display text]]` but NOT `![[...]]` (embeds)
+fn extract_wikilinks(source_media_id: MediaId, content: &str) -> Vec<MarkdownLink> {
+    // Match [[...]] - we'll manually filter out embeds that are preceded by !
+    let re = Regex::new(r"\[\[([^\]|]+)(?:\|([^\]]+))?\]\]").unwrap();
+    let mut links = Vec::new();
+
+    for (line_num, line) in content.lines().enumerate() {
+        for cap in re.captures_iter(line) {
+            let full_match = cap.get(0).unwrap();
+            let match_start = full_match.start();
+
+            // Check if preceded by ! (which would make it an embed, not a wikilink)
+            if match_start > 0 {
+                let bytes = line.as_bytes();
+                if bytes.get(match_start - 1) == Some(&b'!') {
+                    continue; // Skip embeds
+                }
+            }
+
+            let target = cap.get(1).unwrap().as_str().trim();
+            let display_text = cap.get(2).map(|m| m.as_str().trim().to_string());
+
+            let context = extract_context(content, line_num, full_match.start(), full_match.end());
+
+            links.push(MarkdownLink {
+                id: Uuid::now_v7(),
+                source_media_id,
+                target_path: target.to_string(),
+                target_media_id: None, // Will be resolved later
+                link_type: LinkType::Wikilink,
+                link_text: display_text.or_else(|| Some(target.to_string())),
+                line_number: Some(line_num as i32 + 1), // 1-indexed
+                context: Some(context),
+                created_at: chrono::Utc::now(),
+            });
+        }
+    }
+
+    links
+}
+
+/// Extract embeds from content.
+/// Matches: `![[target]]`
+fn extract_embeds(source_media_id: MediaId, content: &str) -> Vec<MarkdownLink> {
+    let re = Regex::new(r"!\[\[([^\]|]+)(?:\|([^\]]+))?\]\]").unwrap();
+    let mut links = Vec::new();
+
+    for (line_num, line) in content.lines().enumerate() {
+        for cap in re.captures_iter(line) {
+            let full_match = cap.get(0).unwrap();
+            let target = cap.get(1).unwrap().as_str().trim();
+            let display_text = cap.get(2).map(|m| m.as_str().trim().to_string());
+
+            let context = extract_context(content, line_num, full_match.start(), full_match.end());
+
+            links.push(MarkdownLink {
+                id: Uuid::now_v7(),
+                source_media_id,
+                target_path: target.to_string(),
+                target_media_id: None,
+                link_type: LinkType::Embed,
+                link_text: display_text.or_else(|| Some(target.to_string())),
+                line_number: Some(line_num as i32 + 1),
+                context: Some(context),
+                created_at: chrono::Utc::now(),
+            });
+        }
+    }
+
+    links
+}
+
+/// Extract markdown links from content.
+/// Matches: `[text](path)` but only for internal paths (no http/https)
+fn extract_markdown_links(source_media_id: MediaId, content: &str) -> Vec<MarkdownLink> {
+    // Match [text](path) where path doesn't start with http:// or https://
+    let re = Regex::new(r"\[([^\]]+)\]\(([^)]+)\)").unwrap();
+    let mut links = Vec::new();
+
+    for (line_num, line) in content.lines().enumerate() {
+        for cap in re.captures_iter(line) {
+            let full_match = cap.get(0).unwrap();
+            let text = cap.get(1).unwrap().as_str().trim();
+            let path = cap.get(2).unwrap().as_str().trim();
+
+            // Skip external links
+            if path.starts_with("http://")
+                || path.starts_with("https://")
+                || path.starts_with("mailto:")
+                || path.starts_with("ftp://")
+            {
+                continue;
+            }
+
+            // Skip anchor-only links
+            if path.starts_with('#') {
+                continue;
+            }
+
+            // Remove any anchor from the path for resolution
+            let target_path = path.split('#').next().unwrap_or(path);
+
+            let context = extract_context(content, line_num, full_match.start(), full_match.end());
+
+            links.push(MarkdownLink {
+                id: Uuid::now_v7(),
+                source_media_id,
+                target_path: target_path.to_string(),
+                target_media_id: None,
+                link_type: LinkType::MarkdownLink,
+                link_text: Some(text.to_string()),
+                line_number: Some(line_num as i32 + 1),
+                context: Some(context),
+                created_at: chrono::Utc::now(),
+            });
+        }
+    }
+
+    links
+}
+
+/// Extract surrounding context for a link.
+fn extract_context(content: &str, line_num: usize, _start: usize, _end: usize) -> String {
+    let lines: Vec<&str> = content.lines().collect();
+    if line_num >= lines.len() {
+        return String::new();
+    }
+
+    let line = lines[line_num];
+    let line_len = line.len();
+
+    // Get surrounding lines for context if the current line is short
+    if line_len < 30 && line_num > 0 {
+        // Include previous line
+        let prev = lines.get(line_num.saturating_sub(1)).unwrap_or(&"");
+        let next = lines.get(line_num + 1).unwrap_or(&"");
+        return format!("{} {} {}", prev.trim(), line.trim(), next.trim())
+            .chars()
+            .take(CONTEXT_CHARS_BEFORE + CONTEXT_CHARS_AFTER + 20)
+            .collect();
+    }
+
+    // Truncate long lines
+    if line_len > CONTEXT_CHARS_BEFORE + CONTEXT_CHARS_AFTER {
+        line.chars()
+            .take(CONTEXT_CHARS_BEFORE + CONTEXT_CHARS_AFTER)
+            .collect()
+    } else {
+        line.to_string()
+    }
+}
+
+/// Link resolution strategies for finding target media items.
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum ResolutionStrategy {
+    /// Direct path match
+    DirectPath,
+    /// Relative to source directory
+    RelativeToSource,
+    /// Filename with .md extension added
+    FilenameWithMd,
+    /// Filename-only search (Obsidian-style)
+    FilenameOnly,
+}
+
+/// Resolve a link target to possible file paths.
+///
+/// Returns a list of candidate paths to check, in order of preference.
+pub fn resolve_link_candidates(
+    target: &str,
+    source_path: &Path,
+    root_dirs: &[std::path::PathBuf],
+) -> Vec<std::path::PathBuf> {
+    let mut candidates = Vec::new();
+
+    // Clean up the target path
+    let target = target.trim();
+
+    // 1. Direct path - if it looks like a path
+    if target.contains('/') || target.contains('\\') {
+        let direct = std::path::PathBuf::from(target);
+        if direct.is_absolute() {
+            candidates.push(direct);
+        } else {
+            // Relative to each root dir
+            for root in root_dirs {
+                candidates.push(root.join(&direct));
+            }
+        }
+    }
+
+    // 2. Relative to source file's directory
+    if let Some(source_dir) = source_path.parent() {
+        let relative = source_dir.join(target);
+        candidates.push(relative.clone());
+
+        // Also try with .md extension
+        if !target.ends_with(".md") {
+            candidates.push(relative.with_extension("md"));
+            let mut with_md = relative.clone();
+            with_md.set_file_name(format!(
+                "{}.md",
+                relative.file_name().unwrap_or_default().to_string_lossy()
+            ));
+            candidates.push(with_md);
+        }
+    }
+
+    // 3. Filename with .md extension in root dirs
+    let target_with_md = if target.ends_with(".md") {
+        target.to_string()
+    } else {
+        format!("{}.md", target)
+    };
+
+    for root in root_dirs {
+        candidates.push(root.join(&target_with_md));
+    }
+
+    // 4. Remove duplicates while preserving order
+    let mut seen = std::collections::HashSet::new();
+    candidates.retain(|p| seen.insert(p.clone()));
+
+    candidates
+}
+
+/// Extract frontmatter aliases from markdown content.
+///
+/// Obsidian uses the `aliases` field in frontmatter to define alternative names
+/// for a note that can be used in wikilinks.
+pub fn extract_aliases(content: &str) -> Result<Vec<String>> {
+    let parsed = gray_matter::Matter::<gray_matter::engine::YAML>::new().parse(content);
+
+    if let Some(data) = parsed.ok().and_then(|p| p.data) {
+        if let gray_matter::Pod::Hash(map) = data {
+            if let Some(aliases) = map.get("aliases") {
+                match aliases {
+                    gray_matter::Pod::Array(arr) => {
+                        return Ok(arr
+                            .iter()
+                            .filter_map(|a| {
+                                if let gray_matter::Pod::String(s) = a {
+                                    Some(s.clone())
+                                } else {
+                                    None
+                                }
+                            })
+                            .collect());
+                    }
+                    gray_matter::Pod::String(s) => {
+                        // Single alias as string
+                        return Ok(vec![s.clone()]);
+                    }
+                    _ => {}
+                }
+            }
+        }
+    }
+
+    Ok(Vec::new())
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    fn test_media_id() -> MediaId {
+        MediaId(Uuid::nil())
+    }
+
+    #[test]
+    fn test_extract_simple_wikilink() {
+        let content = "This is a [[simple link]] in text.";
+        let links = extract_links(test_media_id(), content);
+
+        assert_eq!(links.len(), 1);
+        assert_eq!(links[0].target_path, "simple link");
+        assert_eq!(links[0].link_type, LinkType::Wikilink);
+        assert_eq!(links[0].link_text, Some("simple link".to_string()));
+    }
+
+    #[test]
+    fn test_extract_wikilink_with_display() {
+        let content = "Check out [[target note|this article]] for more.";
+        let links = extract_links(test_media_id(), content);
+
+        assert_eq!(links.len(), 1);
+        assert_eq!(links[0].target_path, "target note");
+        assert_eq!(links[0].link_text, Some("this article".to_string()));
+    }
+
+    #[test]
+    fn test_extract_embed() {
+        let content = "Here is an image: ![[image.png]]";
+        let links = extract_links(test_media_id(), content);
+
+        assert_eq!(links.len(), 1);
+        assert_eq!(links[0].target_path, "image.png");
+        assert_eq!(links[0].link_type, LinkType::Embed);
+    }
+
+    #[test]
+    fn test_extract_markdown_link() {
+        let content = "Read [the documentation](docs/README.md) for details.";
+        let links = extract_links(test_media_id(), content);
+
+        assert_eq!(links.len(), 1);
+        assert_eq!(links[0].target_path, "docs/README.md");
+        assert_eq!(links[0].link_type, LinkType::MarkdownLink);
+        assert_eq!(links[0].link_text, Some("the documentation".to_string()));
+    }
+
+    #[test]
+    fn test_skip_external_links() {
+        let content = "Visit [our site](https://example.com) or [email us](mailto:test@test.com).";
+        let links = extract_links(test_media_id(), content);
+
+        assert!(links.is_empty());
+    }
+
+    #[test]
+    fn test_multiple_links() {
+        let content = r#"
+# My Note
+
+This links to [[Note A]] and also [[Note B|Note B Title]].
+
+We also have a markdown link to [config](./config.md).
+
+And an embedded image: ![[diagram.png]]
+"#;
+        let links = extract_links(test_media_id(), content);
+
+        assert_eq!(links.len(), 4);
+
+        let types: Vec<_> = links.iter().map(|l| l.link_type).collect();
+        assert!(types.contains(&LinkType::Wikilink));
+        assert!(types.contains(&LinkType::Embed));
+        assert!(types.contains(&LinkType::MarkdownLink));
+    }
+
+    #[test]
+    fn test_line_numbers() {
+        let content = "Line 1\n[[link on line 2]]\nLine 3";
+        let links = extract_links(test_media_id(), content);
+
+        assert_eq!(links.len(), 1);
+        assert_eq!(links[0].line_number, Some(2));
+    }
+
+    #[test]
+    fn test_resolve_candidates() {
+        let source_path = std::path::Path::new("/notes/projects/readme.md");
+        let root_dirs = vec![std::path::PathBuf::from("/notes")];
+
+        let candidates = resolve_link_candidates("My Note", source_path, &root_dirs);
+
+        // Should include relative path and .md variations
+        assert!(!candidates.is_empty());
+        assert!(candidates
+            .iter()
+            .any(|p| p.to_string_lossy().contains("My Note.md")));
+    }
+
+    #[test]
+    fn test_extract_aliases() {
+        let content = r#"---
+title: My Note
+aliases:
+  - Alternative Name
+  - Another Alias
+---
+
+# Content here
+"#;
+        let aliases = extract_aliases(content).unwrap();
+        assert_eq!(aliases, vec!["Alternative Name", "Another Alias"]);
+    }
+
+    #[test]
+    fn test_extract_single_alias() {
+        let content = r#"---
+title: My Note
+aliases: Single Alias
+---
+
+# Content
+"#;
+        let aliases = extract_aliases(content).unwrap();
+        assert_eq!(aliases, vec!["Single Alias"]);
+    }
+
+    #[test]
+    fn test_wikilink_not_matching_embed() {
+        let content = "A wikilink [[note]] and an embed ![[image.png]]";
+        let links = extract_links(test_media_id(), content);
+
+        assert_eq!(links.len(), 2);
+        let wikilinks: Vec<_> = links
+            .iter()
+            .filter(|l| l.link_type == LinkType::Wikilink)
+            .collect();
+        let embeds: Vec<_> = links
+            .iter()
+            .filter(|l| l.link_type == LinkType::Embed)
+            .collect();
+
+        assert_eq!(wikilinks.len(), 1);
+        assert_eq!(embeds.len(), 1);
+        assert_eq!(wikilinks[0].target_path, "note");
+        assert_eq!(embeds[0].target_path, "image.png");
+    }
+}

crates/pinakes-core/src/model.rs

@@ -154,6 +154,9 @@ pub struct MediaItem {
 
     /// Soft delete timestamp. If set, the item is in the trash.
     pub deleted_at: Option<DateTime<Utc>>,
+
+    /// When markdown links were last extracted from this file.
+    pub links_extracted_at: Option<DateTime<Utc>>,
 }
 
 #[derive(Debug, Clone, Serialize, Deserialize)]
@@ -486,3 +489,100 @@ impl fmt::Display for ReadingStatus {
         }
     }
 }
+
+// ===== Markdown Links (Obsidian-style) =====
+
+/// Type of markdown link
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum LinkType {
+    /// Wikilink: [[target]] or [[target|display]]
+    Wikilink,
+    /// Markdown link: [text](path)
+    MarkdownLink,
+    /// Embed: ![[target]]
+    Embed,
+}
+
+impl fmt::Display for LinkType {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        match self {
+            Self::Wikilink => write!(f, "wikilink"),
+            Self::MarkdownLink => write!(f, "markdown_link"),
+            Self::Embed => write!(f, "embed"),
+        }
+    }
+}
+
+impl std::str::FromStr for LinkType {
+    type Err = String;
+
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        match s.to_lowercase().as_str() {
+            "wikilink" => Ok(Self::Wikilink),
+            "markdown_link" => Ok(Self::MarkdownLink),
+            "embed" => Ok(Self::Embed),
+            _ => Err(format!("unknown link type: {}", s)),
+        }
+    }
+}
+
+/// A markdown link extracted from a file
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct MarkdownLink {
+    pub id: Uuid,
+    pub source_media_id: MediaId,
+    /// Raw link target as written in the source (wikilink name or path)
+    pub target_path: String,
+    /// Resolved target media_id (None if unresolved)
+    pub target_media_id: Option<MediaId>,
+    pub link_type: LinkType,
+    /// Display text for the link
+    pub link_text: Option<String>,
+    /// Line number in source file (1-indexed)
+    pub line_number: Option<i32>,
+    /// Surrounding text for backlink preview
+    pub context: Option<String>,
+    pub created_at: DateTime<Utc>,
+}
+
+/// Information about a backlink (incoming link)
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct BacklinkInfo {
+    pub link_id: Uuid,
+    pub source_id: MediaId,
+    pub source_title: Option<String>,
+    pub source_path: String,
+    pub link_text: Option<String>,
+    pub line_number: Option<i32>,
+    pub context: Option<String>,
+    pub link_type: LinkType,
+}
+
+/// Graph data for visualization
+#[derive(Debug, Clone, Default, Serialize, Deserialize)]
+pub struct GraphData {
+    pub nodes: Vec<GraphNode>,
+    pub edges: Vec<GraphEdge>,
+}
+
+/// A node in the graph visualization
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct GraphNode {
+    pub id: String,
+    pub label: String,
+    pub title: Option<String>,
+    pub media_type: String,
+    /// Number of outgoing links from this node
+    pub link_count: u32,
+    /// Number of incoming links to this node
+    pub backlink_count: u32,
+}
+
+/// An edge (link) in the graph visualization
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct GraphEdge {
+    pub source: String,
+    pub target: String,
+    pub link_type: LinkType,
+}

crates/pinakes-core/src/storage/mod.rs

@@ -789,6 +789,45 @@ pub trait StorageBackend: Send + Sync + 'static {
 
     /// Count items in trash.
     async fn count_trash(&self) -> Result<u64>;
+
+    // ===== Markdown Links (Obsidian-style) =====
+
+    /// Save extracted markdown links for a media item.
+    /// This replaces any existing links for the source media.
+    async fn save_markdown_links(
+        &self,
+        media_id: MediaId,
+        links: &[crate::model::MarkdownLink],
+    ) -> Result<()>;
+
+    /// Get outgoing links from a media item.
+    async fn get_outgoing_links(&self, media_id: MediaId) -> Result<Vec<crate::model::MarkdownLink>>;
+
+    /// Get backlinks (incoming links) to a media item.
+    async fn get_backlinks(&self, media_id: MediaId) -> Result<Vec<crate::model::BacklinkInfo>>;
+
+    /// Clear all links for a media item.
+    async fn clear_links_for_media(&self, media_id: MediaId) -> Result<()>;
+
+    /// Get graph data for visualization.
+    ///
+    /// If `center_id` is provided, returns nodes within `depth` hops of that node.
+    /// If `center_id` is None, returns the entire graph (limited by internal max).
+    async fn get_graph_data(
+        &self,
+        center_id: Option<MediaId>,
+        depth: u32,
+    ) -> Result<crate::model::GraphData>;
+
+    /// Resolve unresolved links by matching target_path against media item paths.
+    /// Returns the number of links that were resolved.
+    async fn resolve_links(&self) -> Result<u64>;
+
+    /// Update the links_extracted_at timestamp for a media item.
+    async fn mark_links_extracted(&self, media_id: MediaId) -> Result<()>;
+
+    /// Get count of unresolved links (links where target_media_id is NULL).
+    async fn count_unresolved_links(&self) -> Result<u64>;
 }
 
 /// Comprehensive library statistics.

crates/pinakes-core/src/storage/postgres.rs

@@ -200,6 +200,9 @@ fn row_to_media_item(row: &Row) -> Result<MediaItem> {
 
         // Trash support
         deleted_at: row.try_get("deleted_at").ok().flatten(),
+
+        // Markdown links extraction timestamp
+        links_extracted_at: row.try_get("links_extracted_at").ok().flatten(),
     })
 }
 
@@ -6036,6 +6039,425 @@ impl StorageBackend for PostgresBackend {
         let count: i64 = row.get(0);
         Ok(count as u64)
     }
+
+    // ===== Markdown Links (Obsidian-style) =====
+
+    async fn save_markdown_links(
+        &self,
+        media_id: MediaId,
+        links: &[crate::model::MarkdownLink],
+    ) -> Result<()> {
+        let client = self
+            .pool
+            .get()
+            .await
+            .map_err(|e| PinakesError::Database(format!("pool error: {e}")))?;
+
+        let media_id_str = media_id.0.to_string();
+
+        // Delete existing links for this source
+        client
+            .execute(
+                "DELETE FROM markdown_links WHERE source_media_id = $1",
+                &[&media_id_str],
+            )
+            .await
+            .map_err(|e| PinakesError::Database(e.to_string()))?;
+
+        // Insert new links
+        for link in links {
+            let target_media_id = link.target_media_id.map(|id| id.0.to_string());
+            client
+                .execute(
+                    "INSERT INTO markdown_links (
+                        id, source_media_id, target_path, target_media_id,
+                        link_type, link_text, line_number, context, created_at
+                    ) VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9)",
+                    &[
+                        &link.id.to_string(),
+                        &media_id_str,
+                        &link.target_path,
+                        &target_media_id,
+                        &link.link_type.to_string(),
+                        &link.link_text,
+                        &link.line_number,
+                        &link.context,
+                        &link.created_at,
+                    ],
+                )
+                .await
+                .map_err(|e| PinakesError::Database(e.to_string()))?;
+        }
+
+        Ok(())
+    }
+
+    async fn get_outgoing_links(&self, media_id: MediaId) -> Result<Vec<crate::model::MarkdownLink>> {
+        let client = self
+            .pool
+            .get()
+            .await
+            .map_err(|e| PinakesError::Database(format!("pool error: {e}")))?;
+
+        let media_id_str = media_id.0.to_string();
+
+        let rows = client
+            .query(
+                "SELECT id, source_media_id, target_path, target_media_id,
+                        link_type, link_text, line_number, context, created_at
+                 FROM markdown_links
+                 WHERE source_media_id = $1
+                 ORDER BY line_number",
+                &[&media_id_str],
+            )
+            .await
+            .map_err(|e| PinakesError::Database(e.to_string()))?;
+
+        let mut links = Vec::new();
+        for row in rows {
+            links.push(row_to_markdown_link(&row)?);
+        }
+
+        Ok(links)
+    }
+
+    async fn get_backlinks(&self, media_id: MediaId) -> Result<Vec<crate::model::BacklinkInfo>> {
+        let client = self
+            .pool
+            .get()
+            .await
+            .map_err(|e| PinakesError::Database(format!("pool error: {e}")))?;
+
+        let media_id_str = media_id.0.to_string();
+
+        let rows = client
+            .query(
+                "SELECT l.id, l.source_media_id, m.title, m.path,
+                        l.link_text, l.line_number, l.context, l.link_type
+                 FROM markdown_links l
+                 JOIN media_items m ON l.source_media_id = m.id
+                 WHERE l.target_media_id = $1
+                 ORDER BY m.title, l.line_number",
+                &[&media_id_str],
+            )
+            .await
+            .map_err(|e| PinakesError::Database(e.to_string()))?;
+
+        let mut backlinks = Vec::new();
+        for row in rows {
+            let link_id_str: String = row.get(0);
+            let source_id_str: String = row.get(1);
+            let source_title: Option<String> = row.get(2);
+            let source_path: String = row.get(3);
+            let link_text: Option<String> = row.get(4);
+            let line_number: Option<i32> = row.get(5);
+            let context: Option<String> = row.get(6);
+            let link_type_str: String = row.get(7);
+
+            backlinks.push(crate::model::BacklinkInfo {
+                link_id: Uuid::parse_str(&link_id_str)
+                    .map_err(|e| PinakesError::Database(e.to_string()))?,
+                source_id: MediaId(
+                    Uuid::parse_str(&source_id_str)
+                        .map_err(|e| PinakesError::Database(e.to_string()))?,
+                ),
+                source_title,
+                source_path,
+                link_text,
+                line_number,
+                context,
+                link_type: link_type_str
+                    .parse()
+                    .unwrap_or(crate::model::LinkType::Wikilink),
+            });
+        }
+
+        Ok(backlinks)
+    }
+
+    async fn clear_links_for_media(&self, media_id: MediaId) -> Result<()> {
+        let client = self
+            .pool
+            .get()
+            .await
+            .map_err(|e| PinakesError::Database(format!("pool error: {e}")))?;
+
+        let media_id_str = media_id.0.to_string();
+
+        client
+            .execute(
+                "DELETE FROM markdown_links WHERE source_media_id = $1",
+                &[&media_id_str],
+            )
+            .await
+            .map_err(|e| PinakesError::Database(e.to_string()))?;
+
+        Ok(())
+    }
+
+    async fn get_graph_data(
+        &self,
+        center_id: Option<MediaId>,
+        depth: u32,
+    ) -> Result<crate::model::GraphData> {
+        let client = self
+            .pool
+            .get()
+            .await
+            .map_err(|e| PinakesError::Database(format!("pool error: {e}")))?;
+
+        let depth = depth.min(5); // Limit depth
+        let mut nodes = Vec::new();
+        let mut edges = Vec::new();
+        let mut node_ids: std::collections::HashSet<String> = std::collections::HashSet::new();
+
+        if let Some(center) = center_id {
+            // BFS to find connected nodes within depth
+            let mut frontier = vec![center.0.to_string()];
+            let mut visited = std::collections::HashSet::new();
+            visited.insert(center.0.to_string());
+
+            for _ in 0..depth {
+                if frontier.is_empty() {
+                    break;
+                }
+                let mut next_frontier = Vec::new();
+
+                for node_id in &frontier {
+                    // Get outgoing links
+                    let rows = client
+                        .query(
+                            "SELECT target_media_id FROM markdown_links
+                             WHERE source_media_id = $1 AND target_media_id IS NOT NULL",
+                            &[node_id],
+                        )
+                        .await
+                        .map_err(|e| PinakesError::Database(e.to_string()))?;
+
+                    for row in rows {
+                        let id: String = row.get(0);
+                        if !visited.contains(&id) {
+                            visited.insert(id.clone());
+                            next_frontier.push(id);
+                        }
+                    }
+
+                    // Get incoming links
+                    let rows = client
+                        .query(
+                            "SELECT source_media_id FROM markdown_links
+                             WHERE target_media_id = $1",
+                            &[node_id],
+                        )
+                        .await
+                        .map_err(|e| PinakesError::Database(e.to_string()))?;
+
+                    for row in rows {
+                        let id: String = row.get(0);
+                        if !visited.contains(&id) {
+                            visited.insert(id.clone());
+                            next_frontier.push(id);
+                        }
+                    }
+                }
+
+                frontier = next_frontier;
+            }
+
+            node_ids = visited;
+        } else {
+            // Get all markdown files with links (limit to 500)
+            let rows = client
+                .query(
+                    "SELECT DISTINCT id FROM media_items
+                     WHERE media_type = 'markdown' AND deleted_at IS NULL
+                     LIMIT 500",
+                    &[],
+                )
+                .await
+                .map_err(|e| PinakesError::Database(e.to_string()))?;
+
+            for row in rows {
+                let id: String = row.get(0);
+                node_ids.insert(id);
+            }
+        }
+
+        // Build nodes with metadata
+        for node_id in &node_ids {
+            let row = client
+                .query_opt(
+                    "SELECT id, COALESCE(title, file_name) as label, title, media_type
+                     FROM media_items WHERE id = $1",
+                    &[node_id],
+                )
+                .await
+                .map_err(|e| PinakesError::Database(e.to_string()))?;
+
+            if let Some(row) = row {
+                let id: String = row.get(0);
+                let label: String = row.get(1);
+                let title: Option<String> = row.get(2);
+                let media_type: String = row.get(3);
+
+                // Count outgoing links
+                let link_count_row = client
+                    .query_one(
+                        "SELECT COUNT(*) FROM markdown_links WHERE source_media_id = $1",
+                        &[&id],
+                    )
+                    .await
+                    .map_err(|e| PinakesError::Database(e.to_string()))?;
+                let link_count: i64 = link_count_row.get(0);
+
+                // Count incoming links
+                let backlink_count_row = client
+                    .query_one(
+                        "SELECT COUNT(*) FROM markdown_links WHERE target_media_id = $1",
+                        &[&id],
+                    )
+                    .await
+                    .map_err(|e| PinakesError::Database(e.to_string()))?;
+                let backlink_count: i64 = backlink_count_row.get(0);
+
+                nodes.push(crate::model::GraphNode {
+                    id: id.clone(),
+                    label,
+                    title,
+                    media_type,
+                    link_count: link_count as u32,
+                    backlink_count: backlink_count as u32,
+                });
+            }
+        }
+
+        // Build edges
+        for node_id in &node_ids {
+            let rows = client
+                .query(
+                    "SELECT source_media_id, target_media_id, link_type
+                     FROM markdown_links
+                     WHERE source_media_id = $1 AND target_media_id IS NOT NULL",
+                    &[node_id],
+                )
+                .await
+                .map_err(|e| PinakesError::Database(e.to_string()))?;
+
+            for row in rows {
+                let source: String = row.get(0);
+                let target: String = row.get(1);
+                let link_type_str: String = row.get(2);
+
+                if node_ids.contains(&target) {
+                    edges.push(crate::model::GraphEdge {
+                        source,
+                        target,
+                        link_type: link_type_str
+                            .parse()
+                            .unwrap_or(crate::model::LinkType::Wikilink),
+                    });
+                }
+            }
+        }
+
+        Ok(crate::model::GraphData { nodes, edges })
+    }
+
+    async fn resolve_links(&self) -> Result<u64> {
+        let client = self
+            .pool
+            .get()
+            .await
+            .map_err(|e| PinakesError::Database(format!("pool error: {e}")))?;
+
+        // Strategy 1: Exact path match
+        let result1 = client
+            .execute(
+                "UPDATE markdown_links
+                 SET target_media_id = (
+                     SELECT id FROM media_items
+                     WHERE path = markdown_links.target_path
+                     AND deleted_at IS NULL
+                     LIMIT 1
+                 )
+                 WHERE target_media_id IS NULL
+                 AND EXISTS (
+                     SELECT 1 FROM media_items
+                     WHERE path = markdown_links.target_path
+                     AND deleted_at IS NULL
+                 )",
+                &[],
+            )
+            .await
+            .map_err(|e| PinakesError::Database(e.to_string()))?;
+
+        // Strategy 2: Filename match
+        let result2 = client
+            .execute(
+                "UPDATE markdown_links
+                 SET target_media_id = (
+                     SELECT id FROM media_items
+                     WHERE (file_name = markdown_links.target_path
+                            OR file_name = markdown_links.target_path || '.md'
+                            OR REPLACE(file_name, '.md', '') = markdown_links.target_path)
+                     AND deleted_at IS NULL
+                     LIMIT 1
+                 )
+                 WHERE target_media_id IS NULL
+                 AND EXISTS (
+                     SELECT 1 FROM media_items
+                     WHERE (file_name = markdown_links.target_path
+                            OR file_name = markdown_links.target_path || '.md'
+                            OR REPLACE(file_name, '.md', '') = markdown_links.target_path)
+                     AND deleted_at IS NULL
+                 )",
+                &[],
+            )
+            .await
+            .map_err(|e| PinakesError::Database(e.to_string()))?;
+
+        Ok(result1 + result2)
+    }
+
+    async fn mark_links_extracted(&self, media_id: MediaId) -> Result<()> {
+        let client = self
+            .pool
+            .get()
+            .await
+            .map_err(|e| PinakesError::Database(format!("pool error: {e}")))?;
+
+        let media_id_str = media_id.0.to_string();
+        let now = chrono::Utc::now();
+
+        client
+            .execute(
+                "UPDATE media_items SET links_extracted_at = $1 WHERE id = $2",
+                &[&now, &media_id_str],
+            )
+            .await
+            .map_err(|e| PinakesError::Database(e.to_string()))?;
+
+        Ok(())
+    }
+
+    async fn count_unresolved_links(&self) -> Result<u64> {
+        let client = self
+            .pool
+            .get()
+            .await
+            .map_err(|e| PinakesError::Database(format!("pool error: {e}")))?;
+
+        let row = client
+            .query_one(
+                "SELECT COUNT(*) FROM markdown_links WHERE target_media_id IS NULL",
+                &[],
+            )
+            .await
+            .map_err(|e| PinakesError::Database(e.to_string()))?;
+
+        let count: i64 = row.get(0);
+        Ok(count as u64)
+    }
 }
 
 impl PostgresBackend {
@@ -6329,6 +6751,37 @@ fn find_first_fts_param(query: &SearchQuery) -> i32 {
     find_inner(query, &mut offset).unwrap_or(1)
 }
 
+// Helper function to parse a markdown link row
+fn row_to_markdown_link(row: &Row) -> Result<crate::model::MarkdownLink> {
+    let id_str: String = row.get(0);
+    let source_id_str: String = row.get(1);
+    let target_path: String = row.get(2);
+    let target_id: Option<String> = row.get(3);
+    let link_type_str: String = row.get(4);
+    let link_text: Option<String> = row.get(5);
+    let line_number: Option<i32> = row.get(6);
+    let context: Option<String> = row.get(7);
+    let created_at: chrono::DateTime<Utc> = row.get(8);
+
+    Ok(crate::model::MarkdownLink {
+        id: Uuid::parse_str(&id_str).map_err(|e| PinakesError::Database(e.to_string()))?,
+        source_media_id: MediaId(
+            Uuid::parse_str(&source_id_str).map_err(|e| PinakesError::Database(e.to_string()))?,
+        ),
+        target_path,
+        target_media_id: target_id
+            .and_then(|s| Uuid::parse_str(&s).ok())
+            .map(MediaId),
+        link_type: link_type_str
+            .parse()
+            .unwrap_or(crate::model::LinkType::Wikilink),
+        link_text,
+        line_number,
+        context,
+        created_at,
+    })
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;

crates/pinakes-core/src/storage/sqlite.rs

@@ -160,6 +160,14 @@ fn row_to_media_item(row: &Row) -> rusqlite::Result<MediaItem> {
             .flatten()
             .and_then(|s| DateTime::parse_from_rfc3339(&s).ok())
             .map(|dt| dt.with_timezone(&Utc)),
+
+        // Markdown links extraction timestamp
+        links_extracted_at: row
+            .get::<_, Option<String>>("links_extracted_at")
+            .ok()
+            .flatten()
+            .and_then(|s| DateTime::parse_from_rfc3339(&s).ok())
+            .map(|dt| dt.with_timezone(&Utc)),
     })
 }
 
@@ -6379,6 +6387,428 @@ impl StorageBackend for SqliteBackend {
 
         Ok(count)
     }
+
+    // ===== Markdown Links (Obsidian-style) =====
+
+    async fn save_markdown_links(
+        &self,
+        media_id: MediaId,
+        links: &[crate::model::MarkdownLink],
+    ) -> Result<()> {
+        let conn = self.conn.clone();
+        let media_id_str = media_id.0.to_string();
+        let links: Vec<_> = links.to_vec();
+
+        tokio::task::spawn_blocking(move || {
+            let conn = conn.lock().unwrap();
+
+            // Delete existing links for this source
+            conn.execute(
+                "DELETE FROM markdown_links WHERE source_media_id = ?1",
+                [&media_id_str],
+            )?;
+
+            // Insert new links
+            let mut stmt = conn.prepare(
+                "INSERT INTO markdown_links (
+                    id, source_media_id, target_path, target_media_id,
+                    link_type, link_text, line_number, context, created_at
+                ) VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, ?8, ?9)"
+            )?;
+
+            for link in &links {
+                stmt.execute(params![
+                    link.id.to_string(),
+                    media_id_str,
+                    link.target_path,
+                    link.target_media_id.map(|id| id.0.to_string()),
+                    link.link_type.to_string(),
+                    link.link_text,
+                    link.line_number,
+                    link.context,
+                    link.created_at.to_rfc3339(),
+                ])?;
+            }
+
+            Ok::<_, rusqlite::Error>(())
+        })
+        .await
+        .map_err(|e| PinakesError::Database(e.to_string()))??;
+
+        Ok(())
+    }
+
+    async fn get_outgoing_links(&self, media_id: MediaId) -> Result<Vec<crate::model::MarkdownLink>> {
+        let conn = self.conn.clone();
+        let media_id_str = media_id.0.to_string();
+
+        let links = tokio::task::spawn_blocking(move || {
+            let conn = conn.lock().unwrap();
+            let mut stmt = conn.prepare(
+                "SELECT id, source_media_id, target_path, target_media_id,
+                        link_type, link_text, line_number, context, created_at
+                 FROM markdown_links
+                 WHERE source_media_id = ?1
+                 ORDER BY line_number"
+            )?;
+
+            let rows = stmt.query_map([&media_id_str], |row| {
+                row_to_markdown_link(row)
+            })?;
+
+            let mut links = Vec::new();
+            for row in rows {
+                links.push(row?);
+            }
+            Ok::<_, rusqlite::Error>(links)
+        })
+        .await
+        .map_err(|e| PinakesError::Database(e.to_string()))??;
+
+        Ok(links)
+    }
+
+    async fn get_backlinks(&self, media_id: MediaId) -> Result<Vec<crate::model::BacklinkInfo>> {
+        let conn = self.conn.clone();
+        let media_id_str = media_id.0.to_string();
+
+        let backlinks = tokio::task::spawn_blocking(move || {
+            let conn = conn.lock().unwrap();
+            let mut stmt = conn.prepare(
+                "SELECT l.id, l.source_media_id, m.title, m.path,
+                        l.link_text, l.line_number, l.context, l.link_type
+                 FROM markdown_links l
+                 JOIN media_items m ON l.source_media_id = m.id
+                 WHERE l.target_media_id = ?1
+                 ORDER BY m.title, l.line_number"
+            )?;
+
+            let rows = stmt.query_map([&media_id_str], |row| {
+                let link_id_str: String = row.get(0)?;
+                let source_id_str: String = row.get(1)?;
+                let source_title: Option<String> = row.get(2)?;
+                let source_path: String = row.get(3)?;
+                let link_text: Option<String> = row.get(4)?;
+                let line_number: Option<i32> = row.get(5)?;
+                let context: Option<String> = row.get(6)?;
+                let link_type_str: String = row.get(7)?;
+
+                Ok(crate::model::BacklinkInfo {
+                    link_id: parse_uuid(&link_id_str)?,
+                    source_id: MediaId(parse_uuid(&source_id_str)?),
+                    source_title,
+                    source_path,
+                    link_text,
+                    line_number,
+                    context,
+                    link_type: link_type_str.parse().unwrap_or(crate::model::LinkType::Wikilink),
+                })
+            })?;
+
+            let mut backlinks = Vec::new();
+            for row in rows {
+                backlinks.push(row?);
+            }
+            Ok::<_, rusqlite::Error>(backlinks)
+        })
+        .await
+        .map_err(|e| PinakesError::Database(e.to_string()))??;
+
+        Ok(backlinks)
+    }
+
+    async fn clear_links_for_media(&self, media_id: MediaId) -> Result<()> {
+        let conn = self.conn.clone();
+        let media_id_str = media_id.0.to_string();
+
+        tokio::task::spawn_blocking(move || {
+            let conn = conn.lock().unwrap();
+            conn.execute(
+                "DELETE FROM markdown_links WHERE source_media_id = ?1",
+                [&media_id_str],
+            )?;
+            Ok::<_, rusqlite::Error>(())
+        })
+        .await
+        .map_err(|e| PinakesError::Database(e.to_string()))??;
+
+        Ok(())
+    }
+
+    async fn get_graph_data(
+        &self,
+        center_id: Option<MediaId>,
+        depth: u32,
+    ) -> Result<crate::model::GraphData> {
+        let conn = self.conn.clone();
+        let center_id_str = center_id.map(|id| id.0.to_string());
+        let depth = depth.min(5); // Limit depth to prevent huge queries
+
+        let graph_data = tokio::task::spawn_blocking(move || {
+            let conn = conn.lock().unwrap();
+            let mut nodes = Vec::new();
+            let mut edges = Vec::new();
+            let mut node_ids = std::collections::HashSet::new();
+
+            // Get nodes - either all markdown files or those connected to center
+            if let Some(center_id) = center_id_str {
+                // BFS to find connected nodes within depth
+                let mut frontier = vec![center_id.clone()];
+                let mut visited = std::collections::HashSet::new();
+                visited.insert(center_id.clone());
+
+                for _ in 0..depth {
+                    let mut next_frontier = Vec::new();
+
+                    for node_id in &frontier {
+                        // Get outgoing links
+                        let mut stmt = conn.prepare(
+                            "SELECT target_media_id FROM markdown_links
+                             WHERE source_media_id = ?1 AND target_media_id IS NOT NULL"
+                        )?;
+                        let rows = stmt.query_map([node_id], |row| {
+                            let id: String = row.get(0)?;
+                            Ok(id)
+                        })?;
+                        for row in rows {
+                            let id = row?;
+                            if !visited.contains(&id) {
+                                visited.insert(id.clone());
+                                next_frontier.push(id);
+                            }
+                        }
+
+                        // Get incoming links
+                        let mut stmt = conn.prepare(
+                            "SELECT source_media_id FROM markdown_links
+                             WHERE target_media_id = ?1"
+                        )?;
+                        let rows = stmt.query_map([node_id], |row| {
+                            let id: String = row.get(0)?;
+                            Ok(id)
+                        })?;
+                        for row in rows {
+                            let id = row?;
+                            if !visited.contains(&id) {
+                                visited.insert(id.clone());
+                                next_frontier.push(id);
+                            }
+                        }
+                    }
+
+                    frontier = next_frontier;
+                }
+
+                node_ids = visited;
+            } else {
+                // Get all markdown files with links (limit to 500 for performance)
+                let mut stmt = conn.prepare(
+                    "SELECT DISTINCT id FROM media_items
+                     WHERE media_type = 'markdown' AND deleted_at IS NULL
+                     LIMIT 500"
+                )?;
+                let rows = stmt.query_map([], |row| {
+                    let id: String = row.get(0)?;
+                    Ok(id)
+                })?;
+                for row in rows {
+                    node_ids.insert(row?);
+                }
+            }
+
+            // Build nodes with metadata
+            for node_id in &node_ids {
+                let mut stmt = conn.prepare(
+                    "SELECT id, COALESCE(title, file_name) as label, title, media_type
+                     FROM media_items WHERE id = ?1"
+                )?;
+                if let Ok((id, label, title, media_type)) = stmt.query_row([node_id], |row| {
+                    Ok((
+                        row.get::<_, String>(0)?,
+                        row.get::<_, String>(1)?,
+                        row.get::<_, Option<String>>(2)?,
+                        row.get::<_, String>(3)?,
+                    ))
+                }) {
+                    // Count outgoing links
+                    let link_count: i64 = conn.query_row(
+                        "SELECT COUNT(*) FROM markdown_links WHERE source_media_id = ?1",
+                        [&id],
+                        |row| row.get(0),
+                    )?;
+
+                    // Count incoming links
+                    let backlink_count: i64 = conn.query_row(
+                        "SELECT COUNT(*) FROM markdown_links WHERE target_media_id = ?1",
+                        [&id],
+                        |row| row.get(0),
+                    )?;
+
+                    nodes.push(crate::model::GraphNode {
+                        id: id.clone(),
+                        label,
+                        title,
+                        media_type,
+                        link_count: link_count as u32,
+                        backlink_count: backlink_count as u32,
+                    });
+                }
+            }
+
+            // Build edges
+            for node_id in &node_ids {
+                let mut stmt = conn.prepare(
+                    "SELECT source_media_id, target_media_id, link_type
+                     FROM markdown_links
+                     WHERE source_media_id = ?1 AND target_media_id IS NOT NULL"
+                )?;
+                let rows = stmt.query_map([node_id], |row| {
+                    let source: String = row.get(0)?;
+                    let target: String = row.get(1)?;
+                    let link_type_str: String = row.get(2)?;
+                    Ok((source, target, link_type_str))
+                })?;
+                for row in rows {
+                    let (source, target, link_type_str) = row?;
+                    if node_ids.contains(&target) {
+                        edges.push(crate::model::GraphEdge {
+                            source,
+                            target,
+                            link_type: link_type_str.parse().unwrap_or(crate::model::LinkType::Wikilink),
+                        });
+                    }
+                }
+            }
+
+            Ok::<_, rusqlite::Error>(crate::model::GraphData { nodes, edges })
+        })
+        .await
+        .map_err(|e| PinakesError::Database(e.to_string()))??;
+
+        Ok(graph_data)
+    }
+
+    async fn resolve_links(&self) -> Result<u64> {
+        let conn = self.conn.clone();
+
+        let count = tokio::task::spawn_blocking(move || {
+            let conn = conn.lock().unwrap();
+
+            // Find unresolved links and try to resolve them
+            // Strategy 1: Exact path match
+            let updated1 = conn.execute(
+                "UPDATE markdown_links
+                 SET target_media_id = (
+                     SELECT id FROM media_items
+                     WHERE path = markdown_links.target_path
+                     AND deleted_at IS NULL
+                     LIMIT 1
+                 )
+                 WHERE target_media_id IS NULL
+                 AND EXISTS (
+                     SELECT 1 FROM media_items
+                     WHERE path = markdown_links.target_path
+                     AND deleted_at IS NULL
+                 )",
+                [],
+            )?;
+
+            // Strategy 2: Filename match (Obsidian-style)
+            // Match target_path to file_name (with or without .md extension)
+            let updated2 = conn.execute(
+                "UPDATE markdown_links
+                 SET target_media_id = (
+                     SELECT id FROM media_items
+                     WHERE (file_name = markdown_links.target_path
+                            OR file_name = markdown_links.target_path || '.md'
+                            OR REPLACE(file_name, '.md', '') = markdown_links.target_path)
+                     AND deleted_at IS NULL
+                     LIMIT 1
+                 )
+                 WHERE target_media_id IS NULL
+                 AND EXISTS (
+                     SELECT 1 FROM media_items
+                     WHERE (file_name = markdown_links.target_path
+                            OR file_name = markdown_links.target_path || '.md'
+                            OR REPLACE(file_name, '.md', '') = markdown_links.target_path)
+                     AND deleted_at IS NULL
+                 )",
+                [],
+            )?;
+
+            Ok::<_, rusqlite::Error>((updated1 + updated2) as u64)
+        })
+        .await
+        .map_err(|e| PinakesError::Database(e.to_string()))??;
+
+        Ok(count)
+    }
+
+    async fn mark_links_extracted(&self, media_id: MediaId) -> Result<()> {
+        let conn = self.conn.clone();
+        let media_id_str = media_id.0.to_string();
+        let now = chrono::Utc::now().to_rfc3339();
+
+        tokio::task::spawn_blocking(move || {
+            let conn = conn.lock().unwrap();
+            conn.execute(
+                "UPDATE media_items SET links_extracted_at = ?1 WHERE id = ?2",
+                params![now, media_id_str],
+            )?;
+            Ok::<_, rusqlite::Error>(())
+        })
+        .await
+        .map_err(|e| PinakesError::Database(e.to_string()))??;
+
+        Ok(())
+    }
+
+    async fn count_unresolved_links(&self) -> Result<u64> {
+        let conn = self.conn.clone();
+
+        let count = tokio::task::spawn_blocking(move || {
+            let conn = conn.lock().unwrap();
+            let count: i64 = conn.query_row(
+                "SELECT COUNT(*) FROM markdown_links WHERE target_media_id IS NULL",
+                [],
+                |row| row.get(0),
+            )?;
+            Ok::<_, rusqlite::Error>(count as u64)
+        })
+        .await
+        .map_err(|e| PinakesError::Database(e.to_string()))??;
+
+        Ok(count)
+    }
+}
+
+// Helper function to parse a markdown link row
+fn row_to_markdown_link(row: &Row) -> rusqlite::Result<crate::model::MarkdownLink> {
+    let id_str: String = row.get(0)?;
+    let source_id_str: String = row.get(1)?;
+    let target_path: String = row.get(2)?;
+    let target_id: Option<String> = row.get(3)?;
+    let link_type_str: String = row.get(4)?;
+    let link_text: Option<String> = row.get(5)?;
+    let line_number: Option<i32> = row.get(6)?;
+    let context: Option<String> = row.get(7)?;
+    let created_at_str: String = row.get(8)?;
+
+    Ok(crate::model::MarkdownLink {
+        id: parse_uuid(&id_str)?,
+        source_media_id: MediaId(parse_uuid(&source_id_str)?),
+        target_path,
+        target_media_id: target_id
+            .and_then(|s| Uuid::parse_str(&s).ok())
+            .map(MediaId),
+        link_type: link_type_str
+            .parse()
+            .unwrap_or(crate::model::LinkType::Wikilink),
+        link_text,
+        line_number,
+        context,
+        created_at: parse_datetime(&created_at_str),
+    })
 }
 
 // Helper function to parse a share row

crates/pinakes-core/src/upload.rs

@@ -98,6 +98,7 @@ pub async fn process_upload<R: AsyncRead + Unpin>(
         created_at: now,
         updated_at: now,
         deleted_at: None,
+        links_extracted_at: None,
     };
 
     // Store the media item

crates/pinakes-core/tests/common/mod.rs

@@ -57,6 +57,7 @@ pub fn make_test_media(hash: &str) -> MediaItem {
         created_at: now,
         updated_at: now,
         deleted_at: None,
+        links_extracted_at: None,
     }
 }
 
@@ -93,5 +94,6 @@ pub fn create_test_media_item(path: PathBuf, hash: &str) -> MediaItem {
         created_at: now,
         updated_at: now,
         deleted_at: None,
+        links_extracted_at: None,
     }
 }

crates/pinakes-core/tests/integration.rs

@@ -47,6 +47,7 @@ async fn test_media_crud() {
         created_at: now,
         updated_at: now,
         deleted_at: None,
+        links_extracted_at: None,
     };
 
     // Insert
@@ -138,6 +139,7 @@ async fn test_tags() {
         created_at: now,
         updated_at: now,
         deleted_at: None,
+        links_extracted_at: None,
     };
     storage.insert_media(&item).await.unwrap();
     storage.tag_media(id, parent.id).await.unwrap();
@@ -203,6 +205,7 @@ async fn test_collections() {
         created_at: now,
         updated_at: now,
         deleted_at: None,
+        links_extracted_at: None,
     };
     storage.insert_media(&item).await.unwrap();
 
@@ -263,6 +266,7 @@ async fn test_custom_fields() {
         created_at: now,
         updated_at: now,
         deleted_at: None,
+        links_extracted_at: None,
     };
     storage.insert_media(&item).await.unwrap();
 
@@ -342,6 +346,7 @@ async fn test_search() {
             created_at: now,
             updated_at: now,
             deleted_at: None,
+            links_extracted_at: None,
         };
         storage.insert_media(&item).await.unwrap();
     }
@@ -486,6 +491,7 @@ async fn test_library_statistics_with_data() {
         created_at: now,
         updated_at: now,
         deleted_at: None,
+        links_extracted_at: None,
     };
     storage.insert_media(&item).await.unwrap();