fix(vectorstore): walk and read the brain through brainfs (#157, slice 3) (#166)

vectorstore.Sync walks and reads via brainfs (regular files only, handle-checked), so symlinked notes are no longer embedded and their old chunks are purged. brainfs.WalkRegular requires every component of its root to be a real directory (fixes in-brain root aliasing in Sync and search Wing/Wing+Hall). Walk errors, including a symlinked scan root, abort Sync before the delete pass. brainfs gains Lstat. Directory-swap race recorded as accepted residual (#167).

Refs #157. Adversarial review: BLOCKED, BLOCKED, PASS.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Mathias <mthbqv@gmail.com>

ingestion/internal/vectorstore/sync.go

@@ -2,12 +2,15 @@ package vectorstore
 
 import (
 	"context"
+	"errors"
 	"fmt"
+	"io/fs"
 	"log/slog"
-	"os"
-	"path/filepath"
+	"path"
 	"strings"
 	"time"
+
+	"github.com/mathiasbq/hyperguild/ingestion/internal/brainfs"
 )
 
 // Embedder produces dense vectors. The embed package's Client satisfies
@@ -90,31 +93,38 @@ func Sync(ctx context.Context, brainDir string, store Store, embedder Embedder)
 	}
 	seenParents := make(map[string]struct{})
 
+	// Through the brain root, regular files only (hyperguild#157): the brain
+	// is a Syncthing folder, and os.ReadFile followed a synced symlink into
+	// the embeddings and blocked on a FIFO.
+	bfs, err := brainfs.Open(brainDir)
+	if err != nil {
+		return res, fmt.Errorf("open brain: %w", err)
+	}
+	defer func() { _ = bfs.Close() }()
+
 	for _, sub := range scanDirs {
-		root := filepath.Join(brainDir, sub)
-		if _, err := os.Stat(root); os.IsNotExist(err) {
+		if _, err := bfs.Lstat(sub); errors.Is(err, fs.ErrNotExist) {
 			continue
 		}
-
-		err = filepath.WalkDir(root, func(path string, d os.DirEntry, err error) error {
-			if err != nil {
-				return err
+		// Any other walk error, including a scan root that is a symlink,
+		// aborts the run below.
+		var walkErr error
+		onErr := func(rel string, err error) {
+			if walkErr == nil {
+				walkErr = fmt.Errorf("%s: %w", rel, err)
 			}
-			if d.IsDir() || !strings.HasSuffix(path, ".md") || d.Name() == "_index.md" {
+		}
+		err = bfs.WalkRegular(sub, func(relSlash string) error {
+			if !strings.HasSuffix(relSlash, ".md") || path.Base(relSlash) == "_index.md" {
 				return nil
 			}
-			rel, err := filepath.Rel(brainDir, path)
-			if err != nil {
-				return err
-			}
-			relSlash := filepath.ToSlash(rel)
 			seenParents[relSlash] = struct{}{}
 
 			if ps, ok := parents[relSlash]; ok {
 				// File already has chunks in the store. Re-embed only when
 				// the file has been edited since the oldest chunk was
 				// written. Tolerate clock skew with a sub-second grace.
-				info, statErr := d.Info()
+				info, statErr := bfs.Lstat(relSlash)
 				if statErr != nil {
 					res.Errors = append(res.Errors, fmt.Errorf("stat %s: %w", relSlash, statErr))
 					return nil
@@ -132,7 +142,7 @@ func Sync(ctx context.Context, brainDir string, store Store, embedder Embedder)
 				}
 			}
 
-			content, readErr := os.ReadFile(path)
+			content, readErr := bfs.ReadRegular(relSlash)
 			if readErr != nil {
 				res.Errors = append(res.Errors, fmt.Errorf("read %s: %w", relSlash, readErr))
 				return nil
@@ -151,7 +161,12 @@ func Sync(ctx context.Context, brainDir string, store Store, embedder Embedder)
 				res.Added++
 			}
 			return nil
-		})
+		}, onErr)
+		if err == nil {
+			err = walkErr
+		}
+		// Abort before the delete pass: an incomplete walk would leave
+		// seenParents short and delete chunks of files that still exist.
 		if err != nil {
 			return res, fmt.Errorf("walk %s: %w", sub, err)
 		}