refactor: remove chromem-go — pgvector on PostgreSQL replaces it

Remove standalone vector DB (chromem-go) in favor of PostgreSQL pgvector
extension which runs alongside AGE on the same database instance.

Removed:
- chromem-go dependency
- VectorStore type and initialization
- --embed flag from ccg build
- --semantic flag from ccg search

Vector search can be added later via pgvector on the same PostgreSQL
that already runs AGE, without any additional infrastructure.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: tae2089

cmd/ccg/main.go

@@ -6,7 +6,6 @@ import (
 	"os"
 
 	"github.com/mark3labs/mcp-go/server"
-	chromem "github.com/philippgille/chromem-go"
 
 	"gorm.io/driver/mysql"
 	"gorm.io/driver/postgres"
@@ -62,26 +61,13 @@ func main() {
 
 	syncer := incremental.New(st, walkers[".go"])
 
-	// Initialize vector DB for semantic search (persisted to disk)
-	var vectorStore *cli.VectorStore
-	if os.Getenv("OPENAI_API_KEY") != "" {
-		vdb := chromem.NewDB()
-		collection, err := vdb.GetOrCreateCollection("nodes", nil, nil)
-		if err != nil {
-			slog.Warn("vector DB init failed", "error", err)
-		} else {
-			vectorStore = &cli.VectorStore{DB: vdb, Collection: collection}
-		}
-	}
-
 	deps := &cli.Deps{
 		Logger:        logger,
 		DB:            db,
 		Store:         st,
 		SearchBackend: sb,
 		Walkers:       walkers,
 		Syncer:        syncer,
-		VectorDB:      vectorStore,
 		ServeFunc: func(cfg cli.ServeConfig) error {
 			return runServe(cfg.DBDriver, cfg.DSN)
 		},

go.mod

@@ -5,7 +5,6 @@ go 1.25.5
 require (
 	github.com/lib/pq v1.12.3
 	github.com/mark3labs/mcp-go v0.47.1
-	github.com/philippgille/chromem-go v0.7.0
 	github.com/smacker/go-tree-sitter v0.0.0-20240827094217-dd81d9e9be82
 	github.com/spf13/cobra v1.10.2
 	gorm.io/driver/mysql v1.6.0

go.sum

@@ -38,8 +38,6 @@ github.com/mark3labs/mcp-go v0.47.1 h1:A9sJJ20mscl/ssLYHjodfaoBmq6uuhMG7pAPNYaQy
 github.com/mark3labs/mcp-go v0.47.1/go.mod h1:JKTC7R2LLVagkEWK7Kwu7DbmA6iIvnNAod6yrHiQMag=
 github.com/mattn/go-sqlite3 v1.14.22 h1:2gZY6PC6kBnID23Tichd1K+Z0oS6nE/XwU+Vz/5o4kU=
 github.com/mattn/go-sqlite3 v1.14.22/go.mod h1:Uh1q+B4BYcTPb+yiD3kU8Ct7aC0hY9fxUwlHK0RXw+Y=
-github.com/philippgille/chromem-go v0.7.0 h1:4jfvfyKymjKNfGxBUhHUcj1kp7B17NL/I1P+vGh1RvY=
-github.com/philippgille/chromem-go v0.7.0/go.mod h1:hTd+wGEm/fFPQl7ilfCwQXkgEUxceYh86iIdoKMolPo=
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
 github.com/rogpeppe/go-internal v1.9.0 h1:73kH8U+JUqXU8lRuOHeVHaa/SZPifC7BkcraZVejAe8=

internal/cli/build.go

@@ -7,7 +7,6 @@ import (
 	"path/filepath"
 	"strings"
 
-	chromem "github.com/philippgille/chromem-go"
 	"github.com/spf13/cobra"
 
 	"github.com/imtaebin/code-context-graph/internal/model"
@@ -21,7 +20,6 @@ var skipDirs = map[string]bool{
 }
 
 func newBuildCmd(deps *Deps) *cobra.Command {
-	var embed bool
 	var syncGraph bool
 
 	cmd := &cobra.Command{
@@ -175,27 +173,6 @@ func newBuildCmd(deps *Deps) *cobra.Command {
 				}
 			}
 
-			// Build vector embeddings if --embed flag is set
-			if embed && deps.VectorDB != nil {
-				deps.Logger.Info("building vector embeddings")
-				for _, n := range indexNodes {
-					err := deps.VectorDB.Collection.AddDocument(ctx, chromem.Document{
-						ID:      fmt.Sprintf("%d", n.ID),
-						Content: buildContent(n),
-						Metadata: map[string]string{
-							"qualified_name": n.QualifiedName,
-							"kind":           string(n.Kind),
-							"file_path":      n.FilePath,
-							"language":       n.Language,
-						},
-					})
-					if err != nil {
-						deps.Logger.Warn("embed failed", "node", n.QualifiedName, "error", err)
-					}
-				}
-				deps.Logger.Info("vector embeddings built", "documents", len(indexNodes))
-			}
-
 			// Sync to Apache AGE graph if --graph flag is set
 			if syncGraph && deps.AgeStore != nil {
 				deps.Logger.Info("syncing to AGE graph")
@@ -222,7 +199,6 @@ func newBuildCmd(deps *Deps) *cobra.Command {
 		},
 	}
 
-	cmd.Flags().BoolVar(&embed, "embed", false, "Build vector embeddings for semantic search (requires OPENAI_API_KEY)")
 	cmd.Flags().BoolVar(&syncGraph, "graph", false, "Sync graph to Apache AGE (requires AGE_DSN)")
 
 	return cmd

internal/cli/root.go

@@ -5,7 +5,6 @@ import (
 	"log/slog"
 	"os"
 
-	chromem "github.com/philippgille/chromem-go"
 	"github.com/spf13/cobra"
 	"gorm.io/gorm"
 
@@ -25,16 +24,9 @@ type Deps struct {
 	Walkers       map[string]*treesitter.Walker
 	Syncer        *incremental.Syncer
 	ServeFunc     func(cfg ServeConfig) error
-	VectorDB      *VectorStore
 	AgeStore      *agestore.Store
 }
 
-// VectorStore wraps chromem-go for semantic search.
-type VectorStore struct {
-	DB         *chromem.DB
-	Collection *chromem.Collection
-}
-
 // NewRootCmd creates the root cobra command with all subcommands attached.
 // If deps is nil, a minimal Deps with a default logger is created.
 func NewRootCmd(deps *Deps) *cobra.Command {

internal/cli/search.go

@@ -9,45 +9,22 @@ import (
 
 func newSearchCmd(deps *Deps) *cobra.Command {
 	var limit int
-	var semantic bool
 
 	cmd := &cobra.Command{
 		Use:   "search <query>",
-		Short: "Full-text or semantic search for code nodes",
+		Short: "Full-text search for code nodes",
 		Args:  cobra.ExactArgs(1),
 		RunE: func(cmd *cobra.Command, args []string) error {
 			query := args[0]
 			ctx := context.Background()
-			out := stdout(cmd)
-
-			// Semantic search via vector embeddings
-			if semantic {
-				if deps.VectorDB == nil || deps.VectorDB.Collection == nil {
-					return fmt.Errorf("vector embeddings not available — run 'ccg build --embed' first")
-				}
-				results, err := deps.VectorDB.Collection.Query(ctx, query, limit, nil, nil)
-				if err != nil {
-					return fmt.Errorf("semantic search: %w", err)
-				}
-				if len(results) == 0 {
-					fmt.Fprintln(out, "No results")
-					return nil
-				}
-				for _, r := range results {
-					qn := r.Metadata["qualified_name"]
-					kind := r.Metadata["kind"]
-					fp := r.Metadata["file_path"]
-					fmt.Fprintf(out, "%s\t%s\t%s\t(score: %.3f)\n", qn, kind, fp, r.Similarity)
-				}
-				return nil
-			}
 
-			// FTS keyword search
 			nodes, err := deps.SearchBackend.Query(ctx, deps.DB, query, limit)
 			if err != nil {
 				return fmt.Errorf("search: %w", err)
 			}
 
+			out := stdout(cmd)
+
 			if len(nodes) == 0 {
 				fmt.Fprintln(out, "No results")
 				return nil
@@ -62,7 +39,6 @@ func newSearchCmd(deps *Deps) *cobra.Command {
 	}
 
 	cmd.Flags().IntVar(&limit, "limit", 10, "Maximum number of results")
-	cmd.Flags().BoolVar(&semantic, "semantic", false, "Use semantic search via vector embeddings")
 
 	return cmd
 }