feat: automatically repopulate if embedding model is switched

Signed-off-by: Ettore Di Giacinto <[email protected]>

Author: mudler

rag/collection.go

@@ -26,7 +26,7 @@ func NewPersistentChromeCollection(llmClient *openai.Client, collectionName, dbP
 		filepath.Join(dbPath, fmt.Sprintf("%s%s.json", collectionPrefix, collectionName)),
 		filePath,
 		chromemDB,
-		maxChunkSize)
+		maxChunkSize, llmClient, embeddingModel)
 	if err != nil {
 		xlog.Error("Failed to create PersistentKB", err)
 		os.Exit(1)
@@ -44,7 +44,7 @@ func NewPersistentLocalAICollection(llmClient *openai.Client, apiURL, apiKey, co
 		filepath.Join(dbPath, fmt.Sprintf("%s%s.json", collectionPrefix, collectionName)),
 		filePath,
 		ragDB,
-		maxChunkSize)
+		maxChunkSize, llmClient, embeddingModel)
 	if err != nil {
 		xlog.Error("Failed to create PersistentKB", err)
 		os.Exit(1)

rag/engine.go

@@ -8,6 +8,7 @@ import (
 type Engine interface {
 	Store(s string, metadata map[string]string) (engine.Result, error)
 	StoreDocuments(s []string, metadata map[string]string) ([]engine.Result, error)
+	GetEmbeddingDimensions() (int, error)
 	Reset() error
 	Search(s string, similarEntries int) ([]types.Result, error)
 	Count() int

rag/engine/chromem.go

@@ -39,6 +39,11 @@ func NewChromemDBCollection(collection, path string, openaiClient *openai.Client
 	}
 	chromem.collection = c
 
+	count := c.Count()
+	if count > 0 {
+		chromem.index = count + 1
+	}
+
 	return chromem, nil
 }
 
@@ -59,6 +64,20 @@ func (c *ChromemDB) Reset() error {
 	return nil
 }
 
+func (c *ChromemDB) GetEmbeddingDimensions() (int, error) {
+	count := c.collection.Count()
+	if count == 0 {
+		return 0, fmt.Errorf("no documents in collection")
+	}
+
+	doc, err := c.collection.GetByID(context.Background(), fmt.Sprint(count))
+	if err != nil {
+		return 0, fmt.Errorf("error getting document: %v", err)
+	}
+
+	return len(doc.Embedding), nil
+}
+
 func (c *ChromemDB) embedding() chromem.EmbeddingFunc {
 	return chromem.EmbeddingFunc(
 		func(ctx context.Context, text string) ([]float32, error) {

rag/engine/localai.go

@@ -31,6 +31,10 @@ func (db *LocalAIRAGDB) Count() int {
 	return 0
 }
 
+func (db *LocalAIRAGDB) GetEmbeddingDimensions() (int, error) {
+	return 0, fmt.Errorf("not implemented")
+}
+
 func (db *LocalAIRAGDB) StoreDocuments(s []string, metadata map[string]string) ([]Result, error) {
 	results := []Result{}
 	for _, content := range s {

rag/persistency.go

@@ -2,6 +2,7 @@ package rag
 
 import (
 	"bytes"
+	"context"
 	"encoding/json"
 	"fmt"
 	"io"
@@ -15,6 +16,7 @@ import (
 	"github.com/mudler/localrecall/pkg/xlog"
 	"github.com/mudler/localrecall/rag/engine"
 	"github.com/mudler/localrecall/rag/types"
+	"github.com/sashabaranov/go-openai"
 )
 
 // CollectionState represents the persistent state of a collection
@@ -55,7 +57,7 @@ func loadDB(path string) (*CollectionState, error) {
 	return state, nil
 }
 
-func NewPersistentCollectionKB(stateFile, assetDir string, store Engine, maxChunkSize int) (*PersistentKB, error) {
+func NewPersistentCollectionKB(stateFile, assetDir string, store Engine, maxChunkSize int, llmClient *openai.Client, embeddingModel string) (*PersistentKB, error) {
 	// if file exists, try to load an existing state
 	// if file does not exist, create a new state
 	if err := os.MkdirAll(assetDir, 0755); err != nil {
@@ -89,6 +91,24 @@ func NewPersistentCollectionKB(stateFile, assetDir string, store Engine, maxChun
 		index:        state.Index,
 	}
 
+	// TODO: Automatically repopulate if embeddings dimensions are mismatching.
+	// To check if dimensions are mismatching, we can check the number of dimensions of the first embedding in the index if is the same as the
+	// dimension that the embedding model returns.
+	resp, err := llmClient.CreateEmbeddings(context.Background(),
+		openai.EmbeddingRequestStrings{
+			Input: []string{"test"},
+			Model: openai.EmbeddingModel(embeddingModel),
+		},
+	)
+	if err == nil && len(resp.Data) > 0 {
+		embedding := resp.Data[0].Embedding
+		embeddingDimensions, err := db.Engine.GetEmbeddingDimensions()
+		if err == nil && len(embedding) != embeddingDimensions {
+			xlog.Info("Embedding dimensions mismatch, repopulating", "embeddingDimensions", embeddingDimensions, "embedding", embedding)
+			return db, db.Repopulate()
+		}
+	}
+
 	return db, nil
 }