Treat index id as auto-generated field (#27)

* Treat index id as auto-generated field

* fix Can't create index: KeyError: 'embedding' #28

* udpate docs

* Update requirement

* Use lighter default local embedding model

---------

Co-authored-by: ian <ian@cinnamon.is>

libs/ktem/ktem/embeddings/manager.py

@@ -196,4 +196,4 @@ class EmbeddingManager:
         return {vendor.__qualname__: vendor for vendor in self._vendors}
 
 
-embeddings = EmbeddingManager()
+embedding_models_manager = EmbeddingManager()

libs/ktem/ktem/embeddings/ui.py

@@ -5,7 +5,7 @@ import pandas as pd
 import yaml
 from ktem.app import BasePage
 
-from .manager import embeddings
+from .manager import embedding_models_manager
 
 
 def format_description(cls):
@@ -118,12 +118,12 @@ class EmbeddingManagement(BasePage):
             outputs=[self.emb_list],
         )
         self._app.app.load(
-            lambda: gr.update(choices=list(embeddings.vendors().keys())),
+            lambda: gr.update(choices=list(embedding_models_manager.vendors().keys())),
             outputs=[self.emb_choices],
         )
 
     def on_emb_vendor_change(self, vendor):
-        vendor = embeddings.vendors()[vendor]
+        vendor = embedding_models_manager.vendors()[vendor]
 
         required: dict = {}
         desc = vendor.describe()
@@ -224,12 +224,12 @@ class EmbeddingManagement(BasePage):
         try:
             spec = yaml.safe_load(spec)
             spec["__type__"] = (
-                embeddings.vendors()[choices].__module__
+                embedding_models_manager.vendors()[choices].__module__
                 + "."
-                + embeddings.vendors()[choices].__qualname__
+                + embedding_models_manager.vendors()[choices].__qualname__
             )
 
-            embeddings.add(name, spec=spec, default=default)
+            embedding_models_manager.add(name, spec=spec, default=default)
             gr.Info(f'Create Embedding model "{name}" successfully')
         except Exception as e:
             raise gr.Error(f"Failed to create Embedding model {name}: {e}")
@@ -237,7 +237,7 @@ class EmbeddingManagement(BasePage):
     def list_embeddings(self):
         """List the Embedding models"""
         items = []
-        for item in embeddings.info().values():
+        for item in embedding_models_manager.info().values():
             record = {}
             record["name"] = item["name"]
             record["vendor"] = item["spec"].get("__type__", "-").split(".")[-1]
@@ -280,9 +280,9 @@ class EmbeddingManagement(BasePage):
             btn_delete_yes = gr.update(visible=False)
             btn_delete_no = gr.update(visible=False)
 
-            info = deepcopy(embeddings.info()[selected_emb_name])
+            info = deepcopy(embedding_models_manager.info()[selected_emb_name])
             vendor_str = info["spec"].pop("__type__", "-").split(".")[-1]
-            vendor = embeddings.vendors()[vendor_str]
+            vendor = embedding_models_manager.vendors()[vendor_str]
 
             edit_spec = yaml.dump(info["spec"])
             edit_spec_desc = format_description(vendor)
@@ -309,15 +309,19 @@ class EmbeddingManagement(BasePage):
     def save_emb(self, selected_emb_name, default, spec):
         try:
             spec = yaml.safe_load(spec)
-            spec["__type__"] = embeddings.info()[selected_emb_name]["spec"]["__type__"]
-            embeddings.update(selected_emb_name, spec=spec, default=default)
+            spec["__type__"] = embedding_models_manager.info()[selected_emb_name][
+                "spec"
+            ]["__type__"]
+            embedding_models_manager.update(
+                selected_emb_name, spec=spec, default=default
+            )
             gr.Info(f'Save Embedding model "{selected_emb_name}" successfully')
         except Exception as e:
             gr.Error(f'Failed to save Embedding model "{selected_emb_name}": {e}')
 
     def delete_emb(self, selected_emb_name):
         try:
-            embeddings.delete(selected_emb_name)
+            embedding_models_manager.delete(selected_emb_name)
         except Exception as e:
             gr.Error(f'Failed to delete Embedding model "{selected_emb_name}": {e}')
             return selected_emb_name

libs/ktem/ktem/index/file/index.py

@@ -293,10 +293,10 @@ class FileIndex(BaseIndex):
 
     @classmethod
     def get_admin_settings(cls):
-        from ktem.embeddings.manager import embeddings
+        from ktem.embeddings.manager import embedding_models_manager
 
-        embedding_default = embeddings.get_default_name()
-        embedding_choices = list(embeddings.options().keys())
+        embedding_default = embedding_models_manager.get_default_name()
+        embedding_choices = list(embedding_models_manager.options().keys())
 
         return {
             "embedding": {

libs/ktem/ktem/index/file/pipelines.py

@@ -12,7 +12,7 @@ from typing import Optional
 import gradio as gr
 from ktem.components import filestorage_path
 from ktem.db.models import engine
-from ktem.embeddings.manager import embeddings
+from ktem.embeddings.manager import embedding_models_manager
 from llama_index.vector_stores import (
     FilterCondition,
     FilterOperator,
@@ -225,7 +225,9 @@ class DocumentRetrievalPipeline(BaseFileIndexRetriever):
         if not user_settings["use_reranking"]:
             retriever.reranker = None  # type: ignore
 
-        retriever.vector_retrieval.embedding = embeddings[index_settings["embedding"]]
+        retriever.vector_retrieval.embedding = embedding_models_manager[
+            index_settings.get("embedding", embedding_models_manager.get_default_name())
+        ]
         kwargs = {
             ".top_k": int(user_settings["num_retrieval"]),
             ".mmr": user_settings["mmr"],
@@ -436,7 +438,9 @@ class IndexDocumentPipeline(BaseFileIndexIndexing):
         if chunk_overlap:
             obj.file_ingestor.text_splitter.chunk_overlap = chunk_overlap
 
-        obj.indexing_vector_pipeline.embedding = embeddings[index_settings["embedding"]]
+        obj.indexing_vector_pipeline.embedding = embedding_models_manager[
+            index_settings.get("embedding", embedding_models_manager.get_default_name())
+        ]
 
         return obj
 

libs/ktem/ktem/index/manager.py

@@ -58,6 +58,10 @@ class IndexManager:
                 index_cls = import_dotted_string(index_type, safe=False)
                 index = index_cls(app=self._app, id=entry.id, name=name, config=config)
                 index.on_create()
+
+                # update the entry
+                entry.config = index.config
+                sess.commit()
             except Exception as e:
                 sess.delete(entry)
                 sess.commit()
@@ -177,7 +181,7 @@ class IndexManager:
         self.load_index_types()
 
         for index in settings.KH_INDICES:
-            if not self.exists(index["id"]):
+            if not self.exists(name=index["name"]):
                 self.build_index(**index)
 
         with Session(engine) as sess: