update

flowsettings.py

@@ -4,39 +4,43 @@ from inspect import currentframe, getframeinfo
 from pathlib import Path
 
 from decouple import config
+from ktem.utils.lang import SUPPORTED_LANGUAGE_MAP
 from theflow.settings.default import *  # noqa
 
 cur_frame = currentframe()
 if cur_frame is None:
     raise ValueError("Cannot get the current frame.")
 this_file = getframeinfo(cur_frame).filename
-this_dir = Path(this_file).parent
+this_dir = Path("/home/ubuntu/lib-knowledgehub/kotaemon/ktem_app_data")
 
 # change this if your app use a different name
 KH_PACKAGE_NAME = "kotaemon_app"
 
-KH_APP_VERSION = config("KH_APP_VERSION", "local")
+KH_APP_VERSION = config("KH_APP_VERSION", None)
 if not KH_APP_VERSION:
     try:
         # Caution: This might produce the wrong version
         # https://stackoverflow.com/a/59533071
         KH_APP_VERSION = version(KH_PACKAGE_NAME)
-    except Exception as e:
-        print(f"Failed to get app version: {e}")
+    except Exception:
+        KH_APP_VERSION = "local"
+
+KH_GRADIO_SHARE = config("KH_GRADIO_SHARE", default=False, cast=bool)
+KH_ENABLE_FIRST_SETUP = config("KH_ENABLE_FIRST_SETUP", default=True, cast=bool)
+KH_DEMO_MODE = config("KH_DEMO_MODE", default=False, cast=bool)
+KH_OLLAMA_URL = config("KH_OLLAMA_URL", default="http://localhost:11434/v1/")
 
 # App can be ran from anywhere and it's not trivial to decide where to store app data.
 # So let's use the same directory as the flowsetting.py file.
-# KH_APP_DATA_DIR = this_dir / "ktem_app_data"
-
-# override app data dir to fit preview data
-KH_APP_DATA_DIR = Path("/home/ubuntu/lib-knowledgehub/kotaemon/ktem_app_data")
+KH_APP_DATA_DIR = this_dir / "ktem_app_data"
+KH_APP_DATA_EXISTS = KH_APP_DATA_DIR.exists()
 KH_APP_DATA_DIR.mkdir(parents=True, exist_ok=True)
 
 # User data directory
 KH_USER_DATA_DIR = KH_APP_DATA_DIR / "user_data"
 KH_USER_DATA_DIR.mkdir(parents=True, exist_ok=True)
 
-# markdowm output directory
+# markdown output directory
 KH_MARKDOWN_OUTPUT_DIR = KH_APP_DATA_DIR / "markdown_cache_dir"
 KH_MARKDOWN_OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
 
@@ -62,7 +66,14 @@ os.environ["HF_HUB_CACHE"] = str(KH_APP_DATA_DIR / "huggingface")
 KH_DOC_DIR = this_dir / "docs"
 
 KH_MODE = "dev"
-KH_FEATURE_USER_MANAGEMENT = False
+KH_SSO_ENABLED = config("KH_SSO_ENABLED", default=False, cast=bool)
+
+KH_FEATURE_CHAT_SUGGESTION = config(
+    "KH_FEATURE_CHAT_SUGGESTION", default=False, cast=bool
+)
+KH_FEATURE_USER_MANAGEMENT = config(
+    "KH_FEATURE_USER_MANAGEMENT", default=True, cast=bool
+)
 KH_USER_CAN_SEE_PUBLIC = None
 KH_FEATURE_USER_MANAGEMENT_ADMIN = str(
     config("KH_FEATURE_USER_MANAGEMENT_ADMIN", default="admin")
@@ -71,8 +82,12 @@ KH_FEATURE_USER_MANAGEMENT_PASSWORD = str(
     config("KH_FEATURE_USER_MANAGEMENT_PASSWORD", default="admin")
 )
 KH_ENABLE_ALEMBIC = False
-KH_DATABASE = f"sqlite:///file:{KH_USER_DATA_DIR / 'sql.db?mode=ro&uri=true'}"
+KH_DATABASE = f"sqlite:///{KH_USER_DATA_DIR / 'sql.db'}"
 KH_FILESTORAGE_PATH = str(KH_USER_DATA_DIR / "files")
+KH_WEB_SEARCH_BACKEND = (
+    "kotaemon.indices.retrievers.tavily_web_search.WebSearch"
+    # "kotaemon.indices.retrievers.jina_web_search.WebSearch"
+)
 
 KH_DOCSTORE = {
     # "__type__": "kotaemon.storages.ElasticsearchDocumentStore",
@@ -83,10 +98,13 @@ KH_DOCSTORE = {
 KH_VECTORSTORE = {
     # "__type__": "kotaemon.storages.LanceDBVectorStore",
     "__type__": "kotaemon.storages.ChromaVectorStore",
+    # "__type__": "kotaemon.storages.MilvusVectorStore",
+    # "__type__": "kotaemon.storages.QdrantVectorStore",
     "path": str(KH_USER_DATA_DIR / "vectorstore"),
 }
 KH_LLMS = {}
 KH_EMBEDDINGS = {}
+KH_RERANKINGS = {}
 
 # populate options from config
 if config("AZURE_OPENAI_API_KEY", default="") and config(
@@ -122,52 +140,68 @@ if config("AZURE_OPENAI_API_KEY", default="") and config(
             "default": False,
         }
 
-if config("OPENAI_API_KEY", default=""):
+OPENAI_DEFAULT = "<YOUR_OPENAI_KEY>"
+OPENAI_API_KEY = config("OPENAI_API_KEY", default=OPENAI_DEFAULT)
+GOOGLE_API_KEY = config("GOOGLE_API_KEY", default="your-key")
+IS_OPENAI_DEFAULT = len(OPENAI_API_KEY) > 0 and OPENAI_API_KEY != OPENAI_DEFAULT
+
+if OPENAI_API_KEY:
     KH_LLMS["openai"] = {
         "spec": {
             "__type__": "kotaemon.llms.ChatOpenAI",
             "temperature": 0,
             "base_url": config("OPENAI_API_BASE", default="")
             or "https://api.openai.com/v1",
-            "api_key": config("OPENAI_API_KEY", default=""),
-            "model": config("OPENAI_CHAT_MODEL", default="gpt-3.5-turbo"),
+            "api_key": OPENAI_API_KEY,
+            "model": config("OPENAI_CHAT_MODEL", default="gpt-4o-mini"),
             "timeout": 20,
         },
-        "default": True,
+        "default": IS_OPENAI_DEFAULT,
     }
     KH_EMBEDDINGS["openai"] = {
         "spec": {
             "__type__": "kotaemon.embeddings.OpenAIEmbeddings",
             "base_url": config("OPENAI_API_BASE", default="https://api.openai.com/v1"),
-            "api_key": config("OPENAI_API_KEY", default=""),
+            "api_key": OPENAI_API_KEY,
             "model": config(
-                "OPENAI_EMBEDDINGS_MODEL", default="text-embedding-ada-002"
+                "OPENAI_EMBEDDINGS_MODEL", default="text-embedding-3-large"
             ),
             "timeout": 10,
             "context_length": 8191,
         },
-        "default": True,
+        "default": IS_OPENAI_DEFAULT,
     }
 
 if config("LOCAL_MODEL", default=""):
     KH_LLMS["ollama"] = {
         "spec": {
             "__type__": "kotaemon.llms.ChatOpenAI",
-            "base_url": "http://localhost:11434/v1/",
-            "model": config("LOCAL_MODEL", default="llama3.1:8b"),
+            "base_url": KH_OLLAMA_URL,
+            "model": config("LOCAL_MODEL", default="qwen2.5:7b"),
+            "api_key": "ollama",
+        },
+        "default": False,
+    }
+    KH_LLMS["ollama-long-context"] = {
+        "spec": {
+            "__type__": "kotaemon.llms.LCOllamaChat",
+            "base_url": KH_OLLAMA_URL.replace("v1/", ""),
+            "model": config("LOCAL_MODEL", default="qwen2.5:7b"),
+            "num_ctx": 8192,
         },
         "default": False,
     }
+
     KH_EMBEDDINGS["ollama"] = {
         "spec": {
             "__type__": "kotaemon.embeddings.OpenAIEmbeddings",
-            "base_url": "http://localhost:11434/v1/",
+            "base_url": KH_OLLAMA_URL,
             "model": config("LOCAL_MODEL_EMBEDDINGS", default="nomic-embed-text"),
+            "api_key": "ollama",
         },
         "default": False,
     }
-
-    KH_EMBEDDINGS["local-bge-en"] = {
+    KH_EMBEDDINGS["fast_embed"] = {
         "spec": {
             "__type__": "kotaemon.embeddings.FastEmbedEmbeddings",
             "model_name": "BAAI/bge-base-en-v1.5",
@@ -175,13 +209,84 @@ if config("LOCAL_MODEL", default=""):
         "default": False,
     }
 
+# additional LLM configurations
+KH_LLMS["claude"] = {
+    "spec": {
+        "__type__": "kotaemon.llms.chats.LCAnthropicChat",
+        "model_name": "claude-3-5-sonnet-20240620",
+        "api_key": "your-key",
+    },
+    "default": False,
+}
+KH_LLMS["google"] = {
+    "spec": {
+        "__type__": "kotaemon.llms.chats.LCGeminiChat",
+        "model_name": "gemini-1.5-flash",
+        "api_key": GOOGLE_API_KEY,
+    },
+    "default": not IS_OPENAI_DEFAULT,
+}
+KH_LLMS["groq"] = {
+    "spec": {
+        "__type__": "kotaemon.llms.ChatOpenAI",
+        "base_url": "https://api.groq.com/openai/v1",
+        "model": "llama-3.1-8b-instant",
+        "api_key": "your-key",
+    },
+    "default": False,
+}
+KH_LLMS["cohere"] = {
+    "spec": {
+        "__type__": "kotaemon.llms.chats.LCCohereChat",
+        "model_name": "command-r-plus-08-2024",
+        "api_key": config("COHERE_API_KEY", default="your-key"),
+    },
+    "default": False,
+}
+
+# additional embeddings configurations
+KH_EMBEDDINGS["cohere"] = {
+    "spec": {
+        "__type__": "kotaemon.embeddings.LCCohereEmbeddings",
+        "model": "embed-multilingual-v3.0",
+        "cohere_api_key": config("COHERE_API_KEY", default="your-key"),
+        "user_agent": "default",
+    },
+    "default": False,
+}
+KH_EMBEDDINGS["google"] = {
+    "spec": {
+        "__type__": "kotaemon.embeddings.LCGoogleEmbeddings",
+        "model": "models/text-embedding-004",
+        "google_api_key": GOOGLE_API_KEY,
+    },
+    "default": not IS_OPENAI_DEFAULT,
+}
+# KH_EMBEDDINGS["huggingface"] = {
+#     "spec": {
+#         "__type__": "kotaemon.embeddings.LCHuggingFaceEmbeddings",
+#         "model_name": "sentence-transformers/all-mpnet-base-v2",
+#     },
+#     "default": False,
+# }
+
+# default reranking models
+KH_RERANKINGS["cohere"] = {
+    "spec": {
+        "__type__": "kotaemon.rerankings.CohereReranking",
+        "model_name": "rerank-multilingual-v2.0",
+        "cohere_api_key": config("COHERE_API_KEY", default=""),
+    },
+    "default": True,
+}
+
 KH_REASONINGS = [
     "ktem.reasoning.simple.FullQAPipeline",
     "ktem.reasoning.simple.FullDecomposeQAPipeline",
     "ktem.reasoning.react.ReactAgentPipeline",
     "ktem.reasoning.rewoo.RewooAgentPipeline",
 ]
-KH_REASONINGS_USE_MULTIMODAL = False
+KH_REASONINGS_USE_MULTIMODAL = config("USE_MULTIMODAL", default=False, cast=bool)
 KH_VLM_ENDPOINT = "{0}/openai/deployments/{1}/chat/completions?api-version={2}".format(
     config("AZURE_OPENAI_ENDPOINT", default=""),
     config("OPENAI_VISION_DEPLOYMENT_NAME", default="gpt-4o"),
@@ -202,7 +307,7 @@ SETTINGS_REASONING = {
     "lang": {
         "name": "Language",
         "value": "en",
-        "choices": [("English", "en"), ("Japanese", "ja"), ("Vietnamese", "vi")],
+        "choices": [(lang, code) for code, lang in SUPPORTED_LANGUAGE_MAP.items()],
         "component": "dropdown",
     },
     "max_context_length": {
@@ -212,32 +317,51 @@ SETTINGS_REASONING = {
     },
 }
 
+USE_NANO_GRAPHRAG = config("USE_NANO_GRAPHRAG", default=False, cast=bool)
+USE_LIGHTRAG = config("USE_LIGHTRAG", default=True, cast=bool)
+USE_MS_GRAPHRAG = config("USE_MS_GRAPHRAG", default=True, cast=bool)
+
+GRAPHRAG_INDEX_TYPES = []
+
+if USE_MS_GRAPHRAG:
+    GRAPHRAG_INDEX_TYPES.append("ktem.index.file.graph.GraphRAGIndex")
+if USE_NANO_GRAPHRAG:
+    GRAPHRAG_INDEX_TYPES.append("ktem.index.file.graph.NanoGraphRAGIndex")
+if USE_LIGHTRAG:
+    GRAPHRAG_INDEX_TYPES.append("ktem.index.file.graph.LightRAGIndex")
 
 KH_INDEX_TYPES = [
     "ktem.index.file.FileIndex",
-    "ktem.index.file.graph.GraphRAGIndex",
+    *GRAPHRAG_INDEX_TYPES,
 ]
-KH_INDICES = [
+
+GRAPHRAG_INDICES = [
     {
-        "name": "File",
+        "name": graph_type.split(".")[-1].replace("Index", "")
+        + " Collection",  # get last name
         "config": {
             "supported_file_types": (
                 ".png, .jpeg, .jpg, .tiff, .tif, .pdf, .xls, .xlsx, .doc, .docx, "
-                ".pptx, .csv, .html, .mhtml, .txt, .zip"
+                ".pptx, .csv, .html, .mhtml, .txt, .md, .zip"
             ),
-            "private": False,
+            "private": True,
         },
-        "index_type": "ktem.index.file.FileIndex",
-    },
+        "index_type": graph_type,
+    }
+    for graph_type in GRAPHRAG_INDEX_TYPES
+]
+
+KH_INDICES = [
     {
-        "name": "GraphRAG",
+        "name": "File Collection",
         "config": {
             "supported_file_types": (
                 ".png, .jpeg, .jpg, .tiff, .tif, .pdf, .xls, .xlsx, .doc, .docx, "
-                ".pptx, .csv, .html, .mhtml, .txt, .zip"
+                ".pptx, .csv, .html, .mhtml, .txt, .md, .zip"
             ),
-            "private": False,
+            "private": True,
         },
-        "index_type": "ktem.index.file.graph.GraphRAGIndex",
+        "index_type": "ktem.index.file.FileIndex",
     },
+    *GRAPHRAG_INDICES,
 ]