allow LlamaCppChat to auto download model from hf hub (#29)

libs/kotaemon/kotaemon/llms/chats/llamacpp.py

@@ -12,9 +12,14 @@ if TYPE_CHECKING:
 class LlamaCppChat(ChatLLM):
     """Wrapper around the llama-cpp-python's Llama model"""
 
-    model_path: str = Param(
+    model_path: Optional[str] = Param(
         help="Path to the model file. This is required to load the model.",
-        required=True,
+    )
+    repo_id: Optional[str] = Param(
+        help="Id of a repo on the HuggingFace Hub in the form of `user_name/repo_name`."
+    )
+    filename: Optional[str] = Param(
+        help="A filename or glob pattern to match the model file in the repo."
     )
     chat_format: str = Param(
         help=(
@@ -28,7 +33,7 @@ class LlamaCppChat(ChatLLM):
     n_ctx: Optional[int] = Param(512, help="Text context, 0 = from model")
     n_gpu_layers: Optional[int] = Param(
         0,
-        help=("Number of layers to offload to GPU. If -1, all layers are offloaded"),
+        help="Number of layers to offload to GPU. If -1, all layers are offloaded",
     )
     use_mmap: Optional[bool] = Param(
         True,
@@ -36,7 +41,7 @@ class LlamaCppChat(ChatLLM):
     )
     vocab_only: Optional[bool] = Param(
         False,
-        help=("If True, only the vocabulary is loaded. This is useful for debugging."),
+        help="If True, only the vocabulary is loaded. This is useful for debugging.",
     )
 
     _role_mapper: dict[str, str] = {
@@ -57,8 +62,11 @@ class LlamaCppChat(ChatLLM):
             )
 
         errors = []
-        if not self.model_path:
+        if not self.model_path and (not self.repo_id or not self.filename):
-            errors.append("- `model_path` is required to load the model")
+            errors.append(
+                "- `model_path` or `repo_id` and `filename` are required to load the"
+                " model"
+            )
 
         if not self.chat_format:
             errors.append(
@@ -69,15 +77,27 @@ class LlamaCppChat(ChatLLM):
         if errors:
             raise ValueError("\n".join(errors))
 
-        return Llama(
+        if self.model_path:
-            model_path=cast(str, self.model_path),
+            return Llama(
-            chat_format=self.chat_format,
+                model_path=cast(str, self.model_path),
-            lora_base=self.lora_base,
+                chat_format=self.chat_format,
-            n_ctx=self.n_ctx,
+                lora_base=self.lora_base,
-            n_gpu_layers=self.n_gpu_layers,
+                n_ctx=self.n_ctx,
-            use_mmap=self.use_mmap,
+                n_gpu_layers=self.n_gpu_layers,
-            vocab_only=self.vocab_only,
+                use_mmap=self.use_mmap,
-        )
+                vocab_only=self.vocab_only,
+            )
+        else:
+            return Llama.from_pretrained(
+                repo_id=self.repo_id,
+                filename=self.filename,
+                chat_format=self.chat_format,
+                lora_base=self.lora_base,
+                n_ctx=self.n_ctx,
+                n_gpu_layers=self.n_gpu_layers,
+                use_mmap=self.use_mmap,
+                vocab_only=self.vocab_only,
+            )
 
     def prepare_message(
         self, messages: str | BaseMessage | list[BaseMessage]

libs/kotaemon/pyproject.toml

@@ -22,7 +22,7 @@ dependencies = [
     "theflow",
     "llama-index>=0.9.0,<0.10.0",
     "llama-hub",
-    "gradio>=4.0.0,<=4.22.0",
+    "gradio>=4.26.0",
     "openpyxl",
     "cookiecutter",
     "click",
@@ -34,6 +34,7 @@ dependencies = [
     "unstructured",
     "pypdf",
     "html2text",
+    "fastembed",
 ]
 readme = "README.md"
 license = { text = "MIT License" }

libs/ktem/pyproject.toml

@@ -19,7 +19,6 @@ dependencies = [
     "python-decouple",
     "sqlalchemy",
     "sqlmodel",
-    "fastembed",
     "tiktoken",
     "gradio>=4.26.0",
     "markdown",