fix: Docker and sagemaker setup (#1118)

* fix: docker copying extra files

* feat: allow configuring mode through env vars

* feat: Attempt to build and tag a docker image

* fix: run docker on release

* fix: typing in prompt transformation

* chore: remove tutorial comments

private_gpt/components/embedding/embedding_component.py

@@ -13,7 +13,7 @@ class EmbeddingComponent:
     @inject
     def __init__(self) -> None:
         match settings.llm.mode:
-            case "local":
+            case "local" | "sagemaker":
                 from llama_index.embeddings import HuggingFaceEmbedding
 
                 self.embedding_model = HuggingFaceEmbedding(

private_gpt/components/llm/custom/sagemaker.py

@@ -21,8 +21,6 @@ from llama_index.llms.llama_utils import (
 )
 
 if TYPE_CHECKING:
-    from collections.abc import Callable
-
     from llama_index.callbacks import CallbackManager
     from llama_index.llms import (
         CompletionResponseGen,
@@ -113,10 +111,10 @@ class SagemakerLLM(CustomLLM):
     context_window: int = Field(
         description="The maximum number of context tokens for the model."
     )
-    messages_to_prompt: Callable[..., str] = Field(
+    messages_to_prompt: Any = Field(
         description="The function to convert messages to a prompt.", exclude=True
     )
-    completion_to_prompt: Callable[..., str] = Field(
+    completion_to_prompt: Any = Field(
         description="The function to convert a completion to a prompt.", exclude=True
     )
     generate_kwargs: dict[str, Any] = Field(

private_gpt/components/llm/llm_component.py

@@ -37,6 +37,8 @@ class LLMComponent:
 
                 self.llm = SagemakerLLM(
                     endpoint_name=settings.sagemaker.endpoint_name,
+                    messages_to_prompt=messages_to_prompt,
+                    completion_to_prompt=completion_to_prompt,
                 )
             case "openai":
                 from llama_index.llms import OpenAI