refactor get prompt service

aiplanethub · Oct 6, 2023 · 5aa7ff4 · 5aa7ff4
1 parent 53a995f
commit 5aa7ff4
Show file tree

Hide file tree

Showing 2 changed files with 171 additions and 31 deletions.
diff --git a/dummy.ipynb b/dummy.ipynb
@@ -0,0 +1,170 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "initial_id",
+   "metadata": {
+    "collapsed": true,
+    "ExecuteTime": {
+     "end_time": "2023-10-05T13:43:29.749167090Z",
+     "start_time": "2023-10-05T13:43:24.752225360Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from genai_stack.stack.stack import Stack\n",
+    "from genai_stack.model import HuggingFaceModel\n",
+    "from genai_stack.etl.langchain import LangchainETL\n",
+    "from genai_stack.embedding.langchain import LangchainEmbedding\n",
+    "from genai_stack.vectordb.chromadb import ChromaDB\n",
+    "from genai_stack.prompt_engine.engine import PromptEngine\n",
+    "from genai_stack.retriever.langchain import LangChainRetriever\n",
+    "from genai_stack.memory.langchain import ConversationBufferMemory"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "outputs": [],
+   "source": [
+    "etl = LangchainETL.from_kwargs(\n",
+    "    name=\"WebBaseLoader\",\n",
+    "    fields={\"web_path\": [\n",
+    "      \"https://aiplanet.com\",\n",
+    "    ]\n",
+    "  }\n",
+    ")"
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2023-10-05T13:43:29.752762828Z",
+     "start_time": "2023-10-05T13:43:29.749885841Z"
+    }
+   },
+   "id": "58197b4e9c357f27"
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "outputs": [],
+   "source": [
+    "llm = HuggingFaceModel.from_kwargs(model=\"skt/ko-gpt-trinity-1.2B-v0.5\")\n",
+    "config = {\n",
+    "  \"model_name\": \"sentence-transformers/all-mpnet-base-v2\",\n",
+    "  \"model_kwargs\": {\"device\": \"cpu\"},\n",
+    "  \"encode_kwargs\": {\"normalize_embeddings\": False},\n",
+    "}\n",
+    "embedding = LangchainEmbedding.from_kwargs(name=\"HuggingFaceEmbeddings\", fields=config)\n",
+    "chromadb = ChromaDB.from_kwargs()"
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2023-10-05T13:44:37.250590Z",
+     "start_time": "2023-10-05T13:44:37.223926726Z"
+    }
+   },
+   "id": "42fa2dcd7085ac80"
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "outputs": [
+    {
+     "ename": "NameError",
+     "evalue": "name 'PromptEngine' is not defined",
+     "output_type": "error",
+     "traceback": [
+      "\u001B[0;31m---------------------------------------------------------------------------\u001B[0m",
+      "\u001B[0;31mNameError\u001B[0m                                 Traceback (most recent call last)",
+      "Cell \u001B[0;32mIn[1], line 1\u001B[0m\n\u001B[0;32m----> 1\u001B[0m prompt_engine \u001B[38;5;241m=\u001B[39m \u001B[43mPromptEngine\u001B[49m\u001B[38;5;241m.\u001B[39mfrom_kwargs(should_validate\u001B[38;5;241m=\u001B[39m\u001B[38;5;28;01mFalse\u001B[39;00m)\n\u001B[1;32m      2\u001B[0m retriever \u001B[38;5;241m=\u001B[39m LangChainRetriever\u001B[38;5;241m.\u001B[39mfrom_kwargs()\n\u001B[1;32m      3\u001B[0m memory \u001B[38;5;241m=\u001B[39m ConversationBufferMemory\u001B[38;5;241m.\u001B[39mfrom_kwargs()\n",
+      "\u001B[0;31mNameError\u001B[0m: name 'PromptEngine' is not defined"
+     ]
+    }
+   ],
+   "source": [
+    "prompt_engine = PromptEngine.from_kwargs(should_validate=False)\n",
+    "retriever = LangChainRetriever.from_kwargs()\n",
+    "memory = ConversationBufferMemory.from_kwargs()\n",
+    "Stack(\n",
+    "  etl=etl,\n",
+    "  embedding=embedding,\n",
+    "  vectordb=chromadb,\n",
+    "  model=llm,\n",
+    "  prompt_engine=prompt_engine,\n",
+    "  retriever=retriever,\n",
+    "  memory=memory\n",
+    ")\n"
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2023-10-05T13:45:48.582398487Z",
+     "start_time": "2023-10-05T13:45:48.417937857Z"
+    }
+   },
+   "id": "81e991b63b233e6d"
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "outputs": [],
+   "source": [
+    "prompt1 = \"Why choose models from AI Marketplace?\""
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "start_time": "2023-10-05T13:43:49.783528419Z"
+    }
+   },
+   "id": "2e815c9b106810f0"
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "outputs": [],
+   "source": [],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "start_time": "2023-10-05T13:43:49.786045019Z"
+    }
+   },
+   "id": "80d05b20a6921f73"
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "outputs": [],
+   "source": [],
+   "metadata": {
+    "collapsed": false
+   },
+   "id": "8226f4204a117563"
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 2
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython2",
+   "version": "2.7.6"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}
diff --git a/genai_stack/genai_server/services/prompt_engine_service.py b/genai_stack/genai_server/services/prompt_engine_service.py
@@ -21,37 +21,7 @@ def get_prompt(self, data: PromptEngineGetRequestModel) -> PromptEngineGetRespon
             stack_session = session.get(StackSessionSchema, data.session_id)
             if stack_session is None:
                 raise HTTPException(status_code=404, detail=f"Session {data.session_id} not found")
-            prompt_session = (
-                session.query(PromptSchema)
-                .filter_by(stack_session=data.session_id, type=data.type.value)
-                .first()
-            )
-            if prompt_session is not None:
-                template = prompt_session.template
-                prompt_type_map = {
-                    PromptTypeEnum.SIMPLE_CHAT_PROMPT.value: "simple_chat_prompt_template",
-                    PromptTypeEnum.CONTEXTUAL_CHAT_PROMPT.value: "contextual_chat_prompt_template",
-                    PromptTypeEnum.CONTEXTUAL_QA_PROMPT.value: "contextual_qa_prompt_template",
-                }
-                input_variables = ["context", "history", "query"]
-                if data.type == PromptTypeEnum.SIMPLE_CHAT_PROMPT:
-                    input_variables.remove("context")
-                elif data.type == PromptTypeEnum.CONTEXTUAL_QA_PROMPT:
-                    input_variables.remove("history")
-                prompt = PromptTemplate(template=template, input_variables=input_variables)
-                stack = get_current_stack(
-                    engine=session,
-                    config=stack_config,
-                    session=stack_session,
-                    overide_config={
-                        "prompt_engine": {
-                            "should_validate": data.should_validate,
-                            prompt_type_map[data.type.value]: prompt
-                        }
-                    }
-                )
-            else:
-                stack = get_current_stack(config=stack_config, engine=session, session=stack_session)
+            stack = get_current_stack(config=stack_config, engine=session, session=stack_session)
             prompt = stack.prompt_engine.get_prompt_template(promptType=data.type, query=data.query)
             return PromptEngineGetResponseModel(
                 template=prompt.template,