Merge branch 'main' into beta

CHANGELOG.md

@@ -1,9 +1,15 @@
-## [1.1.1](https://github.com/lpm0073/netec-llm/compare/v1.1.0...v1.1.1) (2023-12-01)
+## [1.1.2](https://github.com/lpm0073/hybrid-search-retriever/compare/v1.1.1...v1.1.2) (2023-12-01)
+
+
+### Bug Fixes
+
+* syntax error in examples.prompt ([230b709](https://github.com/lpm0073/hybrid-search-retriever/commit/230b7090c96bdd4d7d8757b182f891ab1b82c6f4))
 
+## [1.1.1](https://github.com/lpm0073/netec-llm/compare/v1.1.0...v1.1.1) (2023-12-01)
 
 ### Bug Fixes
 
-* had to switch to bm25_encoder so that vector store is searchable ([bad6994](https://github.com/lpm0073/netec-llm/commit/bad699481d217dde81877d85124395529652dabe))
+- had to switch to bm25_encoder so that vector store is searchable ([bad6994](https://github.com/lpm0073/netec-llm/commit/bad699481d217dde81877d85124395529652dabe))
 
 # [1.1.0](https://github.com/lpm0073/netec-llm/compare/v1.0.0...v1.1.0) (2023-12-01)
 

Makefile

@@ -3,7 +3,7 @@ SHELL := /bin/bash
 ifneq ("$(wildcard .env)","")
     include .env
 else
-    $(shell echo -e "OPENAI_API_ORGANIZATION=PLEASE-ADD-ME\nOPENAI_API_KEY=PLEASE-ADD-ME\nPINECONE_API_KEY=PLEASE-ADD-ME\nPINECONE_ENVIRONMENT=gcp-starter\nPINECONE_INDEX_NAME=netec-ssm\nDEBUG_MODE=True\n" >> .env)
+    $(shell echo -e "OPENAI_API_ORGANIZATION=PLEASE-ADD-ME\nOPENAI_API_KEY=PLEASE-ADD-ME\nPINECONE_API_KEY=PLEASE-ADD-ME\nPINECONE_ENVIRONMENT=gcp-starter\nPINECONE_INDEX_NAME=hsr\nOPENAI_CHAT_MODEL_NAME=gpt-3.5-turbo\nOPENAI_PROMPT_MODEL_NAME=text-davinci-003\nOPENAI_CHAT_TEMPERATURE=0.0\nOPENAI_CHAT_MAX_RETRIES=3\nDEBUG_MODE=True\n" >> .env)
 endif
 
 .PHONY: analyze init activate test lint clean

models/const.py

@@ -1,6 +1,6 @@
 # -*- coding: utf-8 -*-
 # pylint: disable=too-few-public-methods
-"""Sales Support Model (SSM) for the LangChain project."""
+"""Sales Support Model (hsr) for the LangChain project."""
 
 import os
 
@@ -15,11 +15,26 @@
     OPENAI_API_ORGANIZATION = os.environ["OPENAI_API_ORGANIZATION"]
     PINECONE_API_KEY = os.environ["PINECONE_API_KEY"]
     PINECONE_ENVIRONMENT = os.environ["PINECONE_ENVIRONMENT"]
-    PINECONE_INDEX_NAME = os.environ["PINECONE_INDEX_NAME"]
+    PINECONE_INDEX_NAME = os.environ.get("PINECONE_INDEX_NAME", "hsr")
+    OPENAI_CHAT_MODEL_NAME = os.environ.get("OPENAI_CHAT_MODEL_NAME", "gpt-3.5-turbo")
+    OPENAI_PROMPT_MODEL_NAME = os.environ.get("OPENAI_PROMPT_MODEL_NAME", "text-davinci-003")
+    OPENAI_CHAT_TEMPERATURE = float(os.environ.get("OPENAI_CHAT_TEMPERATURE", 0.0))
+    OPENAI_CHAT_MAX_RETRIES = int(os.environ.get("OPENAI_CHAT_MAX_RETRIES", 3))
+    OPENAI_CHAT_CACHE = bool(os.environ.get("OPENAI_CHAT_CACHE", True))
 else:
     raise FileNotFoundError("No .env file found in root directory of repository")
 
 
+class Config:
+    """Configuration parameters."""
+
+    OPENAI_CHAT_MODEL_NAME: str = OPENAI_CHAT_MODEL_NAME
+    OPENAI_PROMPT_MODEL_NAME: str = OPENAI_PROMPT_MODEL_NAME
+    OPENAI_CHAT_TEMPERATURE: float = OPENAI_CHAT_TEMPERATURE
+    OPENAI_CHAT_MAX_RETRIES: int = OPENAI_CHAT_MAX_RETRIES
+    OPENAI_CHAT_CACHE: bool = OPENAI_CHAT_CACHE
+
+
 class Credentials:
     """Credentials."""
 

models/examples/load.py

@@ -1,15 +1,15 @@
 # -*- coding: utf-8 -*-
-"""Sales Support Model (SSM) Retrieval Augmented Generation (RAG)"""
+"""Sales Support Model (hsr) Retrieval Augmented Generation (RAG)"""
 import argparse
 
 from models.hybrid_search_retreiver import HybridSearchRetriever
 
 
-ssm = HybridSearchRetriever()
+hsr = HybridSearchRetriever()
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(description="RAG example")
     parser.add_argument("filepath", type=str, help="Location of PDF documents")
     args = parser.parse_args()
 
-    ssm.load(filepath=args.filepath)
+    hsr.load(filepath=args.filepath)

models/examples/prompt.py

@@ -1,18 +1,18 @@
 # -*- coding: utf-8 -*-
-"""Sales Support Model (SSM)"""
+"""Sales Support Model (hsr)"""
 import argparse
 
 from models.hybrid_search_retreiver import HybridSearchRetriever
 
 
-ssm = HybridSearchRetriever()
+hsr = HybridSearchRetriever()
 
 
 if __name__ == "__main__":
-    parser = argparse.ArgumentParser(description="SSM examples")
+    parser = argparse.ArgumentParser(description="hsr examples")
     parser.add_argument("system_prompt", type=str, help="A system prompt to send to the model.")
     parser.add_argument("human_prompt", type=str, help="A human prompt to send to the model.")
     args = parser.parse_args()
 
-    result = ssm.cached_chat_request(args.system_prompt, args.human_prompt)
+    result = hsr.cached_chat_request(args.system_prompt, args.human_prompt)
     print(result)

models/examples/rag.py

@@ -1,16 +1,16 @@
 # -*- coding: utf-8 -*-
-"""Sales Support Model (SSM) Retrieval Augmented Generation (RAG)"""
+"""Sales Support Model (hsr) Retrieval Augmented Generation (RAG)"""
 import argparse
 
 from models.hybrid_search_retreiver import HybridSearchRetriever
 
 
-ssm = HybridSearchRetriever()
+hsr = HybridSearchRetriever()
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(description="RAG example")
     parser.add_argument("prompt", type=str, help="A question about the PDF contents")
     args = parser.parse_args()
 
-    result = ssm.rag(prompt=args.prompt)
+    result = hsr.rag(prompt=args.prompt)
     print(result)

models/examples/training_services.py

@@ -1,19 +1,19 @@
 # -*- coding: utf-8 -*-
-"""Sales Support Model (SSM) for the LangChain project."""
+"""Sales Support Model (hsr) for the LangChain project."""
 import argparse
 
 from models.hybrid_search_retreiver import HybridSearchRetriever
 from models.prompt_templates import NetecPromptTemplates
 
 
-ssm = HybridSearchRetriever()
+hsr = HybridSearchRetriever()
 templates = NetecPromptTemplates()
 
 if __name__ == "__main__":
-    parser = argparse.ArgumentParser(description="SSM examples")
+    parser = argparse.ArgumentParser(description="hsr examples")
     parser.add_argument("concept", type=str, help="A kind of training that Netec provides.")
     args = parser.parse_args()
 
     prompt = templates.training_services
-    result = ssm.prompt_with_template(prompt=prompt, concept=args.concept)
+    result = hsr.prompt_with_template(prompt=prompt, concept=args.concept)
     print(result)

models/examples/training_services_oracle.py

@@ -1,19 +1,19 @@
 # -*- coding: utf-8 -*-
-"""Sales Support Model (SSM) for the LangChain project."""
+"""Sales Support Model (hsr) for the LangChain project."""
 import argparse
 
 from models.hybrid_search_retreiver import HybridSearchRetriever
 from models.prompt_templates import NetecPromptTemplates
 
 
-ssm = HybridSearchRetriever()
+hsr = HybridSearchRetriever()
 templates = NetecPromptTemplates()
 
 if __name__ == "__main__":
-    parser = argparse.ArgumentParser(description="SSM Oracle examples")
+    parser = argparse.ArgumentParser(description="hsr Oracle examples")
     parser.add_argument("concept", type=str, help="An Oracle certification exam prep")
     args = parser.parse_args()
 
     prompt = templates.oracle_training_services
-    result = ssm.prompt_with_template(prompt=prompt, concept=args.concept)
+    result = hsr.prompt_with_template(prompt=prompt, concept=args.concept)
     print(result)

models/hybrid_search_retreiver.py

@@ -20,7 +20,6 @@
 import glob
 import os
 import textwrap
-from typing import List
 
 # pinecone integration
 import pinecone
@@ -44,25 +43,23 @@
 from pinecone_text.sparse import BM25Encoder
 
 # this project
-from models.const import Credentials
+from models.const import Config, Credentials
 
 
 ###############################################################################
 # initializations
 ###############################################################################
-DEFAULT_MODEL_NAME = "text-davinci-003"
+DEFAULT_MODEL_NAME = Config.OPENAI_PROMPT_MODEL_NAME
 pinecone.init(api_key=Credentials.PINECONE_API_KEY, environment=Credentials.PINECONE_ENVIRONMENT)
 set_llm_cache(InMemoryCache())
 
 
 class TextSplitter:
     """
-    Custom text splitter that add metadata to the Document object
+    Custom text splitter that adds metadata to the Document object
     which is required by PineconeHybridSearchRetriever.
     """
 
-    # ...
-
     def create_documents(self, texts):
         """Create documents"""
         documents = []
@@ -74,16 +71,16 @@ def create_documents(self, texts):
 
 
 class HybridSearchRetriever:
-    """Sales Support Model (SSM)."""
+    """Hybrid Search Retriever (OpenAI + Pinecone)"""
 
     # prompting wrapper
     chat = ChatOpenAI(
         api_key=Credentials.OPENAI_API_KEY,
         organization=Credentials.OPENAI_API_ORGANIZATION,
-        cache=True,
-        max_retries=3,
-        model="gpt-3.5-turbo",
-        temperature=0.0,
+        cache=Config.OPENAI_CHAT_CACHE,
+        max_retries=Config.OPENAI_CHAT_MAX_RETRIES,
+        model=Config.OPENAI_CHAT_MODEL_NAME,
+        temperature=Config.OPENAI_CHAT_TEMPERATURE,
     )
 
     # embeddings
@@ -112,22 +109,6 @@ def prompt_with_template(self, prompt: PromptTemplate, concept: str, model: str
         retval = llm(prompt.format(concept=concept))
         return retval
 
-    def fit_tf_idf_values(self, corpus: List[str]):
-        """Fit TF-IDF values.
-        1. Fit the BM25 encoder on the corpus
-        2. Encode the corpus
-        3. Store the encoded corpus in Pinecone
-        """
-        corpus = ["foo", "bar", "world", "hello"]
-
-        # fit tf-idf values on your corpus
-        self.bm25_encoder.fit(corpus)
-
-        # persist the values to a json file
-        self.bm25_encoder.dump("bm25_values.json")
-        self.bm25_encoder = BM25Encoder().load("bm25_values.json")
-        self.bm25_encoder.fit(corpus)
-
     def load(self, filepath: str):
         """
         Embed PDF.
@@ -201,9 +182,9 @@ def rag(self, prompt: str):
         document_texts = [doc.page_content for doc in documents]
         leader = textwrap.dedent(
             """\
-            You can assume that the following is true,
-            and you should attempt to incorporate these facts
-            in your response:
+            \n\nYou can assume that the following is true.
+            You should attempt to incorporate these facts
+            into your response:\n\n
         """
         )
 

models/prompt_templates.py

@@ -1,6 +1,6 @@
 # -*- coding: utf-8 -*-
 # pylint: disable=too-few-public-methods
-"""Sales Support Model (SSM) prompt templates"""
+"""Sales Support Model (hsr) prompt templates"""
 
 from langchain.prompts import PromptTemplate
 

models/tests/test_examples.py

@@ -0,0 +1,68 @@
+# -*- coding: utf-8 -*-
+# flake8: noqa: F401
+"""
+Test command line example prompts.
+"""
+from unittest.mock import MagicMock, patch
+
+import pytest  # pylint: disable=unused-import
+
+from models.examples.prompt import hsr as prompt_hrs
+from models.examples.rag import hsr as rag_hsr
+from models.examples.training_services import hsr as training_services_hsr
+from models.examples.training_services_oracle import hsr as training_services_oracle_hsr
+from models.prompt_templates import NetecPromptTemplates
+
+
+HUMAN_PROMPT = 'return the word "SUCCESS" in upper case.'
+
+
+class TestExamples:
+    """Test command line examples."""
+
+    @patch("argparse.ArgumentParser.parse_args")
+    def test_prompt(self, mock_parse_args):
+        """Test prompt example."""
+        mock_args = MagicMock()
+        mock_args.system_prompt = "you are a helpful assistant"
+        mock_args.human_prompt = HUMAN_PROMPT
+        mock_parse_args.return_value = mock_args
+
+        result = prompt_hrs.cached_chat_request(mock_args.system_prompt, mock_args.human_prompt)
+        assert result == "SUCCESS"
+
+    @patch("argparse.ArgumentParser.parse_args")
+    def test_rag(self, mock_parse_args):
+        """Test RAG example."""
+        mock_args = MagicMock()
+        mock_args.human_prompt = HUMAN_PROMPT
+        mock_parse_args.return_value = mock_args
+
+        result = rag_hsr.rag(mock_args.human_prompt)
+        assert result == "SUCCESS"
+
+    @patch("argparse.ArgumentParser.parse_args")
+    def test_training_services(self, mock_parse_args):
+        """Test training services templates."""
+        mock_args = MagicMock()
+        mock_args.human_prompt = HUMAN_PROMPT
+        mock_parse_args.return_value = mock_args
+
+        templates = NetecPromptTemplates()
+        prompt = templates.training_services
+
+        result = training_services_hsr.prompt_with_template(prompt=prompt, concept=mock_args.human_prompt)
+        assert "SUCCESS" in result
+
+    @patch("argparse.ArgumentParser.parse_args")
+    def test_oracle_training_services(self, mock_parse_args):
+        """Test oracle training services."""
+        mock_args = MagicMock()
+        mock_args.human_prompt = HUMAN_PROMPT
+        mock_parse_args.return_value = mock_args
+
+        templates = NetecPromptTemplates()
+        prompt = templates.oracle_training_services
+
+        result = training_services_oracle_hsr.prompt_with_template(prompt=prompt, concept=mock_args.human_prompt)
+        assert "SUCCESS" in result

models/tests/test_hsr.py

@@ -26,8 +26,8 @@ def test_01_basic(self):
     def test_02_class_aatribute_types(self):
         """ensure that class attributes are of the correct type"""
 
-        ssm = HybridSearchRetriever()
-        assert isinstance(ssm.chat, ChatOpenAI)
-        assert isinstance(ssm.pinecone_index, Index)
-        assert isinstance(ssm.text_splitter, TextSplitter)
-        assert isinstance(ssm.openai_embeddings, OpenAIEmbeddings)
+        hsr = HybridSearchRetriever()
+        assert isinstance(hsr.chat, ChatOpenAI)
+        assert isinstance(hsr.pinecone_index, Index)
+        assert isinstance(hsr.text_splitter, TextSplitter)
+        assert isinstance(hsr.openai_embeddings, OpenAIEmbeddings)

models/tests/test_openai.py

@@ -15,8 +15,8 @@ class TestOpenAI:
     def test_03_test_openai_connectivity(self):
         """Ensure that we have connectivity to OpenAI."""
 
-        ssm = HybridSearchRetriever()
-        retval = ssm.cached_chat_request(
+        hsr = HybridSearchRetriever()
+        retval = hsr.cached_chat_request(
             "your are a helpful assistant", "please return the value 'CORRECT' in all upper case."
         )
         assert retval == "CORRECT"

models/tests/test_prompts.py

@@ -12,14 +12,14 @@
 class TestPrompts:
     """Test HybridSearchRetriever class."""
 
-    ssm = HybridSearchRetriever()
+    hsr = HybridSearchRetriever()
     templates = NetecPromptTemplates()
 
     def test_oracle_training_services(self):
         """Test a prompt with the Oracle training services template"""
 
         prompt = self.templates.oracle_training_services
-        result = self.ssm.prompt_with_template(prompt=prompt, concept="Oracle database administrator")
+        result = self.hsr.prompt_with_template(prompt=prompt, concept="Oracle database administrator")
         assert result
         assert "Oracle" in result
         assert "training" in result
@@ -28,7 +28,7 @@ def test_training_services(self):
         """Test a prompt with the training services template"""
 
         prompt = self.templates.training_services
-        result = self.ssm.prompt_with_template(prompt=prompt, concept="Microsoft certified Azure AI engineer associate")
+        result = self.hsr.prompt_with_template(prompt=prompt, concept="Microsoft certified Azure AI engineer associate")
         assert result
         assert "Microsoft" in result
         assert "training" in result