feat: bugfixes and massive simplification of DocumentQnA test (#51)

* VectorStoreIndexCreator, error fixes, tweeking, Documents QnA test

* changing Web and Pdf sources to use propper document

* bugfixes and massive simplification of DocumentQnA test

* bugfix

src/libs/LangChain.Core/Chains/CombineDocuments/BaseCombineDocumentsChain.cs

@@ -32,7 +32,7 @@ public abstract class BaseCombineDocumentsChain(BaseCombineDocumentsChainInput f
     /// <returns></returns>
     public override async Task<IChainValues> CallAsync(IChainValues values)
     {
-        var docs = values.Value[InputKey];
+        var docs = values.Value["input_documents"];
 
         //Other keys are assumed to be needed for LLM prediction
         var otherKeys = values.Value

src/libs/LangChain.Core/Chains/LLM/LLMChainInput.cs

@@ -13,7 +13,7 @@ public class LlmChainInput(
 {
     public BasePromptTemplate Prompt { get; set; } = prompt;
     public IChatModel Llm { get; set; } = llm;
-    public string OutputKey { get; set; }
+    public string OutputKey { get; set; } = "text";
     public bool? Verbose { get; set; }
     public CallbackManager CallbackManager { get; set; }
     public BaseMemory? Memory { get; set; } = memory;

src/libs/LangChain.Core/Chains/RetrievalQA/BaseRetrievalQaChain.cs

@@ -36,14 +36,15 @@ public abstract class BaseRetrievalQaChain(BaseRetrievalQaChainInput fields) : B
     /// <exception cref="NotImplementedException"></exception>
     public override async Task<IChainValues> CallAsync(IChainValues values)
     {
+
         var question = values.Value[_inputKey].ToString();
 
         var docs = (await GetDocsAsync(question)).ToList();
 
         var input = new Dictionary<string, object>
         {
-            [fields.DocumentsKey] = docs,
-            [_inputKey] = question
+            ["input_documents"] = docs,
+            [_inputKey]= question
         };
 
         var answer = await _combineDocumentsChain.Run(input);

src/libs/LangChain.Core/Chains/RetrievalQA/BaseRetrievalQaChainInput.cs

@@ -13,7 +13,6 @@ public class BaseRetrievalQaChainInput(BaseCombineDocumentsChain combineDocument
     public bool ReturnSourceDocuments { get; set; }
 
     public string InputKey { get; set; } = "question";
-    public string DocumentsKey { get; set; } = "input_documents";
     public string OutputKey { get; set; } = "output_text";
     public bool? Verbose { get; set; }
     public CallbackManager? CallbackManager { get; set; }

src/libs/LangChain.Core/Indexes/VectorStoreIndexWrapper.cs

@@ -16,15 +16,14 @@ public VectorStoreIndexWrapper(VectorStore vectorStore)
         _vectorStore = vectorStore;
     }
 
-    public Task<string?> QueryAsync(string question, BaseCombineDocumentsChain llm, string documentsKey= "input_documents", string questionKey="question", string outputKey= "output_text")
+    public Task<string?> QueryAsync(string question, BaseCombineDocumentsChain llm, string inputKey= "question", string outputKey= "output_text")
     {
         var chain = new RetrievalQaChain(
             new RetrievalQaChainInput(
                 llm,
                 _vectorStore.AsRetreiver())
             {
-                InputKey= questionKey,
-                DocumentsKey= documentsKey,
+                InputKey= inputKey,
                 OutputKey= outputKey,
             }
         );

src/tests/LangChain.Providers.LLamaSharp.IntegrationTests/LLamaSharpTests.cs

@@ -1,4 +1,5 @@
-using LangChain.Chains.CombineDocuments;
+using LangChain.Abstractions.Embeddings.Base;
+using LangChain.Chains.CombineDocuments;
 using LangChain.Chains.LLM;
 using LangChain.Databases;
 using LangChain.Databases.InMemory;
@@ -9,6 +10,7 @@
 using LangChain.Providers.Downloader;
 using LangChain.Providers.LLamaSharp;
 using LangChain.TextSplitters;
+using Microsoft.SemanticKernel.AI.Embeddings;
 
 namespace LangChain.Providers.LLamaSharp.IntegrationTests;
 
@@ -103,58 +105,85 @@ public void EmbeddingsTestWithInMemory()
 
         Assert.AreEqual("My dog name is Bob", closest.PageContent);
     }
-
-    [TestMethod]
-#if CONTINUOUS_INTEGRATION_BUILD
-    [Ignore]
-#endif
-    public void DocumentsQuestionAnsweringTest()
+
+
+    #region Helpers
+    IEmbeddings CreateEmbeddings()
     {
-        // setup
         var embeddings = new LLamaSharpEmbeddings(new LLamaSharpConfiguration
         {
             PathToModelFile = ModelPath,
             Temperature = 0
         });
+        return embeddings;
 
+    }
+
+    IChatModel CreateInstructionModel()
+    {
         var model = new LLamaSharpModelInstruction(new LLamaSharpConfiguration
         {
             PathToModelFile = ModelPath,
             Temperature = 0
         });
+        return model;
 
+    }
+
+    VectorStoreIndexWrapper CreateVectorStoreIndex(IEmbeddings embeddings, string[] texts)
+    {
         InMemoryVectorStore vectorStore = new InMemoryVectorStore(embeddings);
+        var textSplitter = new CharacterTextSplitter();
+        VectorStoreIndexCreator indexCreator = new VectorStoreIndexCreator(vectorStore, textSplitter);
+        var index = indexCreator.FromDocumentsAsync(texts.Select(x => new Document(x)).ToList()).Result;
+        return index;
+    }
+
+    PromptTemplate CreatePromptTemplate()
+    {
+        string prompt = "Use the following pieces of context to answer the question at the end. If you don't know the answer, just say that you don't know, don't try to make up an answer.\n\n{context}\n\nQuestion: {question}\nHelpful Answer:";
+        var template = new PromptTemplate(new PromptTemplateInput(prompt, new List<string>() { "context", "question" }));
+        return template;
+    }
+    #endregion
+
+
+    [TestMethod]
+#if CONTINUOUS_INTEGRATION_BUILD
+    [Ignore]
+#endif
+    public void DocumentsQuestionAnsweringTest()
+    {
+        // setup
+        var embeddings = CreateEmbeddings();
+        var model = CreateInstructionModel();
+
         string[] texts = new string[]
         {
             "I spent entire day watching TV",
             "My dog name is Bob",
             "This icecream is delicious",
             "It is cold in space"
         };
-        var textSplitter = new CharacterTextSplitter();
-        VectorStoreIndexCreator indexCreator = new VectorStoreIndexCreator(vectorStore, textSplitter); 
-        var index=indexCreator.FromDocumentsAsync(texts.Select(x=>new Document(x)).ToList()).Result;
+
+        var index = CreateVectorStoreIndex(embeddings, texts);
+        var template = CreatePromptTemplate();
+
+        var chain = new LlmChain(new LlmChainInput(model, template));
 
-        string prompt = "Use the following pieces of context to answer the question at the end. If you don't know the answer, just say that you don't know, don't try to make up an answer.\n\n{context}\n\nQuestion: {question}\nHelpful Answer:";
-        var template = new PromptTemplate(new PromptTemplateInput(prompt, new List<string>() { "context", "question" }));
-        var chain = new LlmChain(new LlmChainInput(model, template)
-        {
-            OutputKey = "text"
-        });
         var stuffDocumentsChain = new StuffDocumentsChain(new StuffDocumentsChainInput(chain)
         {
-            InputKey = "context",
-            DocumentVariableName = "context",
-            OutputKey = "text",
+            DocumentVariableName = "context", // variable name in prompt template
+                                              // for the documents
         });
 
 
         // test
-        var query = "What is the good name for a pet? Tell me only the name, no explanations.";
-        var answer=index.QueryAsync(query, stuffDocumentsChain,
-            documentsKey: "context",
-            questionKey: "question",
-            outputKey: "text").Result;
+        var question = "What is the good name for a pet?";
+        var answer=index.QueryAsync(question, stuffDocumentsChain,
+            inputKey:"question" // variable name in prompt template for the question
+                                // it would be passed by to stuffDocumentsChain
+            ).Result;