Merge branch 'arc53:main' into main

application/api/answer/routes.py

@@ -28,12 +28,15 @@ vectors_collection = db["vectors"]
 prompts_collection = db["prompts"]
 answer = Blueprint('answer', __name__)
 
-if settings.LLM_NAME == "gpt4":
-    gpt_model = 'gpt-4'
+gpt_model = ""
+# to have some kind of default behaviour
+if settings.LLM_NAME == "openai":
+    gpt_model = 'gpt-3.5-turbo'
 elif settings.LLM_NAME == "anthropic":
     gpt_model = 'claude-2'
-else:
-    gpt_model = 'gpt-3.5-turbo'
+
+if settings.MODEL_NAME:  # in case there is particular model name configured
+    gpt_model = settings.MODEL_NAME
 
 # load the prompts
 current_dir = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
@@ -95,7 +98,7 @@ def is_azure_configured():
     return settings.OPENAI_API_BASE and settings.OPENAI_API_VERSION and settings.AZURE_DEPLOYMENT_NAME
 
 
-def complete_stream(question, docsearch, chat_history, api_key, prompt_id, conversation_id):
+def complete_stream(question, docsearch, chat_history, api_key, prompt_id, conversation_id, chunks=2):
     llm = LLMCreator.create_llm(settings.LLM_NAME, api_key=api_key)
 
     if prompt_id == 'default':
@@ -106,8 +109,11 @@ def complete_stream(question, docsearch, chat_history, api_key, prompt_id, conve
         prompt = chat_combine_strict
     else:
         prompt = prompts_collection.find_one({"_id": ObjectId(prompt_id)})["content"]
-
-    docs = docsearch.search(question, k=2)
+        
+    if chunks == 0:
+        docs = []
+    else:
+        docs = docsearch.search(question, k=chunks)
     if settings.LLM_NAME == "llama.cpp":
         docs = [docs[0]]
     # join all page_content together with a newline
@@ -190,6 +196,10 @@ def stream():
         prompt_id = data["prompt_id"]
     else:
         prompt_id = 'default'
+    if 'chunks' in data:
+        chunks = int(data["chunks"])
+    else:
+        chunks = 2
 
     # check if active_docs is set
 
@@ -211,7 +221,8 @@ def stream():
         complete_stream(question, docsearch,
                         chat_history=history, api_key=api_key,
                         prompt_id=prompt_id,
-                        conversation_id=conversation_id), mimetype="text/event-stream"
+                        conversation_id=conversation_id,
+                        chunks=chunks), mimetype="text/event-stream"
     )
 
 
@@ -237,6 +248,10 @@ def api_answer():
         prompt_id = data["prompt_id"]
     else:
         prompt_id = 'default'
+    if 'chunks' in data:
+        chunks = int(data["chunks"])
+    else:
+        chunks = 2
 
     if prompt_id == 'default':
         prompt = chat_combine_template
@@ -260,7 +275,10 @@ def api_answer():
 
 
 
-        docs = docsearch.search(question, k=2)
+        if chunks == 0:
+            docs = []
+        else:
+            docs = docsearch.search(question, k=chunks)
         # join all page_content together with a newline
         docs_together = "\n".join([doc.page_content for doc in docs])
         p_chat_combine = prompt.replace("{summaries}", docs_together)
@@ -359,9 +377,15 @@ def api_search():
         vectorstore = get_vectorstore({"active_docs": data["active_docs"]})
     else:
         vectorstore = ""
+    if 'chunks' in data:
+        chunks = int(data["chunks"])
+    else:
+        chunks = 2
     docsearch = VectorCreator.create_vectorstore(settings.VECTOR_STORE, vectorstore, embeddings_key)
-
-    docs = docsearch.search(question, k=2)
+    if chunks == 0:
+        docs = []
+    else:
+        docs = docsearch.search(question, k=chunks)
 
     source_log_docs = []
     for doc in docs:

application/core/settings.py

@@ -9,6 +9,7 @@ current_dir = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__
 
 class Settings(BaseSettings):
     LLM_NAME: str = "docsgpt"
+    MODEL_NAME: Optional[str] = None # when LLM_NAME is openai, MODEL_NAME can be e.g. gpt-4-turbo-preview or gpt-3.5-turbo
     EMBEDDINGS_NAME: str = "huggingface_sentence-transformers/all-mpnet-base-v2"
     CELERY_BROKER_URL: str = "redis://localhost:6379/0"
     CELERY_RESULT_BACKEND: str = "redis://localhost:6379/1"

docs/package-lock.json

@@ -7,7 +7,7 @@
       "license": "MIT",
       "dependencies": {
         "@vercel/analytics": "^1.1.1",
-        "docsgpt": "^0.3.0",
+        "docsgpt": "^0.3.7",
         "next": "^14.0.4",
         "nextra": "^2.13.2",
         "nextra-theme-docs": "^2.13.2",
@@ -422,6 +422,11 @@
         "node": ">=6.9.0"
       }
     },
+    "node_modules/@bpmn-io/snarkdown": {
+      "version": "2.2.0",
+      "resolved": "https://registry.npmjs.org/@bpmn-io/snarkdown/-/snarkdown-2.2.0.tgz",
+      "integrity": "sha512-bVD7FIoaBDZeCJkMRgnBPDeptPlto87wt2qaCjf5t8iLaevDmTPaREd6FpBEGsHlUdHFFZWRk4qAoEC5So2M0Q=="
+    },
     "node_modules/@braintree/sanitize-url": {
       "version": "6.0.4",
       "resolved": "https://registry.npmjs.org/@braintree/sanitize-url/-/sanitize-url-6.0.4.tgz",
@@ -4958,11 +4963,12 @@
       }
     },
     "node_modules/docsgpt": {
-      "version": "0.3.0",
-      "resolved": "https://registry.npmjs.org/docsgpt/-/docsgpt-0.3.0.tgz",
-      "integrity": "sha512-0yT2m+HAlJ+289p278c3Zi07bu2wr6zULOT/bYXtJ/nb59V2Vpfdj2xFB49+lYLSeVe8H+Ij5fFSNZ6RkVRfMQ==",
+      "version": "0.3.7",
+      "resolved": "https://registry.npmjs.org/docsgpt/-/docsgpt-0.3.7.tgz",
+      "integrity": "sha512-VHrXXOEFtjNTcpA8Blf3IzpLlJxOMhm/S5CM4FDjQEkdK9WWhI8yXd/0Rs/FS8oz7YbFrNxO758mlP7OtQtBBw==",
       "dependencies": {
         "@babel/plugin-transform-flow-strip-types": "^7.23.3",
+        "@bpmn-io/snarkdown": "^2.2.0",
         "@parcel/resolver-glob": "^2.12.0",
         "@parcel/transformer-svg-react": "^2.12.0",
         "@parcel/transformer-typescript-tsc": "^2.12.0",
@@ -4972,6 +4978,7 @@
         "@types/react-dom": "^18.2.19",
         "class-variance-authority": "^0.7.0",
         "clsx": "^2.1.0",
+        "dompurify": "^3.0.9",
         "flow-bin": "^0.229.2",
         "i": "^0.3.7",
         "install": "^0.13.0",
@@ -5029,9 +5036,9 @@
       }
     },
     "node_modules/dompurify": {
-      "version": "3.0.7",
-      "resolved": "https://registry.npmjs.org/dompurify/-/dompurify-3.0.7.tgz",
-      "integrity": "sha512-BViYTZoqP3ak/ULKOc101y+CtHDUvBsVgSxIF1ku0HmK6BRf+C03MC+tArMvOPtVtZp83DDh5puywKDu4sbVjQ=="
+      "version": "3.0.11",
+      "resolved": "https://registry.npmjs.org/dompurify/-/dompurify-3.0.11.tgz",
+      "integrity": "sha512-Fan4uMuyB26gFV3ovPoEoQbxRRPfTu3CvImyZnhGq5fsIEO+gEFLp45ISFt+kQBWsK5ulDdT0oV28jS1UrwQLg=="
     },
     "node_modules/domutils": {
       "version": "2.8.0",
@@ -6206,9 +6213,9 @@
       "integrity": "sha512-gfFQZrcTc8CnKXp6Y4/CBT3fTc0OVuDofpre4aEeEpSBPV5X5v4+Vmx+8snU7RLPrNHPKSgLxGo9YuQzz20o+w=="
     },
     "node_modules/katex": {
-      "version": "0.16.9",
-      "resolved": "https://registry.npmjs.org/katex/-/katex-0.16.9.tgz",
-      "integrity": "sha512-fsSYjWS0EEOwvy81j3vRA8TEAhQhKiqO+FQaKWp0m39qwOzHVBgAUBIXWj1pB+O2W3fIpNa6Y9KSKCVbfPhyAQ==",
+      "version": "0.16.10",
+      "resolved": "https://registry.npmjs.org/katex/-/katex-0.16.10.tgz",
+      "integrity": "sha512-ZiqaC04tp2O5utMsl2TEZTXxa6WSC4yo0fv5ML++D3QZv/vx2Mct0mTlRx3O+uUkjfuAgOkzsCmq5MiUEsDDdA==",
       "funding": [
         "https://opencollective.com/katex",
         "https://github.com/sponsors/katex"

frontend/src/Setting.tsx

@@ -8,6 +8,8 @@ import {
   setPrompt,
   selectSourceDocs,
   setSourceDocs,
+  setChunks,
+  selectChunks,
 } from './preferences/preferenceSlice';
 import { Doc } from './preferences/preferenceApi';
 import { useDarkTheme } from './hooks';
@@ -193,10 +195,13 @@ const Setting: React.FC = () => {
 const General: React.FC = () => {
   const themes = ['Light', 'Dark'];
   const languages = ['English'];
+  const chunks = ['0', '2', '4', '6', '8', '10'];
+  const selectedChunks = useSelector(selectChunks);
   const [isDarkTheme, toggleTheme] = useDarkTheme();
   const [selectedTheme, setSelectedTheme] = useState(
     isDarkTheme ? 'Dark' : 'Light',
   );
+  const dispatch = useDispatch();
   const [selectedLanguage, setSelectedLanguage] = useState(languages[0]);
   return (
     <div className="mt-[59px]">
@@ -211,7 +216,7 @@ const General: React.FC = () => {
           }}
         />
       </div>
-      <div>
+      <div className="mb-4">
         <p className="font-bold text-jet dark:text-bright-gray">
           Select Language
         </p>
@@ -221,6 +226,16 @@ const General: React.FC = () => {
           onSelect={setSelectedLanguage}
         />
       </div>
+      <div>
+        <p className="font-bold text-jet dark:text-bright-gray">
+          Chunks processed per query
+        </p>
+        <Dropdown
+          options={chunks}
+          selectedValue={selectedChunks}
+          onSelect={(value: string) => dispatch(setChunks(value))}
+        />
+      </div>
     </div>
   );
 };

frontend/src/conversation/ConversationBubble.tsx

@@ -160,7 +160,10 @@ const ConversationBubble = forwardRef<
             >
               {message}
             </ReactMarkdown>
-            {DisableSourceFE || type === 'ERROR' ? null : (
+            {DisableSourceFE ||
+            type === 'ERROR' ||
+            !sources ||
+            sources.length === 0 ? null : (
               <>
                 <span className="mt-3 h-px w-full bg-[#DEDEDE]"></span>
                 <div className="mt-3 flex w-full flex-row flex-wrap items-center justify-start gap-2">

frontend/src/conversation/conversationApi.ts

@@ -11,6 +11,7 @@ export function fetchAnswerApi(
   history: Array<any> = [],
   conversationId: string | null,
   promptId: string | null,
+  chunks: string,
 ): Promise<
   | {
       result: any;
@@ -65,6 +66,7 @@ export function fetchAnswerApi(
       active_docs: docPath,
       conversation_id: conversationId,
       prompt_id: promptId,
+      chunks: chunks,
     }),
     signal,
   })
@@ -95,6 +97,7 @@ export function fetchAnswerSteaming(
   history: Array<any> = [],
   conversationId: string | null,
   promptId: string | null,
+  chunks: string,
   onEvent: (event: MessageEvent) => void,
 ): Promise<Answer> {
   let namePath = selectedDocs.name;
@@ -130,6 +133,7 @@ export function fetchAnswerSteaming(
       history: JSON.stringify(history),
       conversation_id: conversationId,
       prompt_id: promptId,
+      chunks: chunks,
     };
     fetch(apiHost + '/stream', {
       method: 'POST',
@@ -192,6 +196,7 @@ export function searchEndpoint(
   selectedDocs: Doc,
   conversation_id: string | null,
   history: Array<any> = [],
+  chunks: string,
 ) {
   /*
   "active_docs": "default",
@@ -223,6 +228,7 @@ export function searchEndpoint(
     active_docs: docPath,
     conversation_id,
     history,
+    chunks: chunks,
   };
   return fetch(`${apiHost}/api/search`, {
     method: 'POST',

frontend/src/conversation/conversationSlice.ts

@@ -28,6 +28,7 @@ export const fetchAnswer = createAsyncThunk<Answer, { question: string }>(
           state.conversation.queries,
           state.conversation.conversationId,
           state.preference.prompt.id,
+          state.preference.chunks,
 
           (event) => {
             const data = JSON.parse(event.data);
@@ -51,6 +52,7 @@ export const fetchAnswer = createAsyncThunk<Answer, { question: string }>(
                 state.preference.selectedDocs!,
                 state.conversation.conversationId,
                 state.conversation.queries,
+                state.preference.chunks,
               ).then((sources) => {
                 //dispatch streaming sources
                 dispatch(
@@ -86,6 +88,7 @@ export const fetchAnswer = createAsyncThunk<Answer, { question: string }>(
           state.conversation.queries,
           state.conversation.conversationId,
           state.preference.prompt.id,
+          state.preference.chunks,
         );
         if (answer) {
           let sourcesPrepped = [];

frontend/src/preferences/preferenceSlice.ts

@@ -10,6 +10,7 @@ interface Preference {
   apiKey: string;
   prompt: { name: string; id: string; type: string };
   selectedDocs: Doc | null;
+  chunks: string;
   sourceDocs: Doc[] | null;
   conversations: { name: string; id: string }[] | null;
 }
@@ -17,6 +18,7 @@ interface Preference {
 const initialState: Preference = {
   apiKey: 'xxx',
   prompt: { name: 'default', id: 'default', type: 'public' },
+  chunks: '2',
   selectedDocs: {
     name: 'default',
     language: 'default',
@@ -51,6 +53,9 @@ export const prefSlice = createSlice({
     setPrompt: (state, action) => {
       state.prompt = action.payload;
     },
+    setChunks: (state, action) => {
+      state.chunks = action.payload;
+    },
   },
 });
 
@@ -60,6 +65,7 @@ export const {
   setSourceDocs,
   setConversations,
   setPrompt,
+  setChunks,
 } = prefSlice.actions;
 export default prefSlice.reducer;
 
@@ -91,6 +97,16 @@ prefListenerMiddleware.startListening({
   },
 });
 
+prefListenerMiddleware.startListening({
+  matcher: isAnyOf(setChunks),
+  effect: (action, listenerApi) => {
+    localStorage.setItem(
+      'DocsGPTChunks',
+      JSON.stringify((listenerApi.getState() as RootState).preference.chunks),
+    );
+  },
+});
+
 export const selectApiKey = (state: RootState) => state.preference.apiKey;
 export const selectApiKeyStatus = (state: RootState) =>
   !!state.preference.apiKey;
@@ -105,3 +121,4 @@ export const selectConversations = (state: RootState) =>
 export const selectConversationId = (state: RootState) =>
   state.conversation.conversationId;
 export const selectPrompt = (state: RootState) => state.preference.prompt;
+export const selectChunks = (state: RootState) => state.preference.chunks;

frontend/src/store.ts

@@ -8,11 +8,13 @@ import {
 const key = localStorage.getItem('DocsGPTApiKey');
 const prompt = localStorage.getItem('DocsGPTPrompt');
 const doc = localStorage.getItem('DocsGPTRecentDocs');
+const chunks = localStorage.getItem('DocsGPTChunks');
 
 const store = configureStore({
   preloadedState: {
     preference: {
       apiKey: key ?? '',
+      chunks: JSON.parse(chunks ?? '2').toString(),
       selectedDocs: doc !== null ? JSON.parse(doc) : null,
       prompt:
         prompt !== null