nasa demo updates

.gitignore

@@ -2,5 +2,4 @@
 *.md
 __pycache__
 documents
-vecstore
-gDriveDocs
+tmp

Dockerfile

@@ -15,5 +15,8 @@ ENV GRADIO_SERVER_NAME=0.0.0.0
 # Install any needed packages specified in requirements.txt
 # RUN pip install --no-cache-dir -r requirements.txt # already installed in base image
 
+#to be moved to ttyd_base image
+RUN python -c "from sentence_transformers import SentenceTransformer; model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2', cache_folder='./tmp/hfEmbModel')"
+
 # Use ENTRYPOINT to allow passing user arguments
 ENTRYPOINT ["python", "app.py"]

app.py

@@ -1,28 +1,15 @@
 from dotenv import load_dotenv
 import datetime
 import openai
-import uuid
 import gradio as gr
-from langchain.embeddings import OpenAIEmbeddings
-from langchain.vectorstores import Chroma
-from langchain.text_splitter import CharacterTextSplitter, RecursiveCharacterTextSplitter
 from langchain.chains import ConversationalRetrievalChain
 from langchain.chains import RetrievalQA
-from langchain.embeddings import SentenceTransformerEmbeddings
 
 import os
-from langchain.chat_models import ChatOpenAI
-from langchain import OpenAI
-from langchain.document_loaders import WebBaseLoader, TextLoader, Docx2txtLoader, PyMuPDFLoader
-from whatsapp_chat_custom import WhatsAppChatLoader # use this instead of from langchain.document_loaders import WhatsAppChatLoader
-
-from ibm_watson_machine_learning.metanames import GenTextParamsMetaNames as GenParams
-from ibm_watson_machine_learning.foundation_models.utils.enums import DecodingMethods
 from ibm_watson_machine_learning.foundation_models import Model
-from ibm_watson_machine_learning.foundation_models.extensions.langchain import WatsonxLLM
 from ibm_watson_machine_learning.foundation_models.utils.enums import ModelTypes
 
-import genai
+from genai.model import Model as genaiModel
 
 from collections import deque
 import re
@@ -68,33 +55,34 @@ if mode.type!='userInputDocs':
 def setOaiApiKey(creds):
     creds = getOaiCreds(creds)
     try:
-        openai.Model.list(api_key=creds.get('oai_key','Null')) # test the API key
+        openAi_models = openai.OpenAI(api_key=creds.get('oai_key','Null')).models.list().data # test the API key
+        openAi_models = sorted([x.id for x in openAi_models if x.owned_by=='openai'])
         api_key_st = creds
-        return 'OpenAI credentials accepted.', *[x.update(interactive=False) for x in credComps_btn_tb], api_key_st
+        return 'OpenAI credentials accepted.', *[gr.update(interactive=False) for x in credComps_btn_tb], api_key_st, gr.update(choices=getModelChoices(openAi_models, ModelTypes, bam_models_def))
     except Exception as e:
         gr.Warning(str(e))
-        return [x.update() for x in credComps_op]
+        return *[gr.update() for x in credComps_op], gr.update()
 
 def setBamApiKey(creds):
     creds = getBamCreds(creds)
     try:
-        bam_models = genai.Model.models(credentials=creds['bam_creds'])
+        bam_models = genaiModel.models(credentials=creds['bam_creds'])
         bam_models = sorted(x.id for x in bam_models)
         api_key_st = creds
-        return 'BAM credentials accepted.', *[x.update(interactive=False) for x in credComps_btn_tb], api_key_st, model_dd.update(choices=getModelChoices(openAi_models, ModelTypes, bam_models))
+        return 'BAM credentials accepted.', *[gr.update(interactive=False) for x in credComps_btn_tb], api_key_st, gr.update(choices=getModelChoices(openAi_models_def, ModelTypes, bam_models))
     except Exception as e:
         gr.Warning(str(e))
-        return *[x.update() for x in credComps_op], model_dd.update()
+        return *[gr.update() for x in credComps_op], gr.update()
 
 def setWxApiKey(key, p_id):
     creds = getWxCreds(key, p_id)
     try:
         Model(model_id='google/flan-ul2', credentials=creds['credentials'], project_id=creds['project_id']) # test the API key
         api_key_st = creds
-        return 'Watsonx credentials accepted.', *[x.update(interactive=False) for x in credComps_btn_tb], api_key_st
+        return 'Watsonx credentials accepted.', *[gr.update(interactive=False) for x in credComps_btn_tb], api_key_st
     except Exception as e:
         gr.Warning(str(e))
-        return [x.update() for x in credComps_op]
+        return [gr.update() for x in credComps_op]
     
 
 # convert user uploaded data to vectorstore
@@ -109,12 +97,9 @@ def uiData_vecStore(userFiles, userUrls, api_key_st, vsDict_st={}, progress=gr.P
     userUrls = [x.strip() for x in userUrls.split(",")] if userUrls else []
     #create documents
     documents = data_ingestion(file_list=file_paths, url_list=userUrls, prog=progress)
-    if documents:
-        for file in file_paths:
-            os.remove(file)
-    else:
+    if not documents:
         gr.Error('No documents found')
-        return {}, '', *[x.update() for x in opComponents]
+        return {}, '', *[gr.update() for x in opComponents]
     # Splitting and Chunks
     docs = split_docs(documents)
     # Embeddings
@@ -122,7 +107,7 @@ def uiData_vecStore(userFiles, userUrls, api_key_st, vsDict_st={}, progress=gr.P
         embeddings = getEmbeddingFunc(api_key_st)
     except Exception as e:
         gr.Error(str(e))
-        return {}, '', *[x.update() for x in opComponents]
+        return {}, '', *[gr.update() for x in opComponents]
     
     progress(0.5, 'Creating Vector Database')
     vsDict_st = getVsDict(embeddings, docs, vsDict_st)
@@ -131,7 +116,9 @@ def uiData_vecStore(userFiles, userUrls, api_key_st, vsDict_st={}, progress=gr.P
     src_str = str(src_str[1]) + ' source document(s) successfully loaded in vector store.'+'\n\n' + src_str[0]
     
     progress(1, 'Data loaded')
-    return vsDict_st, src_str, *[x.update(interactive=False) for x in [data_ingest_btn, upload_fb]], urls_tb.update(interactive=False, placeholder=''), initChatbot_btn.update(interactive=True)
+    for file in file_paths:
+        os.remove(file)
+    return vsDict_st, src_str, *[gr.update(interactive=False) for x in [data_ingest_btn, upload_fb]], gr.update(interactive=False, placeholder=''), gr.update(interactive=True)
 
 # initialize chatbot function sets the QA Chain, and also sets/updates any other components to start chatting. updateQaChain function only updates QA chain and will be called whenever Adv Settings are updated.
 def initializeChatbot(temp, k, modelNameDD, stdlQs, api_key_st, vsDict_st, progress=gr.Progress()):
@@ -146,8 +133,8 @@ def initializeChatbot(temp, k, modelNameDD, stdlQs, api_key_st, vsDict_st, progr
         welMsg = welcomeMsgDefault
     print('Chatbot initialized at ', datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S'))
 
-    return qa_chain_st, chainTuple[1], btn.update(interactive=True), initChatbot_btn.update('Chatbot ready. Now visit the chatbot Tab.', interactive=False)\
-        , status_tb.update(), gr.Tabs.update(selected='cb'), chatbot.update(value=[('', welMsg)])
+    return qa_chain_st, chainTuple[1], gr.update(interactive=True), gr.update('Chatbot ready. Now visit the chatbot Tab.', interactive=False)\
+        , gr.update(), gr.update(selected='cb'), gr.update(value=[('', welMsg)])
 
 # just update the QA Chain, no updates to any UI
 def updateQaChain(temp, k, modelNameDD, stdlQs, api_key_st, vsDict_st):
@@ -171,7 +158,7 @@ def updateQaChain(temp, k, modelNameDD, stdlQs, api_key_st, vsDict_st):
     # settingsUpdated = 'Settings updated:'+ ' Model=' + modelName + ', Temp=' + str(temp)+ ', k=' + str(k)
     # gr.Info(settingsUpdated)
     
-    if 'meta-llama/llama-2' in modelNameDD:
+    if 'meta-llama/' in modelNameDD:
         prompt = promptLlama
     else:
         prompt = None
@@ -193,10 +180,10 @@ def updateQaChain(temp, k, modelNameDD, stdlQs, api_key_st, vsDict_st):
                     rephrase_question=rephQs,
                     return_source_documents=True,
                     return_generated_question=True,
-                    combine_docs_chain_kwargs={'prompt':promptLlama}
+                    combine_docs_chain_kwargs={'prompt':prompt}
                 )
     
-    return qa_chain_st, model_dd.update(value=modelNameDD)
+    return qa_chain_st, gr.update(value=modelNameDD)
         
 
 def respond(message, chat_history, qa_chain):
@@ -206,10 +193,10 @@ def respond(message, chat_history, qa_chain):
     streaming_answer = ""
     for ele in "".join(result['answer']):
         streaming_answer += ele
-        yield "", chat_history + [(message, streaming_answer)], src_docs, btn.update('Please wait...', interactive=False)
+        yield "", chat_history + [(message, streaming_answer)], src_docs, gr.update('Please wait...', interactive=False)
     
     chat_history.extend([(message, result['answer'])])
-    yield "", chat_history, src_docs, btn.update('Send Message', interactive=True)
+    yield "", chat_history, src_docs, gr.update('Send Message', interactive=True)
 
 #####################################################################################################
 
@@ -243,11 +230,11 @@ with gr.Blocks(theme=gr.themes.Default(primary_hue='orange', secondary_hue='gray
                             , info='Internal IBMers only')
                     bamKey_btn = gr.Button("Submit BAM API Key")
             with gr.Row(visible=mode.uiAddDataVis):
-                upload_fb = gr.Files(scale=5, label="Upload (multiple) Files - pdf/txt/docx supported", file_types=['.doc', '.docx', 'text', '.pdf', '.csv', '.ppt', '.pptx'])
-                urls_tb = gr.Textbox(scale=5, label="Enter URLs starting with https (comma separated)"\
+                upload_fb = gr.Files(scale=1, label="Upload (multiple) Files - pdf/txt/docx supported", file_types=['.doc', '.docx', 'text', '.pdf', '.csv', '.ppt', '.pptx'])
+                urls_tb = gr.Textbox(scale=1, label="Enter URLs starting with https (comma separated)"\
                                     , info=url_tb_info\
                                     , placeholder=url_tb_ph)
-                data_ingest_btn = gr.Button("Load Data")
+                data_ingest_btn = gr.Button(scale=0, value="Load Data")
             status_tb = gr.TextArea(label='Status Info')
             initChatbot_btn = gr.Button("Initialize Chatbot", variant="primary", interactive=False)
 
@@ -268,7 +255,7 @@ with gr.Blocks(theme=gr.themes.Default(primary_hue='orange', secondary_hue='gray
                         temp_sld = gr.Slider(minimum=0, maximum=1, step=0.1, value=0.7, label="Temperature", info='Sampling temperature to use when calling LLM. Defaults to 0.7')
                         k_sld = gr.Slider(minimum=1, maximum=10, step=1, value=mode.k, label="K", info='Number of relavant documents to return from Vector Store. Defaults to 4')
                         model_dd = gr.Dropdown(label='Model Name'\
-                                , choices=getModelChoices(openAi_models, ModelTypes, bam_models_old), allow_custom_value=True\
+                                , choices=getModelChoices(openAi_models_def, ModelTypes, bam_models_def), allow_custom_value=True\
                                 , info=model_dd_info)
                     stdlQs_rb = gr.Radio(label='Standalone Question', info=stdlQs_rb_info\
                             , type='index', value=stdlQs_rb_choices[1]\
@@ -277,7 +264,7 @@ with gr.Blocks(theme=gr.themes.Default(primary_hue='orange', secondary_hue='gray
     ### Setup the Gradio Event Listeners
 
     # OpenAI API button
-    oaiKey_btn_args = {'fn':setOaiApiKey, 'inputs':[oaiKey_tb], 'outputs':credComps_op}
+    oaiKey_btn_args = {'fn':setOaiApiKey, 'inputs':[oaiKey_tb], 'outputs':credComps_op+[model_dd]}
     oaiKey_btn.click(**oaiKey_btn_args)
     oaiKey_tb.submit(**oaiKey_btn_args)
 
@@ -311,5 +298,5 @@ with gr.Blocks(theme=gr.themes.Default(primary_hue='orange', secondary_hue='gray
     btn.click(**chat_btn_args)
     msg.submit(**chat_btn_args)
 
-demo.queue(concurrency_count=10)
-demo.launch(show_error=True)
+demo.queue(default_concurrency_limit=10)
+demo.launch(show_error=True, auth=(os.getenv("USERNAME",''), os.getenv("PASSWORD",'')))

ttyd_consts.py

@@ -1,4 +1,4 @@
-from langchain import PromptTemplate
+from langchain.prompts import PromptTemplate
 import os
 from dotenv import load_dotenv
 load_dotenv()
@@ -44,33 +44,9 @@ Question: {question} [/INST]
 
 promptLlama=PromptTemplate(input_variables=['context', 'question'], template=llamaPromptTemplate)
 
-bam_models_old = sorted(['bigscience/bloom',
- 'salesforce/codegen2-16b',
- 'codellama/codellama-34b-instruct',
- 'tiiuae/falcon-40b',
- 'ibm/falcon-40b-8lang-instruct',
- 'google/flan-t5-xl',
- 'google/flan-t5-xxl',
- 'google/flan-ul2',
- 'eleutherai/gpt-neox-20b',
- 'togethercomputer/gpt-neoxt-chat-base-20b',
- 'ibm/granite-13b-sft',
- 'ibm/granite-13b-sft-cft',
- 'ibm/granite-3b-code-v1',
- 'meta-llama/llama-2-13b',
- 'meta-llama/llama-2-13b-chat',
- 'meta-llama/llama-2-13b-chat-beam',
- 'meta-llama/llama-2-70b',
- 'meta-llama/llama-2-70b-chat',
- 'meta-llama/llama-2-7b',
- 'meta-llama/llama-2-7b-chat',
- 'mosaicml/mpt-30b',
- 'ibm/mpt-7b-instruct',
- 'bigscience/mt0-xxl',
- 'bigcode/starcoder',
- 'google/ul2'])
-
-openAi_models = ['gpt-3.5-turbo (openai)', 'gpt-3.5-turbo-16k (openai)', 'gpt-4 (openai)', 'text-davinci-003 (Legacy - openai)', 'text-curie-001 (Legacy - openai)', 'babbage-002 (openai)']
+bam_models_def = ['bigscience/bloom', 'meta-llama/llama-2-13b', 'meta-llama/llama-2-13b-chat', 'add credentials to see full list of models']
+
+openAi_models_def = ['gpt-3.5-turbo', 'gpt-4', 'text-davinci-003', 'add credentials to see full list of models']
 
 model_dd_info = 'Make sure your credentials are submitted before changing the model. You can also input any OpenAI model name or Watsonx/BAM model ID.' 
 

ttyd_functions.py

@@ -3,7 +3,6 @@ import datetime
 import gradio as gr
 import time
 import uuid
-import openai
 from langchain.embeddings import OpenAIEmbeddings
 from langchain.vectorstores import Chroma
 from langchain.text_splitter import RecursiveCharacterTextSplitter
@@ -24,7 +23,7 @@ import tiktoken
 import gdown
 
 from langchain.chat_models import ChatOpenAI
-from langchain import OpenAI
+from langchain.llms import OpenAI
 
 from ibm_watson_machine_learning.metanames import GenTextParamsMetaNames as GenParams
 from ibm_watson_machine_learning.foundation_models.utils.enums import DecodingMethods
@@ -32,7 +31,7 @@ from ibm_watson_machine_learning.foundation_models import Model
 from ibm_watson_machine_learning.foundation_models.extensions.langchain import WatsonxLLM
 
 
-import genai
+from genai.credentials import Credentials
 from genai.extensions.langchain import LangChainInterface
 from genai.schemas import GenerateParams
 
@@ -53,7 +52,7 @@ def getOaiCreds(key):
 def getBamCreds(key):
     key = key if key else 'Null'
     return {'service': 'bam',
-                'bam_creds' : genai.Credentials(key, api_endpoint='https://bam-api.res.ibm.com/v1')
+                'bam_creds' : Credentials(key, api_endpoint='https://bam-api.res.ibm.com/v1')
             }
 
 
@@ -257,7 +256,7 @@ def data_ingestion(inputDir=None, file_list=[], url_list=[], gDriveFolder='', pr
     documents = []
     # Ingestion from Google Drive Folder
     if gDriveFolder:
-        opFolder = './gDriveDocs/'
+        opFolder = './tmp/gDriveDocs/'
         gdown.download_folder(url=gDriveFolder, output=opFolder, quiet=True)
         files = [str(x) for x in Path(opFolder).glob('**/*')]
         documents = ingestFiles(documents, files, prog)
@@ -325,7 +324,7 @@ def getEmbeddingFunc(creds):
         elif creds.get('service')=='watsonx' or creds.get('service')=='bam':
             # testModel = Model(model_id=ModelTypes.FLAN_UL2, credentials=creds['credentials'], project_id=creds['project_id']) # test the API key
             # del testModel
-            embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2") # for now use OpenSource model for embedding as WX doesnt have any embedding model
+            embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2", cache_folder='./tmp/hfEmbModel') # for now use OpenSource model for embedding as WX doesnt have any embedding model
         else:
             raise Exception('Error: Invalid or None Credentials')
         return embeddings
@@ -333,7 +332,7 @@ def getEmbeddingFunc(creds):
 def getVsDict(embeddingFunc, docs, vsDict={}):
     # create chroma client if doesnt exist
     if vsDict.get('chromaClient') is None:
-        vsDict['chromaDir'] = './vecstore/'+str(uuid.uuid1())
+        vsDict['chromaDir'] = './tmp/vecstore/'+str(uuid.uuid1())
         vsDict['chromaClient'] = Chroma(embedding_function=embeddingFunc, persist_directory=vsDict['chromaDir'])
     # clear chroma client before adding new docs
     if vsDict['chromaClient']._collection.count()>0:
@@ -374,4 +373,4 @@ def changeModel(oldModel, newModel):
     return newModel
 
 def getModelChoices(openAi_models, wml_models, bam_models):
-    return [model for model in openAi_models] + [model.value+' (watsonx)' for model in wml_models] + [model + ' (bam)' for model in bam_models]
+    return [model +' (openai)' for model in openAi_models] + [model.value +' (watsonx)' for model in wml_models] + [model + ' (bam)' for model in bam_models]