add more llms

app.py

@@ -18,6 +18,9 @@ st.set_page_config(page_title="Summary", layout="wide")
 
 # Initialize chat history
 chat_history = []
+# Initialize variables for LLM options and chosen LLM
+llm_options = []
+chosen_LLM = "default"
 
 logger.info(f"Build retriever")
 
@@ -29,8 +32,8 @@ def generate_response(prompt_input):
 
 
 @st.cache_resource()
-def factory_transcript(video_id, model):
-    ts = TranscriptSummary(doc_id=video_id, model=model)
+def factory_transcript(video_id, model, llm_provider):
+    ts = TranscriptSummary(doc_id=video_id, model=model, llm_provider=llm_provider)
     logger.info("TranscriptSummary initialized")
     return ts
 
@@ -64,11 +67,38 @@ with st.sidebar:
 
     similarity_top_k = st.number_input("Maximum Number of Results to Display", min_value=1, max_value=100, value=10)
 
-    chosen_LLM = st.selectbox("Choose Language Model", ["gpt-3.5-turbo", "default"])
+    # Selecting the provider
+    chosen_provider = st.selectbox("Choose Provider", ["OpenAI", "Replicate", "Default"])
+
+    # Based on provider, display relevant LLMs
+    if chosen_provider == "OpenAI":
+        llm_options = ["gpt-3.5-turbo-0301", "gpt-3.5-turbo-16k", "gpt-4", "gpt-4-32k-0314"]
+    elif chosen_provider == "Replicate":
+        llm_options = ["mistralai/mistral-7b-v0.1:3e8a0fb6d7812ce30701ba597e5080689bef8a013e5c6a724fafb108cc2426a0",
+                       "mistralai/mistral-7b-instruct-v0.1:83b6a56e7c828e667f21fd596c338fd4f0039b46bcfa18d973e8e70e455fda70",
+                       "joehoover/zephyr-7b-alpha:14ec63365a1141134c41b652fe798633f48b1fd28b356725c4d8842a0ac151ee",
+                       "meta/llama-2-13b-chat:f4e2de70d66816a838a89eeeb621910adffb0dd0baba3976c96980970978018d",
+                       "meta/llama-2-70b-chat:02e509c789964a7ea8736978a43525956ef40397be9033abf9fd2badfe68c9e3",
+                       "meta/llama-2-7b-chat:8e6975e5ed6174911a6ff3d60540dfd4844201974602551e10e9e87ab143d81e", ]
+
+    else:
+        llm_options = ["default"]
+
+    # Allow users to type a custom LLM or choose from list
+    chosen_LLM = st.selectbox("Type or Choose Language Model", llm_options)
+
     api_key = st.text_input("OpenAI API Key", type="password")
 
-if api_key:
-    openai.api_key = api_key
+if api_key and chosen_provider == "OpenAI":
+    logger.info("OpenAI API KEY")
+    try:
+        openai.api_key = api_key
+    except:
+        st.sidebar.write("Incorrect API key provided")
+if api_key and chosen_provider == "Replicate":
+    logger.info("Replicate API KEY")
+
+    os.environ['REPLICATE_API_TOKEN'] = api_key
 
 if youtube_url or uploaded_video:
     video_file_path = os.path.join(output_path_video, f"{video_loader.video_id}.mp3")
@@ -84,7 +114,7 @@ if youtube_url or uploaded_video:
         logger.info(f"Transcription already exists: {transcription_file_path}")
 
     video_retriever = factory_video(video_loader.video_id, top_k=int(similarity_top_k))
-    transcript_summary = factory_transcript(video_loader.video_id, model=chosen_LLM)
+    transcript_summary = factory_transcript(video_loader.video_id, model=chosen_LLM, llm_provider=chosen_provider)
 
     docs = load_transcription(video_loader, output_path_transcription)
 
@@ -144,10 +174,17 @@ if youtube_url or uploaded_video:
             del st.session_state.generated[:]
 
 
+        # def on_input_change():
+        #     user_input = st.session_state.user_input
+        #     st.session_state.past.append(user_input)
+        #     st.session_state.generated.append("The messages from Bot\nWith new line")
+
+
         def on_input_change():
             user_input = st.session_state.user_input
             st.session_state.past.append(user_input)
             st.session_state.generated.append("The messages from Bot\nWith new line")
+            st.session_state.user_input = ""  # This will empty the "User Input:" text box
 
 
         if 'generated' not in st.session_state:

summarization_service/summarizer.py

@@ -3,29 +3,30 @@ import os
 
 from llama_index import (ServiceContext, get_response_synthesizer, Document)
 from llama_index.indices.document_summary import DocumentSummaryIndex
-from llama_index.llms import OpenAI
+from llama_index.llms import OpenAI, Replicate
 
 from config import output_path_transcription
 from logger import logger
 
 
 class TranscriptSummary:
-    def __init__(self, doc_id, model):
+    def __init__(self, doc_id, model="default", llm_provider=None):
         self.doc_id = doc_id
         self.document = self.load_json_file_and_extract_text()
         self.model = model
+        self.llm_provider = llm_provider
 
         # Check for the presence of the openai_key
-        if "gpt" in self.model and "OPENAI_KEY" in os.environ:
-            try:
-                self.llm = OpenAI(temperature=0, model=self.model)
-            except:
-                self.llm = "default"
-
+        if self.llm_provider == "OpenAI":
+            self.llm = OpenAI(model=self.model, temperature=0.9)
+        elif self.llm_provider == "Replicate":
+            self.llm = Replicate(model=self.model)
         else:
             self.llm = "default"
+
         # logger.info(f"LLM ")
-        logger.info(f"LLM used: {self.llm}"*50)
+        logger.info(f"Model used: {self.model}")
+        logger.info(f"LLM Provider: {self.llm_provider}")
 
         self.service_context = ServiceContext.from_defaults(llm=self.llm, chunk_size=1024)
         self.response_synthesizer = get_response_synthesizer(response_mode="tree_summarize", use_async=True)