llm query work.

semsearch.py

@@ -295,48 +295,6 @@ if not client.collections.exists("Chunks") :
           }
         )
 
-###############################################################################
-# text contains prompt for vector DB.
-text = "human-made computer cognitive ability"
-
-
-###############################################################################
-# Initial the the sentence transformer and encode the query prompt.
-logger.info(f"#### Encode text query prompt to create vectors. {text}")
-model = SentenceTransformer('/app/multi-qa-MiniLM-L6-cos-v1')
-
-vector = model.encode(text)
-vectorList = []
-
-logger.debug("#### Print vectors.")
-for vec in vector:
-    vectorList.append(vec)
-logger.debug(f"vectorList: {vectorList[2]}")
-
-# Fetch chunks and print chunks.
-logger.info("#### Retrieve semchunks from db using vectors from prompt.")
-semChunks = wpChunkCollection.query.near_vector(
-    near_vector=vectorList,
-    distance=0.7,
-    limit=3
-)
-logger.debug(f"### semChunks[0]: {semChunks}")
-
-# Print chunks, corresponding document and document title.
-logger.info("#### Print individual retrieved chunks.")
-for chunk in enumerate(semChunks.objects):
-    logger.info(f"#### chunk: {chunk}")
-    webpage_uuid = chunk[1].properties['references']['webpage']
-    logger.info(f"webpage_uuid: {webpage_uuid}")
-    wpFromChunk = wpCollection.query.fetch_object_by_id(webpage_uuid)
-    logger.info(f"### wpFromChunk title: {wpFromChunk.properties['title']}")
-
-
-
-####################################################################
-# 
-collection = client.collections.get("Chunks")
-#model = SentenceTransformer('../multi-qa-MiniLM-L6-cos-v1')
 
 #################################################################
 # Initialize the LLM.
@@ -383,6 +341,43 @@ llm = Llama(model_path,
            )
 
 
+def getRagData(promptText):
+    ###############################################################################
+    # Initial the the sentence transformer and encode the query prompt.
+    logger.info(f"#### Encode text query prompt to create vectors. {text}")
+    model = SentenceTransformer('/app/multi-qa-MiniLM-L6-cos-v1')
+
+    vector = model.encode(promptText)
+    vectorList = []
+
+    logger.debug("#### Print vectors.")
+    for vec in vector:
+        vectorList.append(vec)
+    logger.debug(f"vectorList: {vectorList[2]}")
+
+    # Fetch chunks and print chunks.
+    logger.info("#### Retrieve semchunks from db using vectors from prompt.")
+    semChunks = wpChunkCollection.query.near_vector(
+        near_vector=vectorList,
+        distance=0.7,
+        limit=3
+    )
+    logger.debug(f"### semChunks[0]: {semChunks}")
+
+    # Print chunks, corresponding document and document title.
+    ragData = ""
+    logger.info("#### Print individual retrieved chunks.")
+    for chunk in enumerate(semChunks.objects):
+        logger.info(f"#### chunk: {chunk}")
+        ragData = ragData + "\n" + chunk[0]
+        webpage_uuid = chunk[1].properties['references']['webpage']
+        logger.info(f"webpage_uuid: {webpage_uuid}")
+        wpFromChunk = wpCollection.query.fetch_object_by_id(webpage_uuid)
+        logger.info(f"### wpFromChunk title: {wpFromChunk.properties['title']}")
+    #collection = client.collections.get("Chunks")
+    return  ragData
+
+# Display UI
 display(systemTextArea)
 display(userTextArea)
 display(ragPromptTextArea)
@@ -390,25 +385,38 @@ display(responseTextArea)
 display(selectRag)
 display(submitButton)
 
+def runLLM(prompt):
+    max_tokens = 1000
+    temperature = 0.3
+    top_p = 0.1
+    echo = True
+    stop = ["Q", "\n"]
+
+    modelOutput = llm(
+       prompt,
+       max_tokens=max_tokens,
+       temperature=temperature,
+       top_p=top_p,
+       echo=echo,
+       stop=stop,
+   )
+   result = modelOutput["choices"][0]["text"].strip()
+   return(result)
+
 def setPrompt(pprompt,ragFlag):
     print("\n### setPrompt() entered. ragFlag: ",ragFlag)
     if ragFlag:
-        ragPrompt = setRagPrompt(pprompt)
+        ragPrompt = getRagData(pprompt)
         userPrompt = pprompt + "\n" + ragPrompt
         prompt = userPrompt
+        userPrompt = "Using this information: " + ragPrompt
+             + "process the following statement or question and produce a a response"
+             + intialPrompt
     else:
         userPrompt = pprompt
-    prompt = f""" <s> [INST] <<SYS>> {systemTextArea.value} </SYS>> Q: {userPrompt} A: [/INST]"""
-    return prompt
-
-def runModel(prompt):
-    output = llm.create_completion(
-                  prompt, # Prompt
-                  max_tokens=4096, # Generate up to 32 tokens
-                  #stop = ["Q:", "\n"], # Stop generating just before the model would generate a new question
-                  echo = False # Echo the prompt back in the output
-                )
-    responseTextArea.value = output["choices"][0]["text"]
+    #prompt = f""" <s> [INST] <<SYS>> {systemTextArea.value} </SYS>> Q: {userPrompt} A: [/INST]"""
+    return userPrompt
+
 
 def on_submitButton_clicked(b):
     with output_widget:
@@ -418,7 +426,7 @@ def on_submitButton_clicked(b):
         log.debug(f"### selectRag: {selectRag.value}")
         prompt = setPrompt(userTextArea.value,selectRag.value)
         log.debug("### prompt: " + prompt)
-        runModel(prompt)
+        runLLM(prompt)
 
 submitButton.on_click(on_submitButton_clicked)
 display(output_widget)

startup.sh

@@ -56,13 +56,13 @@ env
 /app/weaviate/weaviate --host 127.0.0.1 --port 8080 --scheme http --write-timeout 600s &  #2>& 1 | tee /data/var/lib/weaviate/ws.log &
 
 echo "#### Before sleep."
-sleep 60
+sleep 30
 
 echo "#### Before /app/semsearch.py"
 python /app/semsearch.py &  #2>& 1 | tee /data/var/lib/weaviate/ss.log &
 
 # Display timestamps.
-for (( ; ; )) do date; sleep 60; done &
+#for (( ; ; )) do date; sleep 60; done &
 
 wait