Deploy dashboard for V3 to cloudFront and Fix authentication

src/workspaces/cookiefactoryv3/assistant/.gitignore

@@ -1,8 +1,8 @@
 .langchain.db
-
+# Dockerfile
 .env
 .python-version
-
+test_run.sh
 *.swp
 package-lock.json
 __pycache__

src/workspaces/cookiefactoryv3/assistant/Dockerfile

@@ -0,0 +1,31 @@
+FROM --platform=linux/amd64 node:20-slim
+RUN npm install -g pnpm
+
+# Install Python and other dependencies
+RUN apt-get update && apt-get install -y \
+    python3 \
+    python3-pip \
+    git \
+    curl \
+    unzip \
+    jq
+
+# Install AWS CLI
+RUN curl "https://awscli.amazonaws.com/awscli-exe-linux-x86_64.zip" -o "awscliv2.zip" && \
+    unzip awscliv2.zip && \
+    ./aws/install && \
+    rm -rf awscliv2.zip aws
+
+
+# Copy your application files
+WORKDIR /app
+COPY . .
+
+# Install dependencies using your install.sh script
+RUN sh install.sh
+
+# Expose the port Chainlit will listen on 
+EXPOSE 8000  
+
+# Start the Chainlit app using your run.sh script
+CMD ["sh", "run.sh"] 

src/workspaces/cookiefactoryv3/assistant/app/bedrock.py

@@ -19,17 +19,12 @@
 from chainlit.context import context
 
 from lib.router import LLMRouterChain, MultiRouteChain, create_routes
-from lib.llm import get_bedrock_text
+from lib.llm import get_bedrock_text_v3_sonnet
 from lib.context_memory import EntityContextMemory
-from lib.tools.qa import init_db
 from lib.initial_diagnosis import InitialDiagnosisChain
 
-print('initializing vector db')
-
-init_db(os.path.join(os.path.dirname(__file__), '..', 'public', './freezer_tunnel_manual.pdf'))
-
-print('done initializing vector db')
-
+## To be implemented with update to chainlit
+#from chainlit.oauth_providers import AWSCognitoOAuthProvider 
 
 welcome_message="""
 Hi, I'm the AI assistant of the Cookie Factory. I'm here to help you diagnose and resolve issues \
@@ -41,15 +36,28 @@
 There is an ongoing event [#{event_id}](https://example.com/issue/{event_id}). Do you want to run an initial diagnosis of the issue?
 """
 
+## To be implemented with update to chainlit
+#cl.oauth_providers = [
+#    AWSCognitoOAuthProvider(
+#        client_id=os.environ["OAUTH_COGNITO_CLIENT_ID"],
+#        client_secret=os.environ["OAUTH_COGNITO_CLIENT_SECRET"],
+#        domain=os.environ["OAUTH_COGNITO_DOMAIN"],
+#    )
+#]
 
 @cl.on_chat_start
 async def start():
+
+  ## To be implemented with udpdate to chainlit
+  #if not cl.user_session.get("is_authenticated"):
+  #      await cl.oauth_providers[0].authorize()
+
   LLMRouterChain.update_forward_refs()
   MultiRouteChain.update_forward_refs()
 
   memory = EntityContextMemory()
   routes = create_routes(memory)
-  llm_chain = MultiRouteChain.from_prompts(llm=get_bedrock_text(), prompt_infos=routes)
+  llm_chain = MultiRouteChain.from_prompts(llm=get_bedrock_text_v3_sonnet(), prompt_infos=routes)
 
   cl.user_session.set("chain", llm_chain)
 
@@ -71,7 +79,7 @@ async def start():
 async def main(message, context):
   llm_chain = cl.user_session.get("chain")
 
-  res = await llm_chain.acall(
+  res = await llm_chain.ainvoke(
     message,
     callbacks=[cl.AsyncLangchainCallbackHandler()])
 
@@ -88,9 +96,9 @@ async def on_action(action):
   await action.remove()
 
   cb = cl.AsyncLangchainCallbackHandler()
-  chain = InitialDiagnosisChain.from_llm(llm=get_bedrock_text())
+  chain = InitialDiagnosisChain.from_llm(llm=get_bedrock_text_v3_sonnet())
 
-  res = await chain.acall({
+  res = await chain.ainvoke({
     'event_title': event_title,
     'event_description': event_description,
     'event_timestamp': event_timestamp
@@ -105,6 +113,7 @@ async def on_action(action):
 @cl.action_callback("agent_actions")
 async def on_action(action):
   event_entity_id = context.session.user_data.get('event_entity_id')
+
   if action.value == "3d":
     await cl.Message(content=f"Navigating to the issue site.").send()
     await action.remove()

src/workspaces/cookiefactoryv3/assistant/app/lib/env.py

@@ -11,3 +11,6 @@ def get_bedrock_region():
 
 def get_workspace_id():
   return os.getenv("WORKSPACE_ID")
+
+def get_knowledge_base_id():
+  return os.getenv("KNOWLEDGE_BASE_ID")

src/workspaces/cookiefactoryv3/assistant/app/lib/initial_diagnosis.py

@@ -5,7 +5,7 @@
 
 from typing import Any, Dict, List, Optional
 
-from langchain import PromptTemplate
+from langchain_core.prompts import PromptTemplate
 from langchain.chains import LLMChain
 
 from langchain.callbacks.manager import (
@@ -15,15 +15,15 @@
 from langchain.chains.base import Chain
 from langchain.schema.language_model import BaseLanguageModel
 
-from .llm import get_bedrock_text, get_processed_prompt_template
+from .llm import get_bedrock_text_v3_sonnet, get_processed_prompt_template_sonnet
 
-default_llm = get_bedrock_text()
+default_llm = get_bedrock_text_v3_sonnet()
 
 question_classifier_prompt = """
 You are a technical assistant to help the cooke line operators to investigate product quality issues. \
 Your task is take the "Collected Information" from alarm systems, summarize the issue and provide prescriptive suggestions as "Initial Diagnosis" based on \
 your knowledge about cookie production to provide initial suggestions for line operators to investigate the issue. Be concise and professional in the response. \
-Translate technical terms to business terms so it's easy for line operators to read and understand, for example, timestamps should be converted to local user friendly format.
+Translate technical terms to business terms so it's easy for line operators to read and understand, for example, timestamps should be converted to local user friendly format and use today's date instead of the one provided for October 23rd 2023.
 
 <example>
 Collected information
@@ -95,7 +95,7 @@ def from_llm(
     ) -> InitialDiagnosisChain:
         router_template = question_classifier_prompt
         router_prompt = PromptTemplate(
-            template=get_processed_prompt_template(router_template),
+            template=get_processed_prompt_template_sonnet(router_template),
             input_variables=["event_title", "event_description", "event_timestamp"],
         )
         llm_chain = LLMChain(llm=llm, prompt=router_prompt)

src/workspaces/cookiefactoryv3/assistant/app/lib/llm.py

@@ -3,25 +3,19 @@
 
 import boto3
 
-from langchain.llms.bedrock import Bedrock
+#from langchain.llms.bedrock import Bedrock
+from langchain_community.chat_models import BedrockChat
 from langchain.embeddings.bedrock import BedrockEmbeddings
 
 from botocore.config import Config
 
 from .env import get_bedrock_region
 
-available_models = [
-    "amazon.titan-tg1-large",
-    "anthropic.claude-v2",
-    "ai21.j2-ultra",
-    "ai21.j2-mid",
-    "anthropic.claude-instant-v1",
-    "anthropic.claude-v1"
-]
-
 # the current model used for text generation
 text_model_id = "anthropic.claude-instant-v1"
+text_v3_haiku_model_id = "anthropic.claude-3-haiku-20240307-v1:0"
 text_v2_model_id = "anthropic.claude-v2"
+text_v3_sonnet_model_id = "anthropic.claude-3-sonnet-20240229-v1:0"
 embedding_model_id = "amazon.titan-embed-text-v1"
 
 model_kwargs = {
@@ -34,6 +28,22 @@
         "temperature": 0.1,
         "top_p": 0.9,
     },
+    "anthropic.claude-3-sonnet-20240229-v1:0": {
+        "anthropic_version": "bedrock-2023-05-31",
+        "max_tokens": 2048,
+        "temperature": 0.0,
+        "top_k": 250,
+        "top_p": 1,
+        "stop_sequences": ["\n\nHuman"]
+    },
+    "anthropic.claude-3-haiku-20240307-v1:0": {
+        "anthropic_version": "bedrock-2023-05-31",
+        "max_tokens": 2048,
+        "temperature": 0.0,
+        "top_k": 250,
+        "top_p": 1,
+        "stop_sequences": ["\n\nHuman"]
+    },
     "anthropic.claude-instant-v1": {
         "max_tokens_to_sample": 2048,
         "temperature": 0.1,
@@ -42,11 +52,15 @@
 }
 
 prompt_template_prefix = {
+    "anthropic.claude-3-haiku-20240307-v1:0": "\n\nHuman: ",
+    "anthropic.claude-3-sonnet-20240229-v1:0": "\n\nHuman: ",
     "anthropic.claude-v2": "\n\nHuman: ",
     "anthropic.claude-instant-v1": "\n\nHuman: "
 }
 
 prompt_template_postfix = {
+    "anthropic.claude-3-haiku-20240307-v1:0": "\n\nAssistant: ",
+    "anthropic.claude-3-sonnet-20240229-v1:0": "\n\nAssistant:",
     "anthropic.claude-v2": "\n\nAssistant:",
     "anthropic.claude-instant-v1": "\n\nAssistant:"
 }
@@ -57,6 +71,8 @@ def template_proc(template):
     return template_proc
 
 prompt_template_procs = {
+    "anthropic.claude-3-haiku-20240307-v1:0": get_template_proc("anthropic.claude-3-haiku-20240307-v1:0"),
+    "anthropic.claude-3-sonnet-20240229-v1:0": get_template_proc("anthropic.claude-3-sonnet-20240229-v1:0"),
     "anthropic.claude-v2": get_template_proc("anthropic.claude-v2"),
     "anthropic.claude-instant-v1": get_template_proc("anthropic.claude-instant-v1")
 }
@@ -73,16 +89,34 @@ def template_proc(template):
         'mode': 'standard'
     }
 ))
+
+bedrock_agents = boto3.client('bedrock-agent-runtime', get_bedrock_region(), config=Config(
+    retries = {
+        'max_attempts': 10,
+        'mode': 'standard'
+    }
+))
+
 response = bedrock.list_foundation_models()
 print(response.get('modelSummaries')) 
 
 def get_bedrock_text():
-    llm = Bedrock(model_id=text_model_id, client=bedrock_runtime)
+    llm = BedrockChat(model_id=text_model_id, streaming=True, client=bedrock_runtime)
     llm.model_kwargs = model_kwargs.get(text_model_id, {})
     return llm
 
+def get_bedrock_text_v3_haiku():
+    llm = BedrockChat(model_id=text_v3_haiku_model_id, streaming=True, client=bedrock_runtime)
+    llm.model_kwargs = model_kwargs.get(text_v3_haiku_model_id, {})
+    return llm
+
+def get_bedrock_text_v3_sonnet():
+    llm = BedrockChat(model_id=text_v3_sonnet_model_id, streaming=True,  client=bedrock_runtime)
+    llm.model_kwargs = model_kwargs.get(text_v3_sonnet_model_id, {})
+    return llm
+
 def get_bedrock_text_v2():
-    llm = Bedrock(model_id=text_v2_model_id, client=bedrock_runtime)
+    llm = BedrockChat(model_id=text_v2_model_id, streaming=True,  client=bedrock_runtime)
     llm.model_kwargs = model_kwargs.get(text_v2_model_id, {})
     return llm
 
@@ -110,3 +144,21 @@ def get_postfix_prompt_template(template):
         return template + prompt_template_postfix[text_model_id]
     else:
         return template
+
+def get_processed_prompt_template_sonnet(template):
+    if text_v3_sonnet_model_id in prompt_template_procs:
+        return prompt_template_procs[text_v3_sonnet_model_id](template)
+    else:
+        return template
+
+def get_prefix_prompt_template_sonnet(template):
+    if text_v3_sonnet_model_id in prompt_template_prefix:
+        return prompt_template_prefix[text_v3_sonnet_model_id] + template
+    else:
+        return template
+
+def get_postfix_prompt_template_sonnet(template):
+    if text_v3_sonnet_model_id in prompt_template_postfix:
+        return template + prompt_template_postfix[text_v3_sonnet_model_id]
+    else:
+        return template

src/workspaces/cookiefactoryv3/assistant/app/lib/router.py

@@ -5,7 +5,7 @@
 
 from typing import Any, Dict, List, Mapping, NamedTuple, Optional
 
-from langchain import PromptTemplate
+from langchain_core.prompts import PromptTemplate
 from langchain.chains import LLMChain
 
 from langchain.callbacks.manager import (
@@ -21,10 +21,11 @@
 from .tools.qa import QAChain
 from .tools.graph import GraphChain
 from .tools.general import GeneralChain
+from .tools.inspect_sensor_data import InspectChain
 
-from .llm import get_bedrock_text, get_processed_prompt_template
+from .llm import get_bedrock_text_v3_sonnet, get_processed_prompt_template_sonnet
 
-default_llm = get_bedrock_text()
+default_llm = get_bedrock_text_v3_sonnet()
 
 question_classifier_prompt = """
 You are given an instruction. The instruction is either a command or a question. You need to decide the type of the instruction provided by user. 
@@ -34,6 +35,7 @@
 - 3dview: if the instruction is a command about manipulating the 3D viewer
 - doc: if the instruction is a question about standard procedures in the knowledge base
 - graph: if the instruction is a question about finding information of the entities in the factory
+- inspect: if the instruction is a question about analyzing historical data for a freezer
 - general: if none of the above applies
 
 You must give an answer using one of the valid options, and you should write out the answer without further explanation.
@@ -46,6 +48,9 @@
 Instruction: how to operate the cookie line?
 Answer: doc
 
+Instruction: Can you analyze the last 15 minutes of data for the freezer tunnel for any potential issues?
+Answer: inspect
+
 Instruction: what are the potential causes of the inconsistent cookie shape?
 Answer: general
 
@@ -134,7 +139,7 @@ def from_llm(
     ) -> LLMRouterChain:
         router_template = question_classifier_prompt
         router_prompt = PromptTemplate(
-            template=get_processed_prompt_template(router_template),
+            template=get_processed_prompt_template_sonnet(router_template),
             input_variables=["question"],
         )
         llm_chain = LLMChain(llm=llm, prompt=router_prompt)
@@ -224,6 +229,10 @@ def create_routes(memory):
             "name": "general",
             "chain": GeneralChain()
         },
+        {
+            "name": "inspect",
+            "chain": InspectChain()
+        },
         {
             "name": "3dview",
             "chain": ViewChain.create(memory=memory)

src/workspaces/cookiefactoryv3/assistant/app/lib/tools/general.py

@@ -3,18 +3,19 @@
 
 from typing import Any, Dict, List, Optional
 
-from langchain import LLMChain, PromptTemplate
+from langchain_core.prompts import PromptTemplate
+from langchain import LLMChain
 from langchain.agents import tool
 from langchain.chains.base import Chain
 from langchain.callbacks.manager import (
     AsyncCallbackManagerForChainRun,
     CallbackManagerForChainRun,
 )
 
-from ..llm import get_bedrock_text, get_processed_prompt_template
+from ..llm import get_bedrock_text_v3_sonnet, get_processed_prompt_template_sonnet
 
-llm = get_bedrock_text()
-prompt_template = get_processed_prompt_template("{question}")
+llm = get_bedrock_text_v3_sonnet()
+prompt_template = get_processed_prompt_template_sonnet("{question}")
 
 def get_tool_metadata():
     return {