Create app.py

app.py

@@ -0,0 +1,123 @@
+from langchain_core.messages import SystemMessage
+from langchain_openai import ChatOpenAI
+from langchain_core.pydantic_v1 import BaseModel, Field
+from langgraph.graph import MessageGraph, END
+from langgraph.checkpoint.sqlite import SqliteSaver
+from langchain_core.messages import HumanMessage
+import streamlit as st
+from typing import List
+import os
+import uuid
+
+template = """Your job is to get information from a user about their profession. We are aiming to generate a profile later
+
+You should get the following information from them:
+
+- Their most recent job
+- Company
+- tools for example for a software engineer(which frameworks/languages)
+
+If you are not able to discerne this info, ask them to clarify! Do not attempt to wildly guess.
+If you're asking anything please be friendly and comment on any of the info you have found e.g working at x company must have been a thrilling challenge
+ 
+After you are able to discerne all the information, call the relevant tool"""
+
+llm = ChatOpenAI(temperature=0)
+
+def get_messages_info(messages):
+    return [SystemMessage(content=template)] + messages
+
+
+class PromptInstructions(BaseModel):
+    """Instructions on how to prompt the LLM."""
+    job: str
+    company: str
+    tools: List[str]
+    skills: List[str]
+
+llm_with_tool = llm.bind_tools([PromptInstructions])
+
+chain = get_messages_info | llm_with_tool
+
+# Helper function for determining if tool was called
+def _is_tool_call(msg):
+    return hasattr(msg, "additional_kwargs") and 'tool_calls' in msg.additional_kwargs
+
+
+# New system prompt
+prompt_system = """Based on the following context, write a good professional profile. Infer the soft skills:
+
+{reqs}"""
+
+# Function to get the messages for the profile
+# Will only get messages AFTER the tool call
+def get_profile_messages(messages):
+    tool_call = None
+    other_msgs = []
+    for m in messages:
+        if _is_tool_call(m):
+            tool_call = m.additional_kwargs['tool_calls'][0]['function']['arguments']
+        elif tool_call is not None:
+            other_msgs.append(m)
+    return [SystemMessage(content=prompt_system.format(reqs=tool_call))] + other_msgs
+
+profile_gen_chain = get_profile_messages | llm
+
+def get_state(messages):
+    if _is_tool_call(messages[-1]):
+        return "profile"
+    elif not isinstance(messages[-1], HumanMessage):
+        return END
+    for m in messages:
+        if _is_tool_call(m):
+            return "profile"
+    return "info"
+
+
+
+memory = SqliteSaver.from_conn_string(":memory:")
+
+nodes = {k:k for k in ['info', 'profile', END]}
+workflow = MessageGraph()
+workflow.add_node("info", chain)
+workflow.add_node("profile", profile_gen_chain)
+workflow.add_conditional_edges("info", get_state, nodes)
+workflow.add_conditional_edges("profile", get_state, nodes)
+workflow.set_entry_point("info")
+graph = workflow.compile(checkpointer=memory)
+
+config = {"configurable": {"thread_id": str(uuid.uuid4())}}
+
+# Helper function to execute the LangChain logic
+def execute_langchain(user_input):
+    output_list = []
+    for output in graph.stream([HumanMessage(content=user_input)], config=config):
+        if "__end__" in output:
+            continue
+        for key, value in output.items():
+            output_list.append((key, value))
+    return output_list
+
+# Streamlit app layout
+st.title("LangChain Chat")
+
+user_input = st.text_input("You:", "")
+
+if st.button("Send"):
+    if user_input:
+        # Execute LangChain logic
+        outputs = execute_langchain(user_input)
+
+        # Display LangChain outputs
+        for key, value in outputs:
+            if key == "Output from node 'info':":
+                st.text("Bot: " + value)
+            elif key == "Output from node 'profile':":
+                st.text("Bot: " + value)
+        
+        st.text("\n---\n")
+
+# Allow the user to quit the chat
+if st.button("Quit"):
+    st.text("Bot: Byebye")
+