Create test.py
Browse files
test.py
ADDED
@@ -0,0 +1,133 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
from typing import TypedDict, Annotated
|
3 |
+
from langgraph.graph import StateGraph
|
4 |
+
from langgraph.checkpoint.memory import MemorySaver
|
5 |
+
from langgraph.graph.message import add_messages
|
6 |
+
from langchain_openai import ChatOpenAI
|
7 |
+
from langchain_community.tools.tavily_search import TavilySearchResults
|
8 |
+
from langchain_core.messages import HumanMessage, ToolMessage, AIMessage
|
9 |
+
from langgraph.prebuilt import tools_condition
|
10 |
+
import os
|
11 |
+
|
12 |
+
# Streamlit UI Header
|
13 |
+
st.title("Checkpoints and Breakpoints")
|
14 |
+
st.caption("Demonstrating LangGraph workflow execution with interruptions and tool invocation.")
|
15 |
+
|
16 |
+
# Fetch API Keys
|
17 |
+
openai_api_key = os.getenv("OPENAI_API_KEY")
|
18 |
+
tavily_api_key = os.getenv("TAVILY_API_KEY")
|
19 |
+
|
20 |
+
if not openai_api_key or not tavily_api_key:
|
21 |
+
st.error("API keys are missing! Set OPENAI_API_KEY and TAVILY_API_KEY in Hugging Face Spaces Secrets.")
|
22 |
+
st.stop()
|
23 |
+
|
24 |
+
os.environ["OPENAI_API_KEY"] = openai_api_key
|
25 |
+
os.environ["TAVILY_API_KEY"] = tavily_api_key
|
26 |
+
|
27 |
+
# Define State Class
|
28 |
+
class State(TypedDict):
|
29 |
+
messages: Annotated[list, add_messages]
|
30 |
+
|
31 |
+
# Initialize LLM and Tools
|
32 |
+
llm = ChatOpenAI(model="gpt-4o-mini")
|
33 |
+
tool = TavilySearchResults(max_results=2)
|
34 |
+
llm_with_tools = llm.bind_tools([tool])
|
35 |
+
|
36 |
+
# Agent Node
|
37 |
+
def Agent(state: State):
|
38 |
+
st.sidebar.write("Agent Input State:", state["messages"])
|
39 |
+
response = llm_with_tools.invoke(state["messages"])
|
40 |
+
st.sidebar.write("Agent Response:", response)
|
41 |
+
return {"messages": [response]}
|
42 |
+
|
43 |
+
# Tools Execution Node
|
44 |
+
def ExecuteTools(state: State):
|
45 |
+
tool_calls = state["messages"][-1].tool_calls
|
46 |
+
responses = []
|
47 |
+
|
48 |
+
if tool_calls:
|
49 |
+
for call in tool_calls:
|
50 |
+
tool_name = call["name"]
|
51 |
+
args = call["args"]
|
52 |
+
st.sidebar.write("Tool Call Detected:", tool_name, args)
|
53 |
+
|
54 |
+
if tool_name == "tavily_search_results_json":
|
55 |
+
tool_response = tool.invoke({"query": args["query"]})
|
56 |
+
st.sidebar.write("Tool Response:", tool_response)
|
57 |
+
responses.append(ToolMessage(content=str(tool_response), tool_call_id=call["id"]))
|
58 |
+
return {"messages": responses}
|
59 |
+
|
60 |
+
# Memory Checkpoint
|
61 |
+
memory = MemorySaver()
|
62 |
+
|
63 |
+
# Build the Graph
|
64 |
+
graph = StateGraph(State)
|
65 |
+
graph.add_node("Agent", Agent)
|
66 |
+
graph.add_node("ExecuteTools", ExecuteTools)
|
67 |
+
|
68 |
+
# Add Conditional Edge to Check for Tools
|
69 |
+
def custom_tools_condition(state: State):
|
70 |
+
return "True" if state["messages"][-1].tool_calls else "False"
|
71 |
+
|
72 |
+
graph.add_conditional_edges("Agent", custom_tools_condition, {"True": "ExecuteTools", "False": "Agent"})
|
73 |
+
graph.add_edge("ExecuteTools", "Agent")
|
74 |
+
graph.set_entry_point("Agent")
|
75 |
+
|
76 |
+
# Compile the Graph
|
77 |
+
app = graph.compile(checkpointer=memory, interrupt_before=["ExecuteTools"])
|
78 |
+
|
79 |
+
# Display Graph Visualization
|
80 |
+
st.subheader("Graph Visualization")
|
81 |
+
st.image(app.get_graph().draw_mermaid_png(), caption="Workflow Graph", use_container_width=True)
|
82 |
+
|
83 |
+
# Run the Workflow
|
84 |
+
st.subheader("Run the Workflow")
|
85 |
+
user_input = st.text_input("Enter a message to start the graph:", "Search for the weather in Uttar Pradesh")
|
86 |
+
thread_id = st.text_input("Thread ID", "1")
|
87 |
+
|
88 |
+
if st.button("Execute Workflow"):
|
89 |
+
thread = {"configurable": {"thread_id": thread_id}}
|
90 |
+
input_message = {"messages": [HumanMessage(content=user_input)]}
|
91 |
+
|
92 |
+
st.write("### Execution Outputs")
|
93 |
+
outputs = []
|
94 |
+
|
95 |
+
try:
|
96 |
+
# Stream the graph execution
|
97 |
+
for event in app.stream(input_message, thread, stream_mode="values"):
|
98 |
+
output_message = event["messages"][-1]
|
99 |
+
st.code(output_message.content)
|
100 |
+
outputs.append(output_message.content)
|
101 |
+
st.sidebar.write("Intermediate State:", event["messages"])
|
102 |
+
|
103 |
+
# Display Intermediate Outputs
|
104 |
+
if outputs:
|
105 |
+
st.subheader("Intermediate Outputs")
|
106 |
+
for idx, output in enumerate(outputs, start=1):
|
107 |
+
st.write(f"**Step {idx}:**")
|
108 |
+
st.code(output)
|
109 |
+
else:
|
110 |
+
st.warning("No outputs generated. Check the workflow or tool calls.")
|
111 |
+
|
112 |
+
# Snapshot of Current State
|
113 |
+
st.subheader("Current State Snapshot")
|
114 |
+
snapshot = app.get_state(thread)
|
115 |
+
current_message = snapshot.values["messages"][-1]
|
116 |
+
st.code(current_message.pretty_print())
|
117 |
+
|
118 |
+
# Manual Update for Interrupted State
|
119 |
+
if hasattr(current_message, "tool_calls") and current_message.tool_calls:
|
120 |
+
tool_call_id = current_message.tool_calls[0]["id"]
|
121 |
+
manual_response = st.text_area("Manual Tool Response", "Enter response to continue execution...")
|
122 |
+
if st.button("Update State"):
|
123 |
+
new_messages = [
|
124 |
+
ToolMessage(content=manual_response, tool_call_id=tool_call_id),
|
125 |
+
AIMessage(content=manual_response),
|
126 |
+
]
|
127 |
+
app.update_state(thread, {"messages": new_messages})
|
128 |
+
st.success("State updated successfully!")
|
129 |
+
st.code(app.get_state(thread).values["messages"][-1].pretty_print())
|
130 |
+
else:
|
131 |
+
st.warning("No tool calls detected to update the state.")
|
132 |
+
except Exception as e:
|
133 |
+
st.error(f"Error during execution: {e}")
|