Spaces:
Paused
Paused
angry-meow
commited on
Commit
·
06dc1b7
1
Parent(s):
954011f
updated graph stucture
Browse files
graph.py
CHANGED
@@ -1,4 +1,4 @@
|
|
1 |
-
from typing import Dict, TypedDict, Annotated, Sequence
|
2 |
from langgraph.graph import Graph, StateGraph, END
|
3 |
from langgraph.prebuilt import ToolExecutor
|
4 |
from langchain.schema import StrOutputParser
|
@@ -12,11 +12,14 @@ from operator import itemgetter
|
|
12 |
# Define the state structure
|
13 |
class State(TypedDict):
|
14 |
messages: Sequence[str]
|
|
|
15 |
research_data: Dict[str, str]
|
16 |
-
|
|
|
17 |
final_post: str
|
18 |
|
19 |
|
|
|
20 |
# Research Agent Pieces
|
21 |
qdrant_research_chain = (
|
22 |
{"context": itemgetter("topic") | models.compression_retriever, "topic": itemgetter("topic")}
|
@@ -34,10 +37,10 @@ tavily_chain = (
|
|
34 |
|
35 |
def query_qdrant(state: State) -> State:
|
36 |
# Extract the last message as the input
|
37 |
-
|
38 |
|
39 |
# Run the chain
|
40 |
-
result = qdrant_research_chain.invoke({"topic":
|
41 |
|
42 |
# Update the state with the research results
|
43 |
state["research_data"]["qdrant_results"] = result
|
@@ -46,7 +49,7 @@ def query_qdrant(state: State) -> State:
|
|
46 |
|
47 |
def web_search(state: State) -> State:
|
48 |
# Extract the last message as the topic
|
49 |
-
topic = state["
|
50 |
|
51 |
# Get the Qdrant results from the state
|
52 |
qdrant_results = state["research_data"].get("qdrant_results", "No previous results available.")
|
@@ -99,11 +102,15 @@ research_graph.add_node("research_supervisor", research_supervisor)
|
|
99 |
|
100 |
research_graph.add_edge("query_qdrant", "research_supervisor")
|
101 |
research_graph.add_edge("web_search", "research_supervisor")
|
102 |
-
research_graph.
|
103 |
-
|
104 |
-
|
|
|
|
|
|
|
105 |
|
106 |
research_graph.set_entry_point("research_supervisor")
|
|
|
107 |
|
108 |
# Create the writing team graph
|
109 |
writing_graph = StateGraph(State)
|
@@ -114,15 +121,25 @@ writing_graph.add_node("voice_editing", voice_editing)
|
|
114 |
writing_graph.add_node("post_review", post_review)
|
115 |
writing_graph.add_node("writing_supervisor", writing_supervisor)
|
116 |
|
117 |
-
writing_graph.add_edge("
|
118 |
-
writing_graph.add_edge("
|
119 |
-
writing_graph.add_edge("
|
120 |
-
writing_graph.add_edge("voice_editing", "post_review")
|
121 |
writing_graph.add_edge("post_review", "writing_supervisor")
|
122 |
-
writing_graph.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
123 |
|
124 |
writing_graph.set_entry_point("writing_supervisor")
|
125 |
|
|
|
|
|
126 |
# Create the overall graph
|
127 |
overall_graph = StateGraph(State)
|
128 |
|
@@ -136,11 +153,15 @@ overall_graph.add_node("overall_supervisor", overall_supervisor)
|
|
136 |
overall_graph.set_entry_point("overall_supervisor")
|
137 |
|
138 |
# Connect the nodes
|
139 |
-
overall_graph.add_edge("overall_supervisor", "research_team")
|
140 |
overall_graph.add_edge("research_team", "overall_supervisor")
|
141 |
-
overall_graph.add_edge("overall_supervisor", "writing_team")
|
142 |
overall_graph.add_edge("writing_team", "overall_supervisor")
|
143 |
-
overall_graph.
|
|
|
|
|
|
|
|
|
|
|
|
|
144 |
|
145 |
# Compile the graph
|
146 |
app = overall_graph.compile()
|
|
|
1 |
+
from typing import Dict, List, TypedDict, Annotated, Sequence
|
2 |
from langgraph.graph import Graph, StateGraph, END
|
3 |
from langgraph.prebuilt import ToolExecutor
|
4 |
from langchain.schema import StrOutputParser
|
|
|
12 |
# Define the state structure
|
13 |
class State(TypedDict):
|
14 |
messages: Sequence[str]
|
15 |
+
topic: str
|
16 |
research_data: Dict[str, str]
|
17 |
+
team_members: List[str]
|
18 |
+
draft_posts: Sequence[str]
|
19 |
final_post: str
|
20 |
|
21 |
|
22 |
+
research_members = ["Qdrant_researcher", "Web_researcher"]
|
23 |
# Research Agent Pieces
|
24 |
qdrant_research_chain = (
|
25 |
{"context": itemgetter("topic") | models.compression_retriever, "topic": itemgetter("topic")}
|
|
|
37 |
|
38 |
def query_qdrant(state: State) -> State:
|
39 |
# Extract the last message as the input
|
40 |
+
topic = state["topic"]
|
41 |
|
42 |
# Run the chain
|
43 |
+
result = qdrant_research_chain.invoke({"topic": topic})
|
44 |
|
45 |
# Update the state with the research results
|
46 |
state["research_data"]["qdrant_results"] = result
|
|
|
49 |
|
50 |
def web_search(state: State) -> State:
|
51 |
# Extract the last message as the topic
|
52 |
+
topic = state["topic"]
|
53 |
|
54 |
# Get the Qdrant results from the state
|
55 |
qdrant_results = state["research_data"].get("qdrant_results", "No previous results available.")
|
|
|
102 |
|
103 |
research_graph.add_edge("query_qdrant", "research_supervisor")
|
104 |
research_graph.add_edge("web_search", "research_supervisor")
|
105 |
+
research_graph.add_conditional_edges(
|
106 |
+
"research_supervisor",
|
107 |
+
lambda x: x["next"],
|
108 |
+
{"query_qdrant": "query_qdrant", "web_search": "web_search", "FINISH": END},
|
109 |
+
)
|
110 |
+
#research_graph.add_edge("research_supervisor", END)
|
111 |
|
112 |
research_graph.set_entry_point("research_supervisor")
|
113 |
+
research_graph_comp = research_graph.compile()
|
114 |
|
115 |
# Create the writing team graph
|
116 |
writing_graph = StateGraph(State)
|
|
|
121 |
writing_graph.add_node("post_review", post_review)
|
122 |
writing_graph.add_node("writing_supervisor", writing_supervisor)
|
123 |
|
124 |
+
writing_graph.add_edge("post_creation", "writing_supervisor")
|
125 |
+
writing_graph.add_edge("copy_editing", "writing_supervisor")
|
126 |
+
writing_graph.add_edge("voice_editing", "writing_supervisor")
|
|
|
127 |
writing_graph.add_edge("post_review", "writing_supervisor")
|
128 |
+
writing_graph.add_conditional_edges(
|
129 |
+
"writing_supervisor",
|
130 |
+
lambda x: x["next"],
|
131 |
+
{"post_creation": "post_creation",
|
132 |
+
"copy_editing": "copy_editing",
|
133 |
+
"voice_editing": "voice_editing",
|
134 |
+
"post_review": "post_review",
|
135 |
+
"FINISH": END},
|
136 |
+
)
|
137 |
+
#writing_graph.add_edge("writing_supervisor", END)
|
138 |
|
139 |
writing_graph.set_entry_point("writing_supervisor")
|
140 |
|
141 |
+
writing_graph_comp = research_graph.compile()
|
142 |
+
|
143 |
# Create the overall graph
|
144 |
overall_graph = StateGraph(State)
|
145 |
|
|
|
153 |
overall_graph.set_entry_point("overall_supervisor")
|
154 |
|
155 |
# Connect the nodes
|
|
|
156 |
overall_graph.add_edge("research_team", "overall_supervisor")
|
|
|
157 |
overall_graph.add_edge("writing_team", "overall_supervisor")
|
158 |
+
overall_graph.add_conditional_edges(
|
159 |
+
"overall_supervisor",
|
160 |
+
lambda x: x["next"],
|
161 |
+
{"research_team": "research_team",
|
162 |
+
"writing_team": "writing_team",
|
163 |
+
"FINISH": END},
|
164 |
+
)
|
165 |
|
166 |
# Compile the graph
|
167 |
app = overall_graph.compile()
|