Spaces:
Sleeping
Sleeping
Update services/graph_service.py
Browse files- services/graph_service.py +27 -19
services/graph_service.py
CHANGED
|
@@ -7,7 +7,10 @@ from langchain_openai import ChatOpenAI
|
|
| 7 |
from langchain_core.messages import BaseMessage, AIMessage, HumanMessage, SystemMessage
|
| 8 |
from langchain.agents import create_openai_tools_agent, AgentExecutor
|
| 9 |
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
|
|
|
|
| 10 |
from langgraph.graph import StateGraph, END
|
|
|
|
|
|
|
| 11 |
from tools.analysis_tools import trigger_interview_analysis
|
| 12 |
|
| 13 |
class AgentState(TypedDict):
|
|
@@ -18,11 +21,11 @@ class AgentState(TypedDict):
|
|
| 18 |
class GraphInterviewProcessor:
|
| 19 |
"""
|
| 20 |
Cette classe encapsule la logique d'un entretien en utilisant LangGraph.
|
| 21 |
-
Elle prépare toutes les données nécessaires à l'initialisation
|
| 22 |
"""
|
| 23 |
def __init__(self, payload: Dict[str, Any]):
|
| 24 |
logging.info("Initialisation de GraphInterviewProcessor...")
|
| 25 |
-
|
| 26 |
self.user_id = payload["user_id"]
|
| 27 |
self.job_offer_id = payload["job_offer_id"]
|
| 28 |
self.job_offer = payload["job_offer"]
|
|
@@ -41,7 +44,6 @@ class GraphInterviewProcessor:
|
|
| 41 |
logging.info("GraphInterviewProcessor initialisé avec succès.")
|
| 42 |
|
| 43 |
def _load_prompt_template(self, file_path: str) -> str:
|
| 44 |
-
"""Charge le template du prompt depuis un fichier."""
|
| 45 |
try:
|
| 46 |
with open(file_path, 'r', encoding='utf-8') as f:
|
| 47 |
return f.read()
|
|
@@ -50,26 +52,21 @@ class GraphInterviewProcessor:
|
|
| 50 |
return "Vous êtes un assistant RH."
|
| 51 |
|
| 52 |
def _format_cv_for_prompt(self) -> str:
|
| 53 |
-
"""Formate le CV pour l'injecter dans le prompt."""
|
| 54 |
return json.dumps(self.cv_data, indent=2, ensure_ascii=False)
|
| 55 |
|
| 56 |
def _extract_skills_summary(self) -> str:
|
| 57 |
-
"""Extrait un résumé des compétences avec niveaux."""
|
| 58 |
competences = self.cv_data.get('analyse_competences', [])
|
| 59 |
-
if not competences:
|
| 60 |
-
return "Aucune analyse de compétences disponible."
|
| 61 |
summary = [f"{comp.get('skill', '')}: {comp.get('level', 'débutant')}" for comp in competences]
|
| 62 |
return "Niveaux de compétences du candidat: " + " | ".join(summary)
|
| 63 |
|
| 64 |
def _extract_reconversion_info(self) -> str:
|
| 65 |
-
"""Extrait les informations de reconversion."""
|
| 66 |
reconversion = self.cv_data.get('reconversion', {})
|
| 67 |
if reconversion.get('is_reconversion'):
|
| 68 |
return f"CANDIDAT EN RECONVERSION: {reconversion.get('analysis', '')}"
|
| 69 |
return "Le candidat n'est pas identifié comme étant en reconversion."
|
| 70 |
|
| 71 |
def _create_agent_executor(self) -> AgentExecutor:
|
| 72 |
-
"""Crée l'agent executor avec un prompt minimaliste."""
|
| 73 |
prompt = ChatPromptTemplate.from_messages([
|
| 74 |
SystemMessage(content="{system_prompt_content}"),
|
| 75 |
MessagesPlaceholder(variable_name="messages"),
|
|
@@ -81,7 +78,6 @@ class GraphInterviewProcessor:
|
|
| 81 |
return AgentExecutor(agent=agent, tools=tools, verbose=True)
|
| 82 |
|
| 83 |
def _agent_node(self, state: AgentState):
|
| 84 |
-
"""Prépare le prompt système dynamiquement et appelle l'agent."""
|
| 85 |
system_prompt_content = self.system_prompt_template.format(
|
| 86 |
entreprise=self.job_offer.get('entreprise', 'notre entreprise'),
|
| 87 |
poste=self.job_offer.get('poste', 'ce poste'),
|
|
@@ -102,24 +98,33 @@ class GraphInterviewProcessor:
|
|
| 102 |
"conversation_history": state["messages"]
|
| 103 |
})
|
| 104 |
return {"messages": [response['output']]}
|
| 105 |
-
|
| 106 |
def _router(self, state: AgentState):
|
| 107 |
"""Décide du chemin à suivre après la réponse de l'agent."""
|
| 108 |
last_message = state["messages"][-1]
|
| 109 |
-
if hasattr(last_message, 'tool_calls') and last_message.tool_calls:
|
| 110 |
-
return "
|
| 111 |
-
|
| 112 |
-
|
| 113 |
def _build_graph(self) -> any:
|
| 114 |
"""Construit et compile le graphe d'états."""
|
|
|
|
|
|
|
| 115 |
graph = StateGraph(AgentState)
|
| 116 |
graph.add_node("agent", self._agent_node)
|
|
|
|
|
|
|
|
|
|
| 117 |
graph.add_conditional_edges(
|
| 118 |
"agent",
|
| 119 |
self._router,
|
| 120 |
-
{
|
|
|
|
|
|
|
|
|
|
| 121 |
)
|
| 122 |
-
|
|
|
|
|
|
|
| 123 |
return graph.compile()
|
| 124 |
|
| 125 |
def invoke(self, messages: List[Dict[str, Any]]):
|
|
@@ -135,9 +140,12 @@ class GraphInterviewProcessor:
|
|
| 135 |
final_state = self.graph.invoke(initial_state)
|
| 136 |
|
| 137 |
last_message = final_state['messages'][-1]
|
| 138 |
-
|
|
|
|
|
|
|
|
|
|
| 139 |
|
| 140 |
return {
|
| 141 |
"response": last_message.content,
|
| 142 |
"status": status
|
| 143 |
-
}
|
|
|
|
| 7 |
from langchain_core.messages import BaseMessage, AIMessage, HumanMessage, SystemMessage
|
| 8 |
from langchain.agents import create_openai_tools_agent, AgentExecutor
|
| 9 |
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
|
| 10 |
+
# Nouveaux imports pour la gestion correcte du graphe
|
| 11 |
from langgraph.graph import StateGraph, END
|
| 12 |
+
from langgraph.prebuilt import ToolNode
|
| 13 |
+
|
| 14 |
from tools.analysis_tools import trigger_interview_analysis
|
| 15 |
|
| 16 |
class AgentState(TypedDict):
|
|
|
|
| 21 |
class GraphInterviewProcessor:
|
| 22 |
"""
|
| 23 |
Cette classe encapsule la logique d'un entretien en utilisant LangGraph.
|
| 24 |
+
Elle prépare toutes les données nécessaires à l'initialisation.
|
| 25 |
"""
|
| 26 |
def __init__(self, payload: Dict[str, Any]):
|
| 27 |
logging.info("Initialisation de GraphInterviewProcessor...")
|
| 28 |
+
|
| 29 |
self.user_id = payload["user_id"]
|
| 30 |
self.job_offer_id = payload["job_offer_id"]
|
| 31 |
self.job_offer = payload["job_offer"]
|
|
|
|
| 44 |
logging.info("GraphInterviewProcessor initialisé avec succès.")
|
| 45 |
|
| 46 |
def _load_prompt_template(self, file_path: str) -> str:
|
|
|
|
| 47 |
try:
|
| 48 |
with open(file_path, 'r', encoding='utf-8') as f:
|
| 49 |
return f.read()
|
|
|
|
| 52 |
return "Vous êtes un assistant RH."
|
| 53 |
|
| 54 |
def _format_cv_for_prompt(self) -> str:
|
|
|
|
| 55 |
return json.dumps(self.cv_data, indent=2, ensure_ascii=False)
|
| 56 |
|
| 57 |
def _extract_skills_summary(self) -> str:
|
|
|
|
| 58 |
competences = self.cv_data.get('analyse_competences', [])
|
| 59 |
+
if not competences: return "Aucune analyse de compétences disponible."
|
|
|
|
| 60 |
summary = [f"{comp.get('skill', '')}: {comp.get('level', 'débutant')}" for comp in competences]
|
| 61 |
return "Niveaux de compétences du candidat: " + " | ".join(summary)
|
| 62 |
|
| 63 |
def _extract_reconversion_info(self) -> str:
|
|
|
|
| 64 |
reconversion = self.cv_data.get('reconversion', {})
|
| 65 |
if reconversion.get('is_reconversion'):
|
| 66 |
return f"CANDIDAT EN RECONVERSION: {reconversion.get('analysis', '')}"
|
| 67 |
return "Le candidat n'est pas identifié comme étant en reconversion."
|
| 68 |
|
| 69 |
def _create_agent_executor(self) -> AgentExecutor:
|
|
|
|
| 70 |
prompt = ChatPromptTemplate.from_messages([
|
| 71 |
SystemMessage(content="{system_prompt_content}"),
|
| 72 |
MessagesPlaceholder(variable_name="messages"),
|
|
|
|
| 78 |
return AgentExecutor(agent=agent, tools=tools, verbose=True)
|
| 79 |
|
| 80 |
def _agent_node(self, state: AgentState):
|
|
|
|
| 81 |
system_prompt_content = self.system_prompt_template.format(
|
| 82 |
entreprise=self.job_offer.get('entreprise', 'notre entreprise'),
|
| 83 |
poste=self.job_offer.get('poste', 'ce poste'),
|
|
|
|
| 98 |
"conversation_history": state["messages"]
|
| 99 |
})
|
| 100 |
return {"messages": [response['output']]}
|
|
|
|
| 101 |
def _router(self, state: AgentState):
|
| 102 |
"""Décide du chemin à suivre après la réponse de l'agent."""
|
| 103 |
last_message = state["messages"][-1]
|
| 104 |
+
if hasattr(last_message, 'tool_calls') and last_message.tool_calls and len(last_message.tool_calls) > 0:
|
| 105 |
+
return "call_tool"
|
| 106 |
+
else:
|
| 107 |
+
return "end_turn"
|
| 108 |
def _build_graph(self) -> any:
|
| 109 |
"""Construit et compile le graphe d'états."""
|
| 110 |
+
tool_node = ToolNode([trigger_interview_analysis])
|
| 111 |
+
|
| 112 |
graph = StateGraph(AgentState)
|
| 113 |
graph.add_node("agent", self._agent_node)
|
| 114 |
+
graph.add_node("tools", tool_node)
|
| 115 |
+
|
| 116 |
+
graph.set_entry_point("agent")
|
| 117 |
graph.add_conditional_edges(
|
| 118 |
"agent",
|
| 119 |
self._router,
|
| 120 |
+
{
|
| 121 |
+
"call_tool": "tools",
|
| 122 |
+
"end_turn": END
|
| 123 |
+
}
|
| 124 |
)
|
| 125 |
+
|
| 126 |
+
graph.add_edge("tools", "agent")
|
| 127 |
+
|
| 128 |
return graph.compile()
|
| 129 |
|
| 130 |
def invoke(self, messages: List[Dict[str, Any]]):
|
|
|
|
| 140 |
final_state = self.graph.invoke(initial_state)
|
| 141 |
|
| 142 |
last_message = final_state['messages'][-1]
|
| 143 |
+
|
| 144 |
+
# Le statut est 'finished' si le dernier message contient un appel à un outil.
|
| 145 |
+
# Cela signifie que le graphe a pris le chemin de l'analyse.
|
| 146 |
+
status = "finished" if hasattr(last_message, 'tool_calls') and last_message.tool_calls else "interviewing"
|
| 147 |
|
| 148 |
return {
|
| 149 |
"response": last_message.content,
|
| 150 |
"status": status
|
| 151 |
+
}
|