File size: 19,453 Bytes
2601697
 
 
 
 
 
 
 
 
 
 
d8faa94
2601697
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d8faa94
2601697
d8faa94
2601697
 
 
 
 
 
 
d8faa94
2601697
 
 
 
 
 
 
d8faa94
2601697
 
 
 
 
 
d8faa94
2601697
 
d8faa94
2601697
d8faa94
 
2601697
d8faa94
 
 
 
 
 
 
 
 
 
 
2601697
d8faa94
 
2601697
 
d8faa94
2601697
 
 
 
d8faa94
2601697
 
d8faa94
 
 
 
 
 
 
 
 
 
 
 
 
 
2601697
 
 
 
d8faa94
 
 
 
2601697
 
 
 
d8faa94
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2601697
 
 
 
d8faa94
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2601697
d8faa94
2601697
 
 
 
d8faa94
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2601697
 
 
d8faa94
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2601697
d8faa94
2601697
d8faa94
 
 
 
2601697
 
d8faa94
 
 
 
2601697
 
d8faa94
2601697
 
d8faa94
2601697
 
d8faa94
 
 
 
 
 
 
 
 
 
 
 
 
2601697
 
 
d8faa94
 
2601697
 
 
 
d8faa94
2601697
 
d8faa94
 
 
 
 
 
 
2601697
 
d8faa94
 
 
 
2601697
 
 
 
d8faa94
 
 
 
 
2601697
 
d8faa94
 
 
 
 
2601697
 
d8faa94
 
 
2601697
 
 
d8faa94
2601697
 
 
 
d8faa94
 
 
 
 
 
2601697
d8faa94
 
 
 
 
 
2601697
 
 
 
d8faa94
 
 
 
2601697
 
 
d253655
d8faa94
 
 
 
2601697
 
 
 
d8faa94
2601697
 
 
 
d8faa94
2601697
 
d8faa94
2601697
 
 
 
d8faa94
 
 
 
 
 
 
2601697
 
 
 
d8faa94
 
2601697
 
d8faa94
 
2601697
 
 
d8faa94
 
 
2601697
 
d8faa94
 
 
 
 
2601697
 
 
 
d8faa94
 
2601697
 
 
d8faa94
 
2601697
d8faa94
2601697
d8faa94
 
2601697
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393

# ==============================================================================
# 1. IMPORTS
# All necessary libraries for the application.
# ==============================================================================
import os
import chromadb
from dotenv import load_dotenv
from typing import Dict, List, Any, TypedDict
from datetime import datetime
import streamlit as st
import httpx # ADDED THIS IMPORT

from langchain_core.runnables import RunnablePassthrough
from langchain.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser
from langchain_core.tools import tool
from langchain.agents import create_tool_calling_agent, AgentExecutor
from langchain_openai import OpenAIEmbeddings, ChatOpenAI
from langchain_community.vectorstores import Chroma

from langgraph.graph import StateGraph, END, START
from pydantic import BaseModel

from groq import Groq
from mem0 import MemoryClient

# ==============================================================================
# 2. SETUP & CONFIGURATION
# Load secrets and initialize core models (LLM, Embeddings).
# This section uses environment variables, which is correct for deployment.
# ==============================================================================
# On Hugging Face Spaces, these will be set as secrets.
load_dotenv()

openai_api_key = os.environ.get("OPENAI_API_KEY")
openai_api_base = os.environ.get("OPENAI_API_BASE")
groq_api_key = os.environ.get('GROQ_API_KEY')
mem0_api_key = os.environ.get('MEM0_API_KEY')

# Initialize the Chat OpenAI model
llm = ChatOpenAI(
    openai_api_base=openai_api_base,
    openai_api_key=openai_api_key,
    model="gpt-4o-mini",
    streaming=False
)

# Initialize the OpenAI Embeddings model
embedding_model = OpenAIEmbeddings(
    openai_api_base=openai_api_base,
    openai_api_key=openai_api_key,
    model='text-embedding-ada-002'
)


# ==============================================================================
# 3. ADVANCED RAG AGENT WORKFLOW
# This is the complete, self-contained logic for the LangGraph agent.
# ==============================================================================

# 3.1. Define Agent State
class AgentState(TypedDict):
    query: str
    expanded_query: str
    context: List[Dict[str, Any]]
    response: Any
    precision_score: float
    groundedness_score: float
    groundedness_loop_count: int
    precision_loop_count: int
    feedback: str
    query_feedback: str
    loop_max_iter: int

# 3.2. Load the Vector Store
# This points to the pre-built database that will be in the Docker container.
vector_store = Chroma(
    collection_name='nutritional_hypotheticals',
    persist_directory="./nutritional_db", # The path inside the Docker container
    embedding_function=embedding_model
)
retriever = vector_store.as_retriever(search_type='similarity', search_kwargs={'k': 5})

# 3.3. Define All Workflow Nodes (Functions)
def expand_query(state):
    print("---------Expanding Query---------")
    system_message = '''You are an expert at query expansion. Your goal is to rewrite the user's query to be more specific and comprehensive, making it ideal for a vector database search focused on nutritional disorders.
When expanding the query, consider the following:
- **Clarify Ambiguities**: Resolve any vague terms or phrases.
- **Add Synonyms and Related Terms**: Include alternative names for disorders, symptoms, or treatments.
- **Specify Context**: Frame the query within the context of nutritional health, deficiencies, symptoms, causes, and treatments.
- **Use Feedback**: Incorporate suggestions from previous refinement steps to improve the query.
Provide only the expanded query as a single, continuous string.'''
    expand_prompt = ChatPromptTemplate.from_messages([
        ("system", system_message),
        ("user", "Expand this query: {query} using the feedback: {query_feedback}")
    ])
    chain = expand_prompt | llm | StrOutputParser()
    expanded_query = chain.invoke({"query": state['query'], "query_feedback":state["query_feedback"]})
    state["expanded_query"] = expanded_query
    return state

def retrieve_context(state):
    print("---------retrieve_context---------")
    query = state['expanded_query']
    docs = retriever.invoke(query)
    context = [{"content": doc.page_content, "metadata": doc.metadata} for doc in docs]
    state['context'] = context
    return state

def craft_response(state: Dict) -> Dict:
    print("---------craft_response---------")
    system_message = '''You are a knowledgeable and precise AI assistant specializing in nutritional disorders. Your task is to provide a clear and accurate answer to the user's query based *strictly* on the provided context.
Follow these guidelines:
1.  **Ground Your Answer**: Base your entire response on the information found in the context. Do not use any external knowledge.
2.  **Be Direct**: Address the user's query directly and concisely.
3.  **Acknowledge Limitations**: If the context does not contain the information needed to answer the query, clearly state that the information is not available in the provided documents.
4.  **Incorporate Feedback**: Use the provided feedback to refine your response and address any previous shortcomings.'''
    response_prompt = ChatPromptTemplate.from_messages([
        ("system", system_message),
        ("user", "Query: {query}\nContext: {context}\n\nfeedback: {feedback}")
    ])
    chain = response_prompt | llm
    response = chain.invoke({
        "query": state['query'],
        "context": "\n".join([doc["metadata"].get("original_content", "") for doc in state['context']]),
        "feedback": state['feedback']
    })
    state['response'] = response
    return state

def score_groundedness(state: Dict) -> Dict:
    print("---------check_groundedness---------")
    system_message = '''You are a groundedness scoring expert. Your role is to evaluate whether an AI-generated response is factually supported by the given context.
- **Score**: Provide a numerical score from 0.0 to 1.0.
  - **1.0**: The response is fully and accurately supported by the context.
  - **0.0**: The response is not supported by the context or contains fabricated information.
- **Crucial Rule**: If the provided context is empty or does not contain the information needed to answer the query, but the response still provides a factual answer, the score must be 0.0.
- **Output**: Return only the numerical score. Do not add any explanation or extra text.'''
    groundedness_prompt = ChatPromptTemplate.from_messages([
        ("system", system_message),
        ("user", "Context: {context}\nResponse: {response}\n\nGroundedness score:")
    ])
    chain = groundedness_prompt | llm | StrOutputParser()
    groundedness_score = float(chain.invoke({
        "context": "\n".join([doc["metadata"].get("original_content", "") for doc in state['context']]),
        "response": state['response'].content
    }))
    state['groundedness_loop_count'] += 1
    state['groundedness_score'] = groundedness_score
    return state

def check_precision(state: Dict) -> Dict:
    print("---------check_precision---------")
    system_message = '''You are a precision scoring expert. Your role is to evaluate how well an AI-generated response addresses a specific user query.
- **Score**: Provide a numerical score from 0.0 to 1.0.
  - **1.0**: The response is perfectly precise, comprehensive, and directly answers the user's query.
  - **0.0**: The response is completely irrelevant or fails to answer the query.
- **Output**: Return only the numerical score. Do not add any explanation or extra text.'''
    precision_prompt = ChatPromptTemplate.from_messages([
        ("system", system_message),
        ("user", "Query: {query}\nResponse: {response}\n\nPrecision score:")
    ])
    chain = precision_prompt | llm | StrOutputParser()
    precision_score = float(chain.invoke({
        "query": state['query'],
        "response": state['response'].content
    }))
    state['precision_score'] = precision_score
    state['precision_loop_count'] += 1
    return state

def refine_response(state: Dict) -> Dict:
    print("---------refine_response---------")
    system_message = '''You are a response refinement expert. Your task is to provide constructive feedback on an AI-generated response based on a user's query.
Analyze the response for:
- **Gaps**: Is any crucial information from the query missing?
- **Ambiguities**: Are there any unclear or vague statements?
- **Inaccuracies**: Does the response contradict the user's intent (even if it's based on the context)?
- **Completeness**: Could the response be more thorough while remaining concise?
**Do not rewrite the response.** Instead, provide specific, actionable suggestions for improvement.'''
    refine_response_prompt = ChatPromptTemplate.from_messages([
        ("system", system_message),
        ("user", "Query: {query}\\nResponse: {response}\\n\\n"
                 "What improvements can be made to enhance accuracy and completeness?")
    ])
    chain = refine_response_prompt | llm| StrOutputParser()
    feedback = f"Previous Response: {state['response'].content}\\nSuggestions: {chain.invoke({'query': state['query'], 'response': state['response'].content})}"
    state['feedback'] = feedback
    return state

def refine_query(state: Dict) -> Dict:
    print("---------refine_query---------")
    system_message = '''You are a query refinement expert. Your task is to analyze an original user query and its expanded version to suggest improvements for a more effective vector database search.
Review the expanded query for:
- **Missing Keywords**: Are there essential terms or synonyms that should be added?
- **Lack of Specificity**: Could the query be narrowed down to a more precise topic?
- **Scope Refinements**: Is the query too broad or too narrow?
**Do not rewrite the query.** Instead, provide structured, actionable suggestions for improvement based on the original query's intent.'''
    refine_query_prompt = ChatPromptTemplate.from_messages([
        ("system", system_message),
        ("user", "Original Query: {query}\\nExpanded Query: {expanded_query}\\n\\n"
                 "What improvements can be made for a better search?")
    ])
    chain = refine_query_prompt | llm | StrOutputParser()
    query_feedback = f"Previous Expanded Query: {state['expanded_query']}\\nSuggestions: {chain.invoke({'query': state['query'], 'expanded_query': state['expanded_query']})}"
    state['query_feedback'] = query_feedback
    return state

# 3.4. Define Conditional Edges
def should_continue_groundedness(state):
    if state['groundedness_score'] >= 0.7:
        return "check_precision"
    else:
        return "max_iterations_reached" if state["groundedness_loop_count"] >= state['loop_max_iter'] else "refine_response"

def should_continue_precision(state: Dict) -> str:
    if state['precision_score'] >= 0.7:
        return "pass"
    else:
        return "max_iterations_reached" if state['precision_loop_count'] >= state['loop_max_iter'] else "refine_query"

def max_iterations_reached(state: Dict) -> Dict:
    state['response'] = "I'm unable to refine the response further. Please provide more context or clarify your question."
    return state

# 3.5. Assemble the Workflow Graph
def create_workflow() -> StateGraph:
    workflow = StateGraph(AgentState)
    workflow.add_node("expand_query", expand_query)
    workflow.add_node("retrieve_context", retrieve_context)
    workflow.add_node("craft_response", craft_response)
    workflow.add_node("score_groundedness", score_groundedness)
    workflow.add_node("refine_response", refine_response)
    workflow.add_node("check_precision", check_precision)
    workflow.add_node("refine_query", refine_query)
    workflow.add_node("max_iterations_reached", max_iterations_reached)

    workflow.add_edge(START, "expand_query")
    workflow.add_edge("expand_query", "retrieve_context")
    workflow.add_edge("retrieve_context", "craft_response")
    workflow.add_edge("craft_response", "score_groundedness")
    workflow.add_conditional_edges("score_groundedness", should_continue_groundedness, {"check_precision": "check_precision", "refine_response": "refine_response", "max_iterations_reached": "max_iterations_reached"})
    workflow.add_edge("refine_response", "craft_response")
    workflow.add_conditional_edges("check_precision", should_continue_precision, {"pass": END, "refine_query": "refine_query", "max_iterations_reached": "max_iterations_reached"})
    workflow.add_edge("refine_query", "expand_query")
    workflow.add_edge("max_iterations_reached", END)
    return workflow

WORKFLOW_APP = create_workflow().compile()

# 3.6. Create the Agentic RAG Tool
@tool
def agentic_rag(query: str):
    """Runs the RAG-based agent for context-aware responses."""
    inputs = {
        "query": query, "expanded_query": "", "context": [], "response": "",
        "precision_score": 0.0, "groundedness_score": 0.0,
        "groundedness_loop_count": 0, "precision_loop_count": 0,
        "feedback": "", "query_feedback": "", "loop_max_iter": 3
    }
    output = WORKFLOW_APP.invoke(inputs)
    final_response = output.get('response')
    if hasattr(final_response, 'content'):
        return final_response.content
    return str(final_response)


# ==============================================================================
# 4. SAFETY GUARDRAIL
# ==============================================================================
# MODIFIED THIS SECTION TO FIX THE RUNTIME ERROR
llama_guard_client = Groq(
    api_key=groq_api_key,
    http_client=httpx.Client() # Manually pass a standard httpx client
)
def filter_input_with_llama_guard(user_input, model="meta-llama/llama-guard-4-12b"):
    try:
        response = llama_guard_client.chat.completions.create(
            messages=[{"role": "user", "content": user_input}],
            model=model,
            temperature=0.0
        )
        return response.choices[0].message.content.strip()
    except Exception as e:
        print(f"Error with Llama Guard (Groq): {e}")
        return "safe" # Fail-safe


# ==============================================================================
# 5. NUTRITION BOT CLASS (with Memory)
# This class encapsulates the agent, memory, and interaction logic.
# ==============================================================================
class NutritionBot:
    def __init__(self):
        self.memory = MemoryClient(api_key=mem0_api_key)
        self.client = ChatOpenAI(
            model_name="gpt-4o-mini",
            openai_api_key=openai_api_key,
            openai_api_base=openai_api_base,
            temperature=0
        )
        tools = [agentic_rag]
        system_prompt = """You are a Medical Support Agent specializing ONLY in nutritional disorders...""" # (Your full, robust prompt here)
        prompt = ChatPromptTemplate.from_messages([
            ("system", system_prompt),
            ("human", "{input}"),
            ("placeholder", "{agent_scratchpad}")
        ])
        agent = create_tool_calling_agent(self.client, tools, prompt)
        self.agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)

    def store_customer_interaction(self, user_id: str, message: str, response: str, metadata: Dict = None):
        if metadata is None: metadata = {}
        metadata["timestamp"] = datetime.now().isoformat()
        conversation = [{"role": "user", "content": message}, {"role": "assistant", "content": response}]
        self.memory.add(messages=conversation, user_id=user_id, metadata=metadata)

    def get_relevant_history(self, user_id: str, query: str) -> str:
        memories = self.memory.search(query=query, user_id=user_id, limit=5)
        context = "Previous relevant interactions:\\n"
        if not memories:
            return "No previous relevant interactions found.\\n"
        for mem in memories:
             context += f"Summary of past interaction: {mem.get('memory', 'N/A')}\\n---\\n"
        return context

    def handle_customer_query(self, user_id: str, query: str) -> str:
        context = self.get_relevant_history(user_id, query)
        prompt = f"Context:\\n{context}\\nCurrent customer query: {query}\\nProvide a helpful response that takes into account any relevant past interactions."
        response = self.agent_executor.invoke({"input": prompt})
        self.store_customer_interaction(user_id=user_id, message=query, response=response["output"])
        return response['output']


# ==============================================================================
# 6. STREAMLIT UI
# This is the entry point and user interface for the application.
# ==============================================================================
def nutrition_disorder_streamlit():
    st.title("Nutrition Disorder Specialist")
    st.write("Ask me anything about nutrition disorders, symptoms, causes, treatments, and more.")

    if 'chat_history' not in st.session_state:
        st.session_state.chat_history = []
    if 'user_id' not in st.session_state:
        st.session_state.user_id = None
    if 'chatbot' not in st.session_state:
        st.session_state.chatbot = None

    if st.session_state.user_id is None:
        with st.form("login_form"):
            user_id = st.text_input("Please enter your name to begin:")
            submit_button = st.form_submit_button("Login")
            if submit_button and user_id:
                st.session_state.user_id = user_id
                st.session_state.chatbot = NutritionBot()
                welcome_msg = f"Welcome, {user_id}! How can I help you with nutrition disorders today?"
                st.session_state.chat_history.append({"role": "assistant", "content": welcome_msg})
                st.rerun()
    else:
        for message in st.session_state.chat_history:
            with st.chat_message(message["role"]):
                st.write(message["content"])

        if user_query := st.chat_input("Ask about a nutrition disorder..."):
            st.session_state.chat_history.append({"role": "user", "content": user_query})
            with st.chat_message("user"):
                st.write(user_query)

            filtered_result = filter_input_with_llama_guard(user_query)
            if "safe" in filtered_result:
                try:
                    with st.spinner("Thinking..."):
                        response = st.session_state.chatbot.handle_customer_query(st.session_state.user_id, user_query)
                    st.session_state.chat_history.append({"role": "assistant", "content": response})
                    with st.chat_message("assistant"):
                        st.write(response)
                except Exception as e:
                    error_msg = f"Sorry, I encountered an error: {e}"
                    st.session_state.chat_history.append({"role": "assistant", "content": error_msg})
                    with st.chat_message("assistant"):
                        st.write(error_msg)
            else:
                inappropriate_msg = "I apologize, but I cannot process that input as it may be inappropriate."
                st.session_state.chat_history.append({"role": "assistant", "content": inappropriate_msg})
                with st.chat_message("assistant"):
                    st.write(inappropriate_msg)

if __name__ == "__main__":
    nutrition_disorder_streamlit()