File size: 20,166 Bytes
5092552
41f9740
 
5092552
 
d4557ee
 
 
ca98093
41f9740
 
1fa6961
41f9740
ca98093
41f9740
 
 
ca98093
41f9740
a35ea13
41f9740
 
 
 
 
 
 
 
ca98093
41f9740
ca98093
41f9740
 
 
 
 
 
 
 
 
 
 
 
 
 
ca98093
41f9740
f4505e9
86c8869
41f9740
f0e66e7
f4505e9
 
86c8869
41f9740
f0e66e7
1fa6961
f4505e9
86c8869
41f9740
f0e66e7
cc467c2
f4505e9
86c8869
41f9740
86c8869
f4505e9
 
 
 
86c8869
41f9740
f0e66e7
a55679f
f4505e9
41f9740
 
25c1140
41f9740
86c8869
41f9740
 
 
 
 
25c1140
d4557ee
0f81d99
f4505e9
41f9740
 
25c1140
41f9740
 
 
 
 
 
25c1140
41f9740
a55679f
41f9740
 
 
ca98093
41f9740
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ca98093
41f9740
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ca98093
41f9740
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0c69489
41f9740
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ca98093
41f9740
 
ca98093
41f9740
0c69489
41f9740
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ca98093
41f9740
 
 
 
 
 
 
 
 
 
 
 
 
5092552
41f9740
 
ca98093
41f9740
a55679f
41f9740
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5092552
41f9740
 
 
 
 
 
 
 
5092552
41f9740
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5092552
41f9740
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5092552
41f9740
 
 
 
0f81d99
41f9740
 
 
 
 
 
 
 
 
 
 
 
72c7dbb
41f9740
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5092552
41f9740
 
 
0c69489
41f9740
 
 
 
 
 
 
5092552
41f9740
5092552
ca98093
5092552
41f9740
5092552
 
86c8869
41f9740
 
 
 
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
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
"""
Enhanced Multi-LLM Agent System with Supabase FAISS Integration
Complete system for document insertion, retrieval, and question answering
"""

import os
import time
import random
import operator
from typing import List, Dict, Any, TypedDict, Annotated, Optional
from dotenv import load_dotenv

from langchain_core.tools import tool
from langchain_community.tools.tavily_search import TavilySearchResults
from langchain_community.document_loaders import WikipediaLoader
from langgraph.graph import StateGraph, END
from langgraph.checkpoint.memory import MemorySaver
from langchain_core.messages import SystemMessage, HumanMessage, AIMessage
from langchain_groq import ChatGroq

# Supabase and FAISS imports
import faiss
import numpy as np
from sentence_transformers import SentenceTransformer
from supabase import create_client, Client
import pandas as pd
import json
import pickle

load_dotenv()

# Enhanced system prompt for question-answering
ENHANCED_SYSTEM_PROMPT = (
    "You are a helpful assistant tasked with answering questions using a set of tools. "
    "You must provide accurate, comprehensive answers based on available information. "
    "When answering questions, follow these guidelines:\n"
    "1. Use available tools to gather information when needed\n"
    "2. Provide precise, factual answers\n"
    "3. For numbers: don't use commas or units unless specified\n"
    "4. For strings: don't use articles or abbreviations, write digits in plain text\n"
    "5. For lists: apply above rules based on element type\n"
    "6. Always end with 'FINAL ANSWER: [YOUR ANSWER]'\n"
    "7. Be concise but thorough in your reasoning\n"
    "8. If you cannot find the answer, state that clearly"
)

# ---- Tool Definitions ----
@tool
def multiply(a: int, b: int) -> int:
    """Multiply two integers and return the product."""
    return a * b

@tool
def add(a: int, b: int) -> int:
    """Add two integers and return the sum."""
    return a + b

@tool
def subtract(a: int, b: int) -> int:
    """Subtract the second integer from the first and return the difference."""
    return a - b

@tool
def divide(a: int, b: int) -> float:
    """Divide the first integer by the second and return the quotient."""
    if b == 0:
        raise ValueError("Cannot divide by zero.")
    return a / b

@tool
def modulus(a: int, b: int) -> int:
    """Return the remainder when dividing the first integer by the second."""
    return a % b

@tool
def optimized_web_search(query: str) -> str:
    """Perform an optimized web search using TavilySearchResults."""
    try:
        time.sleep(random.uniform(0.7, 1.5))
        search_tool = TavilySearchResults(max_results=3)
        docs = search_tool.invoke({"query": query})
        return "\n\n---\n\n".join(
            f"<Doc url='{d.get('url','')}'>{d.get('content','')[:800]}</Doc>"
            for d in docs
        )
    except Exception as e:
        return f"Web search failed: {e}"

@tool
def optimized_wiki_search(query: str) -> str:
    """Perform an optimized Wikipedia search and return content snippets."""
    try:
        time.sleep(random.uniform(0.3, 1))
        docs = WikipediaLoader(query=query, load_max_docs=2).load()
        return "\n\n---\n\n".join(
            f"<Doc src='{d.metadata.get('source','Wikipedia')}'>{d.page_content[:1000]}</Doc>"
            for d in docs
        )
    except Exception as e:
        return f"Wikipedia search failed: {e}"

# ---- Supabase FAISS Vector Database Integration ----
class SupabaseFAISSVectorDB:
    """Enhanced vector database combining FAISS with Supabase for persistent storage"""
    
    def __init__(self):
        # Initialize Supabase client
        self.supabase_url = os.getenv("SUPABASE_URL")
        self.supabase_key = os.getenv("SUPABASE_SERVICE_KEY")
        if self.supabase_url and self.supabase_key:
            self.supabase: Client = create_client(self.supabase_url, self.supabase_key)
        else:
            self.supabase = None
            print("Supabase credentials not found, running without vector database")
        
        # Initialize embedding model
        self.embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
        self.embedding_dim = self.embedding_model.get_sentence_embedding_dimension()
        
        # Initialize FAISS index
        self.index = faiss.IndexFlatL2(self.embedding_dim)
        self.document_store = []  # Local cache for documents
    
    def insert_question_data(self, data: Dict[str, Any]) -> bool:
        """Insert question data into both Supabase and FAISS"""
        try:
            question_text = data.get("Question", "")
            embedding = self.embedding_model.encode([question_text])[0]
            
            # Insert into Supabase if available
            if self.supabase:
                question_data = {
                    "task_id": data.get("task_id"),
                    "question": question_text,
                    "final_answer": data.get("Final answer"),
                    "level": data.get("Level"),
                    "file_name": data.get("file_name", ""),
                    "embedding": embedding.tolist()
                }
                self.supabase.table("questions").insert(question_data).execute()
            
            # Add to local FAISS index
            self.index.add(embedding.reshape(1, -1).astype('float32'))
            self.document_store.append({
                "task_id": data.get("task_id"),
                "question": question_text,
                "answer": data.get("Final answer"),
                "level": data.get("Level")
            })
            
            return True
        except Exception as e:
            print(f"Error inserting data: {e}")
            return False
    
    def search_similar_questions(self, query: str, k: int = 3) -> List[Dict[str, Any]]:
        """Search for similar questions using vector similarity"""
        try:
            if self.index.ntotal == 0:
                return []
            
            query_embedding = self.embedding_model.encode([query])[0]
            k = min(k, self.index.ntotal)
            distances, indices = self.index.search(
                query_embedding.reshape(1, -1).astype('float32'), k
            )
            
            results = []
            for i, idx in enumerate(indices[0]):
                if 0 <= idx < len(self.document_store):
                    doc = self.document_store[idx]
                    results.append({
                        "task_id": doc["task_id"],
                        "question": doc["question"],
                        "answer": doc["answer"],
                        "similarity_score": 1 / (1 + distances[0][i]),
                        "distance": float(distances[0][i])
                    })
            
            return results
        except Exception as e:
            print(f"Error searching similar questions: {e}")
            return []

# ---- Enhanced Agent State ----
class EnhancedAgentState(TypedDict):
    """State structure for the enhanced multi-LLM agent system."""
    messages: Annotated[List[HumanMessage | AIMessage], operator.add]
    query: str
    agent_type: str
    final_answer: str
    perf: Dict[str, Any]
    agno_resp: str
    tools_used: List[str]
    reasoning: str
    similar_questions: List[Dict[str, Any]]

# ---- Enhanced Multi-LLM System ----
class HybridLangGraphMultiLLMSystem:
    """
    Advanced question-answering system with multi-LLM support and vector database integration
    """
    
    def __init__(self, provider="groq"):
        self.provider = provider
        self.tools = [
            multiply, add, subtract, divide, modulus,
            optimized_web_search, optimized_wiki_search
        ]
        
        # Initialize vector database
        self.vector_db = SupabaseFAISSVectorDB()
        
        self.graph = self._build_graph()

    def _llm(self, model_name: str) -> ChatGroq:
        """Create a Groq LLM instance."""
        return ChatGroq(
            model=model_name,
            temperature=0,
            api_key=os.getenv("GROQ_API_KEY")
        )

    def _build_graph(self) -> StateGraph:
        """Build the LangGraph state machine with enhanced capabilities."""
        # Initialize LLMs
        llama8_llm = self._llm("llama3-8b-8192")
        llama70_llm = self._llm("llama3-70b-8192")
        deepseek_llm = self._llm("deepseek-chat")

        def router(st: EnhancedAgentState) -> EnhancedAgentState:
            """Route queries to appropriate LLM based on complexity and content analysis."""
            q = st["query"].lower()
            
            # Enhanced routing logic
            if any(keyword in q for keyword in ["calculate", "compute", "math", "multiply", "add", "subtract", "divide"]):
                t = "llama70"  # Use more powerful model for calculations
            elif any(keyword in q for keyword in ["search", "find", "lookup", "wikipedia", "information about"]):
                t = "search_enhanced"  # Use search-enhanced processing
            elif "deepseek" in q or any(keyword in q for keyword in ["analyze", "reasoning", "complex"]):
                t = "deepseek"
            elif "llama-8" in q:
                t = "llama8"
            elif len(q.split()) > 20:  # Complex queries
                t = "llama70"
            else:
                t = "llama8"  # Default for simple queries
                
            # Search for similar questions
            similar_questions = self.vector_db.search_similar_questions(st["query"], k=3)
            
            return {**st, "agent_type": t, "tools_used": [], "reasoning": "", "similar_questions": similar_questions}

        def llama8_node(st: EnhancedAgentState) -> EnhancedAgentState:
            """Process query with Llama-3 8B model."""
            t0 = time.time()
            try:
                # Add similar questions context if available
                context = ""
                if st.get("similar_questions"):
                    context = "\n\nSimilar questions for reference:\n"
                    for sq in st["similar_questions"][:2]:
                        context += f"Q: {sq['question']}\nA: {sq['answer']}\n"
                
                enhanced_query = f"""
                Question: {st["query"]}
                {context}
                Please provide a direct, accurate answer to this question.
                """
                
                sys = SystemMessage(content=ENHANCED_SYSTEM_PROMPT)
                res = llama8_llm.invoke([sys, HumanMessage(content=enhanced_query)])
                
                answer = res.content.strip()
                if "FINAL ANSWER:" in answer:
                    answer = answer.split("FINAL ANSWER:")[-1].strip()
                
                return {**st,
                        "final_answer": answer,
                        "reasoning": "Used Llama-3 8B with similar questions context",
                        "perf": {"time": time.time() - t0, "prov": "Groq-Llama3-8B"}}
            except Exception as e:
                return {**st, "final_answer": f"Error: {e}", "perf": {"error": str(e)}}

        def llama70_node(st: EnhancedAgentState) -> EnhancedAgentState:
            """Process query with Llama-3 70B model."""
            t0 = time.time()
            try:
                # Add similar questions context if available
                context = ""
                if st.get("similar_questions"):
                    context = "\n\nSimilar questions for reference:\n"
                    for sq in st["similar_questions"][:2]:
                        context += f"Q: {sq['question']}\nA: {sq['answer']}\n"
                
                enhanced_query = f"""
                Question: {st["query"]}
                {context}
                Please provide a direct, accurate answer to this question.
                """
                
                sys = SystemMessage(content=ENHANCED_SYSTEM_PROMPT)
                res = llama70_llm.invoke([sys, HumanMessage(content=enhanced_query)])
                
                answer = res.content.strip()
                if "FINAL ANSWER:" in answer:
                    answer = answer.split("FINAL ANSWER:")[-1].strip()
                
                return {**st,
                        "final_answer": answer,
                        "reasoning": "Used Llama-3 70B for complex reasoning with context",
                        "perf": {"time": time.time() - t0, "prov": "Groq-Llama3-70B"}}
            except Exception as e:
                return {**st, "final_answer": f"Error: {e}", "perf": {"error": str(e)}}

        def deepseek_node(st: EnhancedAgentState) -> EnhancedAgentState:
            """Process query with DeepSeek model."""
            t0 = time.time()
            try:
                # Add similar questions context if available
                context = ""
                if st.get("similar_questions"):
                    context = "\n\nSimilar questions for reference:\n"
                    for sq in st["similar_questions"][:2]:
                        context += f"Q: {sq['question']}\nA: {sq['answer']}\n"
                
                enhanced_query = f"""
                Question: {st["query"]}
                {context}
                Please provide a direct, accurate answer to this question.
                """
                
                sys = SystemMessage(content=ENHANCED_SYSTEM_PROMPT)
                res = deepseek_llm.invoke([sys, HumanMessage(content=enhanced_query)])
                
                answer = res.content.strip()
                if "FINAL ANSWER:" in answer:
                    answer = answer.split("FINAL ANSWER:")[-1].strip()
                
                return {**st,
                        "final_answer": answer,
                        "reasoning": "Used DeepSeek for advanced reasoning and analysis",
                        "perf": {"time": time.time() - t0, "prov": "Groq-DeepSeek"}}
            except Exception as e:
                return {**st, "final_answer": f"Error: {e}", "perf": {"error": str(e)}}

        def search_enhanced_node(st: EnhancedAgentState) -> EnhancedAgentState:
            """Process query with search enhancement."""
            t0 = time.time()
            tools_used = []
            
            try:
                # Determine search strategy
                query = st["query"]
                search_results = ""
                
                if any(keyword in query.lower() for keyword in ["wikipedia", "wiki"]):
                    search_results = optimized_wiki_search.invoke({"query": query})
                    tools_used.append("wikipedia_search")
                else:
                    search_results = optimized_web_search.invoke({"query": query})
                    tools_used.append("web_search")
                
                # Add similar questions context
                context = ""
                if st.get("similar_questions"):
                    context = "\n\nSimilar questions for reference:\n"
                    for sq in st["similar_questions"][:2]:
                        context += f"Q: {sq['question']}\nA: {sq['answer']}\n"
                
                enhanced_query = f"""
                Original Question: {query}
                
                Search Results:
                {search_results}
                {context}
                
                Based on the search results and similar questions above, provide a direct answer to the original question.
                """
                
                sys = SystemMessage(content=ENHANCED_SYSTEM_PROMPT)
                res = llama70_llm.invoke([sys, HumanMessage(content=enhanced_query)])
                
                answer = res.content.strip()
                if "FINAL ANSWER:" in answer:
                    answer = answer.split("FINAL ANSWER:")[-1].strip()
                
                return {**st,
                        "final_answer": answer,
                        "tools_used": tools_used,
                        "reasoning": "Used search enhancement with similar questions context",
                        "perf": {"time": time.time() - t0, "prov": "Search-Enhanced-Llama70"}}
            except Exception as e:
                return {**st, "final_answer": f"Error: {e}", "perf": {"error": str(e)}}

        # Build graph
        g = StateGraph(EnhancedAgentState)
        g.add_node("router", router)
        g.add_node("llama8", llama8_node)
        g.add_node("llama70", llama70_node)
        g.add_node("deepseek", deepseek_node)
        g.add_node("search_enhanced", search_enhanced_node)
        
        g.set_entry_point("router")
        g.add_conditional_edges("router", lambda s: s["agent_type"], {
            "llama8": "llama8",
            "llama70": "llama70",
            "deepseek": "deepseek",
            "search_enhanced": "search_enhanced"
        })
        
        for node in ["llama8", "llama70", "deepseek", "search_enhanced"]:
            g.add_edge(node, END)
            
        return g.compile(checkpointer=MemorySaver())

    def process_query(self, q: str) -> str:
        """Process a query through the enhanced multi-LLM system."""
        state = {
            "messages": [HumanMessage(content=q)],
            "query": q,
            "agent_type": "",
            "final_answer": "",
            "perf": {},
            "agno_resp": "",
            "tools_used": [],
            "reasoning": "",
            "similar_questions": []
        }
        cfg = {"configurable": {"thread_id": f"enhanced_qa_{hash(q)}"}}
        
        try:
            out = self.graph.invoke(state, cfg)
            answer = out.get("final_answer", "").strip()
            
            # Ensure we don't return the question as the answer
            if answer == q or answer.startswith(q):
                return "Information not available"
            
            return answer if answer else "No answer generated"
        except Exception as e:
            return f"Error processing query: {e}"
    
    def load_metadata_from_jsonl(self, jsonl_file_path: str) -> int:
        """Load question metadata from JSONL file into vector database"""
        success_count = 0
        
        try:
            with open(jsonl_file_path, 'r', encoding='utf-8') as file:
                for line_num, line in enumerate(file, 1):
                    try:
                        data = json.loads(line.strip())
                        if self.vector_db.insert_question_data(data):
                            success_count += 1
                        
                        if line_num % 10 == 0:
                            print(f"Processed {line_num} records, {success_count} successful")
                            
                    except json.JSONDecodeError as e:
                        print(f"JSON decode error on line {line_num}: {e}")
                    except Exception as e:
                        print(f"Error processing line {line_num}: {e}")
        
        except FileNotFoundError:
            print(f"File not found: {jsonl_file_path}")
        
        print(f"Loaded {success_count} questions into vector database")
        return success_count

def build_graph(provider: str | None = None) -> StateGraph:
    """Build and return the graph for the enhanced agent system."""
    return HybridLangGraphMultiLLMSystem(provider or "groq").graph

if __name__ == "__main__":
    # Initialize and test the system
    system = HybridLangGraphMultiLLMSystem()
    
    # Load metadata if available
    if os.path.exists("metadata.jsonl"):
        system.load_metadata_from_jsonl("metadata.jsonl")
    
    # Test queries
    test_questions = [
        "How many studio albums were published by Mercedes Sosa between 2000 and 2009?",
        "What is 25 multiplied by 17?",
        "Find information about artificial intelligence on Wikipedia"
    ]
    
    for question in test_questions:
        print(f"Question: {question}")
        answer = system.process_query(question)
        print(f"Answer: {answer}")
        print("-" * 50)