File size: 13,485 Bytes
835936b
 
 
 
 
 
 
 
 
 
 
695f802
835936b
 
 
 
8d36e0e
835936b
 
 
 
8d36e0e
835936b
8d36e0e
835936b
 
 
 
 
695f802
 
 
 
 
835936b
8d36e0e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
695f802
835936b
 
695f802
 
835936b
 
 
695f802
 
835936b
 
 
695f802
 
835936b
 
 
695f802
 
835936b
 
 
 
 
 
695f802
835936b
 
695f802
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8d36e0e
 
 
 
 
 
 
 
 
835936b
 
695f802
 
8d36e0e
695f802
 
 
 
8d36e0e
835936b
695f802
 
 
 
835936b
 
 
695f802
 
8d36e0e
695f802
 
 
 
8d36e0e
835936b
695f802
 
 
 
835936b
 
8d36e0e
 
695f802
8d36e0e
 
 
 
695f802
8d36e0e
 
 
695f802
8d36e0e
835936b
695f802
 
 
8d36e0e
695f802
 
 
 
 
 
 
8d36e0e
 
 
 
695f802
 
8d36e0e
 
 
 
 
 
 
 
695f802
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
835936b
695f802
835936b
 
 
695f802
 
8d36e0e
695f802
 
8d36e0e
695f802
 
 
8d36e0e
 
835936b
695f802
 
 
 
835936b
695f802
835936b
8d36e0e
695f802
8d36e0e
 
 
695f802
8d36e0e
 
 
 
 
 
695f802
8d36e0e
 
695f802
8d36e0e
 
 
 
695f802
8d36e0e
835936b
 
 
 
8d36e0e
835936b
695f802
 
8d36e0e
695f802
8d36e0e
835936b
 
 
695f802
8d36e0e
695f802
 
8d36e0e
 
 
 
 
 
 
 
 
7343388
8d36e0e
 
 
 
 
 
 
 
 
835936b
8d36e0e
835936b
8d36e0e
 
 
835936b
 
8d36e0e
695f802
 
8d36e0e
695f802
 
 
8d36e0e
 
835936b
 
8d36e0e
695f802
 
 
 
 
8d36e0e
695f802
 
8d36e0e
695f802
 
8d36e0e
695f802
 
 
 
 
 
835936b
8d36e0e
835936b
 
 
695f802
8d36e0e
835936b
 
 
695f802
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
import os
from dotenv import load_dotenv

# Load environment variables
load_dotenv()

# Set protobuf implementation to avoid C++ extension issues
os.environ["PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION"] = "python"

# Load keys from environment
hf_token = os.getenv("HUGGINGFACE_INFERENCE_TOKEN")
serper_api_key = os.getenv("SERPER_API_KEY")

# ---- Imports ----
from langgraph.graph import START, StateGraph, MessagesState
from langgraph.prebuilt import tools_condition, ToolNode
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_community.tools.tavily_search import TavilySearchResults
from langchain_community.document_loaders import WikipediaLoader, ArxivLoader
from langchain_community.vectorstores import Chroma
from langchain_core.documents import Document
from langchain_core.messages import SystemMessage, HumanMessage, AIMessage
from langchain_core.tools import tool
from langchain_core.language_models.base import BaseLanguageModel
from langchain.tools.retriever import create_retriever_tool
from langchain.vectorstores import Chroma
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.schema import Document
import json
import requests
from typing import List, Dict, Any
import re
import math
from datetime import datetime

# Custom HuggingFace LLM wrapper
class SimpleHuggingFaceLLM(BaseLanguageModel):
    def __init__(self, repo_id: str, hf_token: str):
        super().__init__()
        self.repo_id = repo_id
        self.hf_token = hf_token
        self.api_url = f"https://api-inference.huggingface.co/models/{repo_id}"
        self.headers = {"Authorization": f"Bearer {hf_token}"}
    
    def _generate(self, messages, stop=None, run_manager=None, **kwargs):
        # Convert messages to a single prompt
        if isinstance(messages, list):
            prompt = messages[-1].content if messages else ""
        else:
            prompt = str(messages)
        
        payload = {
            "inputs": prompt,
            "parameters": {
                "max_new_tokens": 512,
                "temperature": 0.1,
                "return_full_text": False
            }
        }
        
        try:
            response = requests.post(self.api_url, headers=self.headers, json=payload)
            if response.status_code == 200:
                result = response.json()
                if isinstance(result, list) and len(result) > 0:
                    generated_text = result[0].get('generated_text', '')
                else:
                    generated_text = str(result)
                
                from langchain_core.outputs import LLMResult, Generation
                return LLMResult(generations=[[Generation(text=generated_text)]])
            else:
                return LLMResult(generations=[[Generation(text=f"Error: {response.status_code}")]])
        except Exception as e:
            return LLMResult(generations=[[Generation(text=f"Error: {str(e)}")]])
    
    def invoke(self, input, config=None, **kwargs):
        if isinstance(input, list):
            prompt = input[-1].content if input else ""
        else:
            prompt = str(input)
        
        result = self._generate(prompt)
        generated_text = result.generations[0][0].text
        return AIMessage(content=generated_text)
    
    @property
    def _llm_type(self):
        return "huggingface_custom"

# ---- Enhanced Tools ----

@tool
def multiply(a: float, b: float) -> float:
    """Multiply two numbers"""
    return a * b

@tool
def add(a: float, b: float) -> float:
    """Add two numbers"""
    return a + b

@tool
def subtract(a: float, b: float) -> float:
    """Subtract two numbers"""
    return a - b

@tool
def divide(a: float, b: float) -> float:
    """Divide two numbers"""
    if b == 0:
        raise ValueError("Cannot divide by zero.")
    return a / b

@tool
def modulus(a: int, b: int) -> int:
    """Calculate modulus of two integers"""
    return a % b

@tool
def power(a: float, b: float) -> float:
    """Calculate a raised to the power of b"""
    return a ** b

@tool
def square_root(a: float) -> float:
    """Calculate square root of a number"""
    return math.sqrt(a)

@tool
def factorial(n: int) -> int:
    """Calculate factorial of a number"""
    if n < 0:
        raise ValueError("Factorial is not defined for negative numbers")
    if n == 0 or n == 1:
        return 1
    result = 1
    for i in range(2, n + 1):
        result *= i
    return result

@tool
def gcd(a: int, b: int) -> int:
    """Calculate greatest common divisor"""
    while b:
        a, b = b, a % b
    return a

@tool
def lcm(a: int, b: int) -> int:
    """Calculate least common multiple"""
    return abs(a * b) // gcd(a, b)

@tool
def percentage(part: float, whole: float) -> float:
    """Calculate percentage"""
    return (part / whole) * 100

@tool
def compound_interest(principal: float, rate: float, time: float, n: int = 1) -> float:
    """Calculate compound interest"""
    return principal * (1 + rate/n) ** (n * time)

@tool
def calculate_average(numbers: str) -> float:
    """Calculate average of comma-separated numbers"""
    try:
        nums = [float(x.strip()) for x in numbers.split(',')]
        return sum(nums) / len(nums)
    except:
        return 0.0

@tool
def wiki_search(query: str) -> str:
    """Search Wikipedia for information"""
    try:
        search_docs = WikipediaLoader(query=query, load_max_docs=2).load()
        if not search_docs:
            return "No Wikipedia results found."
        
        formatted = "\n\n---\n\n".join([
            f'Wikipedia: {doc.metadata.get("title", "Unknown")}\n{doc.page_content[:1500]}'
            for doc in search_docs
        ])
        return formatted
    except Exception as e:
        return f"Wikipedia search error: {str(e)}"

@tool
def web_search(query: str) -> str:
    """Search the web using Tavily"""
    try:
        search_docs = TavilySearchResults(max_results=2).invoke(query=query)
        if not search_docs:
            return "No web search results found."
        
        formatted = "\n\n---\n\n".join([
            f'Web: {doc.get("title", "Unknown")}\n{doc.get("content", "")[:1500]}'
            for doc in search_docs
        ])
        return formatted
    except Exception as e:
        return f"Web search error: {str(e)}"

@tool
def simple_calculation(expression: str) -> str:
    """Safely evaluate simple mathematical expressions"""
    try:
        # Remove any non-mathematical characters for safety
        safe_chars = set('0123456789+-*/.() ')
        if not all(c in safe_chars for c in expression):
            return "Invalid characters in expression"
        
        # Evaluate the expression
        result = eval(expression)
        return str(result)
    except Exception as e:
        return f"Calculation error: {str(e)}"

# ---- Embedding & Vector Store Setup ----
def setup_vector_store():
    try:
        embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
        
        # Check if metadata.jsonl exists and load it
        if os.path.exists('metadata.jsonl'):
            json_QA = []
            with open('metadata.jsonl', 'r') as jsonl_file:
                for line in jsonl_file:
                    if line.strip():  # Skip empty lines
                        try:
                            json_QA.append(json.loads(line))
                        except:
                            continue
            
            if json_QA:
                documents = []
                for sample in json_QA:
                    if sample.get('Question') and sample.get('Final answer'):
                        doc = Document(
                            page_content=f"Question: {sample['Question']}\n\nAnswer: {sample['Final answer']}",
                            metadata={"source": sample.get("task_id", "unknown")}
                        )
                        documents.append(doc)
                
                if documents:
                    vector_store = Chroma.from_documents(
                        documents=documents,
                        embedding=embeddings,
                        persist_directory="./chroma_db",
                        collection_name="my_collection"
                    )
                    vector_store.persist()
                    print(f"Vector store created with {len(documents)} documents")
                    return vector_store
        
        # Create empty vector store if no data
        vector_store = Chroma(
            embedding_function=embeddings,
            persist_directory="./chroma_db",
            collection_name="my_collection"
        )
        print("Empty vector store created")
        return vector_store
        
    except Exception as e:
        print(f"Vector store setup error: {e}")
        return None

vector_store = setup_vector_store()

@tool
def similar_question_search(query: str) -> str:
    """Search for similar questions in the knowledge base"""
    if not vector_store:
        return "No similar questions available"
    
    try:
        matched_docs = vector_store.similarity_search(query, k=2)
        if not matched_docs:
            return "No similar questions found"
        
        formatted = "\n\n".join([
            f'Similar Q&A:\n{doc.page_content[:800]}'
            for doc in matched_docs
        ])
        return formatted
    except Exception as e:
        return f"Similar question search error: {str(e)}"

# ---- Enhanced System Prompt ----
system_prompt = """
You are an expert assistant that can solve various types of questions using available tools.

Available tools:
- Math: add, subtract, multiply, divide, modulus, power, square_root, factorial, gcd, lcm, percentage, compound_interest, calculate_average, simple_calculation
- Search: wiki_search, web_search, similar_question_search

Instructions:
1. Read the question carefully
2. Break down complex problems into steps
3. Use appropriate tools to gather information or perform calculations
4. Think step by step and show your reasoning
5. Provide accurate, concise answers

IMPORTANT: Always end your response with:
FINAL ANSWER: [your answer here]

For the final answer:
- Numbers: Use plain digits (no commas, units, or symbols unless requested)
- Text: Use exact names without articles
- Lists: Comma-separated values

Think carefully and use tools when needed.
"""

sys_msg = SystemMessage(content=system_prompt)

# ---- Tool List ----
tools = [
    # Math tools
    multiply, add, subtract, divide, modulus, power, square_root, 
    factorial, gcd, lcm, percentage, compound_interest, calculate_average, simple_calculation,
    # Search tools
    wiki_search, web_search, similar_question_search
]

# ---- Graph Definition ----
def build_graph(provider: str = "huggingface"):
    """Build the agent graph with custom HuggingFace integration"""
    
    if provider == "huggingface":
        # Use custom HuggingFace LLM with fallback models
        models_to_try = [
            "google/flan-t5-base",
            "microsoft/DialoGPT-medium", 
            "bigscience/bloom-560m"
        ]
        
        llm = None
        for model_id in models_to_try:
            try:
                llm = SimpleHuggingFaceLLM(repo_id=model_id, hf_token=hf_token)
                print(f"Successfully initialized model: {model_id}")
                break
            except Exception as e:
                print(f"Failed to initialize {model_id}: {e}")
                continue
        
        if llm is None:
            raise ValueError("Failed to initialize any HuggingFace model")
    else:
        raise ValueError("Only 'huggingface' provider is supported")

    # Simple tool binding simulation
    def llm_with_tools(messages):
        return llm.invoke(messages)

    def assistant(state: MessagesState):
        """Assistant node with enhanced error handling"""
        try:
            messages = state["messages"]
            response = llm_with_tools(messages)
            return {"messages": [response]}
        except Exception as e:
            print(f"Assistant error: {e}")
            fallback_response = AIMessage(content="I encountered an error processing your request. Let me try a simpler approach.")
            return {"messages": [fallback_response]}

    def retriever(state: MessagesState):
        """Enhanced retriever with context injection"""
        messages = state["messages"]
        user_query = messages[-1].content if messages else ""
        
        context_messages = [sys_msg]
        
        # Add similar question context if available
        if vector_store:
            try:
                similar = vector_store.similarity_search(user_query, k=1)
                if similar:
                    context_msg = HumanMessage(
                        content=f"Here's a similar example:\n{similar[0].page_content[:500]}"
                    )
                    context_messages.append(context_msg)
            except Exception as e:
                print(f"Retriever error: {e}")
        
        return {"messages": context_messages + messages}

    # Build simplified graph (without complex tool routing for now)
    builder = StateGraph(MessagesState)
    builder.add_node("retriever", retriever)
    builder.add_node("assistant", assistant)
    
    # Simple linear flow
    builder.add_edge(START, "retriever")
    builder.add_edge("retriever", "assistant")

    return builder.compile()