Spaces:
Sleeping
Sleeping
Removed useless code and cleaned the pipeline.py
Browse files- pipeline.py +168 -361
pipeline.py
CHANGED
@@ -10,7 +10,6 @@ from collections import OrderedDict
|
|
10 |
import pandas as pd
|
11 |
from pydantic import BaseModel, Field, ValidationError, validator
|
12 |
|
13 |
-
# NLTK for input validation
|
14 |
import nltk
|
15 |
from nltk.corpus import words
|
16 |
try:
|
@@ -19,7 +18,6 @@ except LookupError:
|
|
19 |
nltk.download('words')
|
20 |
english_words = set(words.words())
|
21 |
|
22 |
-
# LangChain / Groq / LLM imports
|
23 |
from langchain_groq import ChatGroq
|
24 |
from langchain_community.embeddings import HuggingFaceEmbeddings
|
25 |
from langchain_community.vectorstores import FAISS
|
@@ -28,35 +26,15 @@ from langchain.prompts import PromptTemplate
|
|
28 |
from langchain.docstore.document import Document
|
29 |
from langchain_core.caches import BaseCache
|
30 |
from langchain_core.callbacks import Callbacks
|
31 |
-
# from langchain_core.callbacks import CallbackManager
|
32 |
-
# from langchain.callbacks.base import BaseCallbacks # Updated import
|
33 |
-
# from langchain.callbacks.manager import CallbackManager
|
34 |
-
# from langchain.callbacks import StdOutCallbackHandler
|
35 |
|
36 |
-
# Custom chain imports
|
37 |
-
# from groq_client import GroqClient
|
38 |
from chain.classification_chain import get_classification_chain
|
39 |
from chain.refusal_chain import get_refusal_chain
|
40 |
from chain.tailor_chain import get_tailor_chain
|
41 |
from chain.cleaner_chain import get_cleaner_chain
|
42 |
from chain.tailor_chain_wellnessBrand import get_tailor_chain_wellnessBrand
|
43 |
|
44 |
-
# Mistral moderation
|
45 |
from mistralai import Mistral
|
46 |
|
47 |
-
# Google Gemini LLM
|
48 |
-
# from langchain_google_genai import ChatGoogleGenerativeAI
|
49 |
-
|
50 |
-
# Web search
|
51 |
-
# from smolagents import DuckDuckGoSearchTool, ManagedAgent, HfApiModel, CodeAgent
|
52 |
-
# from openinference.instrumentation.smolagents import SmolagentsInstrumentor
|
53 |
-
# from phoenix.otel import register
|
54 |
-
|
55 |
-
|
56 |
-
# register()
|
57 |
-
# SmolagentsInstrumentor().instrument(skip_dep_check=True)
|
58 |
-
|
59 |
-
|
60 |
from smolagents import (
|
61 |
CodeAgent,
|
62 |
DuckDuckGoSearchTool,
|
@@ -65,9 +43,7 @@ from smolagents import (
|
|
65 |
VisitWebpageTool,
|
66 |
)
|
67 |
|
68 |
-
|
69 |
-
from chain.prompts import selfharm_prompt, frustration_prompt, ethical_conflict_prompt,classification_prompt, refusal_prompt, tailor_prompt, cleaner_prompt
|
70 |
-
|
71 |
|
72 |
logging.basicConfig(level=logging.INFO)
|
73 |
logger = logging.getLogger(__name__)
|
@@ -75,19 +51,12 @@ logger = logging.getLogger(__name__)
|
|
75 |
from langchain_core.tracers import LangChainTracer
|
76 |
from langsmith import Client
|
77 |
|
|
|
|
|
|
|
|
|
78 |
|
79 |
-
os.environ["LANGCHAIN_TRACING_V2"]="true"
|
80 |
-
os.environ["LANGSMITH_ENDPOINT"]="https://api.smith.langchain.com"
|
81 |
-
# langsmith_client = Client()
|
82 |
-
os.environ["LANGCHAIN_API_KEY"]=os.getenv("LANGCHAIN_API_KEY")
|
83 |
-
os.environ["LANGCHAIN_PROJECT"]=os.getenv("LANGCHAIN_PROJECT")
|
84 |
-
# tracer = LangChainTracer(project_name=os.environ.get("LANGCHAIN_PROJECT", "healthy_ai_expert"))
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
# -------------------------------------------------------
|
89 |
# Basic Models
|
90 |
-
# -------------------------------------------------------
|
91 |
class QueryInput(BaseModel):
|
92 |
query: str = Field(..., min_length=1)
|
93 |
|
@@ -115,9 +84,7 @@ class ProcessingMetrics(BaseModel):
|
|
115 |
/ self.total_requests
|
116 |
)
|
117 |
|
118 |
-
# -------------------------------------------------------
|
119 |
# Mistral Moderation
|
120 |
-
# -------------------------------------------------------
|
121 |
class ModerationResult(BaseModel):
|
122 |
is_safe: bool
|
123 |
categories: Dict[str, bool]
|
@@ -127,9 +94,7 @@ mistral_api_key = os.environ.get("MISTRAL_API_KEY")
|
|
127 |
client = Mistral(api_key=mistral_api_key)
|
128 |
|
129 |
def moderate_text(query: str) -> ModerationResult:
|
130 |
-
"""
|
131 |
-
Uses Mistral's moderation to detect unsafe content.
|
132 |
-
"""
|
133 |
try:
|
134 |
query_input = QueryInput(query=query)
|
135 |
response = client.classifiers.moderate_chat(
|
@@ -161,53 +126,27 @@ def moderate_text(query: str) -> ModerationResult:
|
|
161 |
raise RuntimeError(f"Moderation failed: {e}")
|
162 |
|
163 |
def compute_moderation_severity(mresult: ModerationResult) -> float:
|
|
|
164 |
severity = 0.0
|
165 |
for flag in mresult.categories.values():
|
166 |
if flag:
|
167 |
severity += 0.3
|
168 |
return min(severity, 1.0)
|
169 |
|
170 |
-
# -------------------------------------------------------
|
171 |
# Models
|
172 |
-
# -------------------------------------------------------
|
173 |
GROQ_MODELS = {
|
174 |
-
"default":
|
175 |
"classification": "qwen-qwq-32b",
|
176 |
-
"moderation":
|
177 |
-
"combination":
|
178 |
}
|
179 |
|
180 |
MAX_RETRIES = 3
|
181 |
RATE_LIMIT_REQUESTS = 60
|
182 |
CACHE_SIZE_LIMIT = 1000
|
183 |
|
184 |
-
# Google Gemini (primary)
|
185 |
-
# GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY")
|
186 |
-
# gemini_llm = ChatGoogleGenerativeAI(
|
187 |
-
# model="gemini-2.0-flash",
|
188 |
-
# temperature=0.5,
|
189 |
-
# max_tokens=None,
|
190 |
-
# timeout=None,
|
191 |
-
# max_retries=2,
|
192 |
-
# )
|
193 |
-
|
194 |
-
# # Fallback
|
195 |
-
# fallback_groq_api_key = os.environ.get("GROQ_API_KEY_FALLBACK", "GROQ_API_KEY")
|
196 |
-
|
197 |
-
# # Attempt to initialize ChatGroq without a cache
|
198 |
-
# try:
|
199 |
-
# groq_fallback_llm = ChatGroq(
|
200 |
-
# model=GROQ_MODELS["default"],
|
201 |
-
# temperature=0.7,
|
202 |
-
# # groq_api_key=fallback_groq_api_key,
|
203 |
-
# max_tokens=2048
|
204 |
-
# )
|
205 |
-
# except Exception as e:
|
206 |
-
# logger.error(f"Failed to initialize ChatGroq: {e}")
|
207 |
-
# raise RuntimeError("ChatGroq initialization failed.") from e
|
208 |
-
# Define a simple no-op cache class
|
209 |
class NoCache(BaseCache):
|
210 |
-
"""
|
211 |
def __init__(self):
|
212 |
pass
|
213 |
|
@@ -220,28 +159,27 @@ class NoCache(BaseCache):
|
|
220 |
def clear(self):
|
221 |
pass
|
222 |
|
223 |
-
# Rebuild the ChatGroq model after defining NoCache
|
224 |
ChatGroq.model_rebuild()
|
225 |
-
|
226 |
try:
|
227 |
fallback_groq_api_key = os.environ.get("GROQ_API_KEY_FALLBACK", os.environ.get("GROQ_API_KEY"))
|
228 |
if not fallback_groq_api_key:
|
229 |
logger.warning("No Groq API key found for fallback LLM")
|
230 |
groq_fallback_llm = ChatGroq(
|
231 |
-
model=GROQ_MODELS["default"],
|
232 |
temperature=0.7,
|
233 |
groq_api_key=fallback_groq_api_key,
|
234 |
max_tokens=2048,
|
235 |
-
cache=NoCache(),
|
236 |
-
callbacks=[]
|
237 |
)
|
238 |
except Exception as e:
|
239 |
logger.error(f"Failed to initialize fallback Groq LLM: {e}")
|
240 |
raise RuntimeError("ChatGroq initialization failed.") from e
|
241 |
-
|
242 |
# Rate-limit & Cache
|
243 |
-
# -------------------------------------------------------
|
244 |
def handle_rate_limiting(state: "PipelineState") -> bool:
|
|
|
245 |
current_time = time.time()
|
246 |
one_min_ago = current_time - 60
|
247 |
state.request_timestamps = [t for t in state.request_timestamps if t > one_min_ago]
|
@@ -251,6 +189,7 @@ def handle_rate_limiting(state: "PipelineState") -> bool:
|
|
251 |
return True
|
252 |
|
253 |
def manage_cache(state: "PipelineState", query: str, response: str = None) -> Optional[str]:
|
|
|
254 |
cache_key = query.strip().lower()
|
255 |
if response is None:
|
256 |
return state.cache.get(cache_key)
|
@@ -262,17 +201,16 @@ def manage_cache(state: "PipelineState", query: str, response: str = None) -> Op
|
|
262 |
return None
|
263 |
|
264 |
def create_error_response(error_type: str, details: str = "") -> str:
|
|
|
265 |
templates = {
|
266 |
"validation": "I couldn't process your query: {details}",
|
267 |
"processing": "I encountered an error while processing: {details}",
|
268 |
"rate_limit": "Too many requests. Please try again soon.",
|
269 |
-
"general":
|
270 |
}
|
271 |
return templates.get(error_type, templates["general"]).format(details=details)
|
272 |
|
273 |
-
# -------------------------------------------------------
|
274 |
# Web Search
|
275 |
-
# -------------------------------------------------------
|
276 |
web_search_cache: Dict[str, str] = {}
|
277 |
|
278 |
def store_websearch_result(query: str, result: str):
|
@@ -282,6 +220,7 @@ def retrieve_websearch_result(query: str) -> Optional[str]:
|
|
282 |
return web_search_cache.get(query.strip().lower())
|
283 |
|
284 |
def do_web_search(query: str) -> str:
|
|
|
285 |
try:
|
286 |
cached = retrieve_websearch_result(query)
|
287 |
if cached:
|
@@ -289,26 +228,17 @@ def do_web_search(query: str) -> str:
|
|
289 |
return cached
|
290 |
|
291 |
logger.info("Performing a new web search for: '%s'", query)
|
292 |
-
# model = HfApiModel()
|
293 |
-
# search_tool = DuckDuckGoSearchTool()
|
294 |
-
# web_agent = CodeAgent(tools=[search_tool], model=model)
|
295 |
-
|
296 |
-
# managed_web_agent = ManagedAgent(
|
297 |
-
# agent=web_agent,
|
298 |
-
# name="web_search",
|
299 |
-
# description="Runs a web search. Provide your query."
|
300 |
-
# )
|
301 |
search_agent = ToolCallingAgent(
|
302 |
-
|
303 |
-
|
304 |
-
|
305 |
-
|
306 |
)
|
307 |
|
308 |
manager_agent = CodeAgent(
|
309 |
tools=[],
|
310 |
-
model=
|
311 |
-
managed_agents=[
|
312 |
)
|
313 |
|
314 |
new_search_result = manager_agent.run(f"Search for information about: {query}")
|
@@ -319,34 +249,21 @@ def do_web_search(query: str) -> str:
|
|
319 |
return ""
|
320 |
|
321 |
def is_greeting(query: str) -> bool:
|
322 |
-
"""
|
323 |
-
Returns True if the query is a greeting. This check is designed to be
|
324 |
-
lenient enough to catch common greetings even with minor spelling mistakes
|
325 |
-
or punctuation.
|
326 |
-
"""
|
327 |
-
# Define a set of common greeting words (you can add variants or use fuzzy matching if needed)
|
328 |
greetings = {"hello", "hi", "hey", "hii", "hola", "greetings"}
|
329 |
-
|
330 |
-
# Remove punctuation and extra whitespace, and lower the case.
|
331 |
cleaned = re.sub(r'[^\w\s]', '', query).strip().lower()
|
332 |
-
|
333 |
-
# Split the cleaned text into words.
|
334 |
words_in_query = set(cleaned.split())
|
335 |
-
|
336 |
-
# Return True if any of the greeting words are in the query.
|
337 |
return not words_in_query.isdisjoint(greetings)
|
338 |
|
339 |
-
|
340 |
-
# -------------------------------------------------------
|
341 |
# Vector Stores & RAG
|
342 |
-
# -------------------------------------------------------
|
343 |
def build_or_load_vectorstore(csv_path: str, store_dir: str) -> FAISS:
|
|
|
344 |
if os.path.exists(store_dir):
|
345 |
logger.info(f"Loading existing FAISS store from {store_dir}")
|
346 |
embeddings = HuggingFaceEmbeddings(
|
347 |
model_name="sentence-transformers/multi-qa-mpnet-base-dot-v1"
|
348 |
)
|
349 |
-
return FAISS.load_local(store_dir, embeddings,allow_dangerous_deserialization=True)
|
350 |
else:
|
351 |
logger.info(f"Building new FAISS store from {csv_path}")
|
352 |
df = pd.read_csv(csv_path)
|
@@ -373,8 +290,9 @@ def build_or_load_vectorstore(csv_path: str, store_dir: str) -> FAISS:
|
|
373 |
vectorstore = FAISS.from_documents(docs, embedding=embeddings)
|
374 |
vectorstore.save_local(store_dir)
|
375 |
return vectorstore
|
376 |
-
|
377 |
def build_rag_chain(vectorstore: FAISS, llm) -> RetrievalQA:
|
|
|
378 |
prompt = PromptTemplate(
|
379 |
template="""
|
380 |
[INST] You are an AI wellness assistant speaking directly to a user who has asked: "{question}"
|
@@ -408,8 +326,9 @@ def build_rag_chain(vectorstore: FAISS, llm) -> RetrievalQA:
|
|
408 |
}
|
409 |
)
|
410 |
return chain
|
411 |
-
|
412 |
def build_rag_chain2(vectorstore: FAISS, llm) -> RetrievalQA:
|
|
|
413 |
prompt = PromptTemplate(
|
414 |
template="""
|
415 |
[INST] You are the brand strategy advisor for Healthy AI Expert. A team member has asked: "{question}"
|
@@ -425,7 +344,6 @@ def build_rag_chain2(vectorstore: FAISS, llm) -> RetrievalQA:
|
|
425 |
|
426 |
Remember our key brand pillars: AI-driven personalization, scientific credibility, user-centric design, and innovation leadership.
|
427 |
[/INST]
|
428 |
-
|
429 |
""",
|
430 |
input_variables=["context", "question"]
|
431 |
)
|
@@ -444,9 +362,7 @@ def build_rag_chain2(vectorstore: FAISS, llm) -> RetrievalQA:
|
|
444 |
)
|
445 |
return chain
|
446 |
|
447 |
-
# -------------------------------------------------------
|
448 |
# PipelineState
|
449 |
-
# -------------------------------------------------------
|
450 |
class PipelineState:
|
451 |
_instance = None
|
452 |
|
@@ -462,6 +378,7 @@ class PipelineState:
|
|
462 |
self._initialize()
|
463 |
|
464 |
def _initialize(self):
|
|
|
465 |
try:
|
466 |
self.metrics = ProcessingMetrics()
|
467 |
self.error_count = 0
|
@@ -478,52 +395,31 @@ class PipelineState:
|
|
478 |
raise RuntimeError("Pipeline initialization failed.") from e
|
479 |
|
480 |
def _setup_chains(self):
|
481 |
-
|
482 |
self.tailor_chainWellnessBrand = get_tailor_chain_wellnessBrand()
|
483 |
self.classification_chain = get_classification_chain()
|
484 |
-
self.refusal_chain
|
485 |
-
self.tailor_chain
|
486 |
-
self.cleaner_chain
|
487 |
|
488 |
-
# Specialized chain for self-harm
|
489 |
-
from chain.prompts import selfharm_prompt
|
490 |
-
# self.self_harm_chain = LLMChain(llm=gemini_llm, prompt=selfharm_prompt, verbose=False)
|
491 |
-
|
492 |
self.self_harm_chain = LLMChain(llm=groq_fallback_llm, prompt=selfharm_prompt, verbose=False)
|
493 |
-
|
494 |
-
|
495 |
-
# NEW: chain for frustration/harsh queries
|
496 |
-
from chain.prompts import frustration_prompt
|
497 |
-
# self.frustration_chain = LLMChain(llm=gemini_llm, prompt=frustration_prompt, verbose=False)
|
498 |
self.frustration_chain = LLMChain(llm=groq_fallback_llm, prompt=frustration_prompt, verbose=False)
|
499 |
-
|
500 |
-
|
501 |
-
# NEW: chain for ethical conflict queries
|
502 |
-
from chain.prompts import ethical_conflict_prompt
|
503 |
-
# self.ethical_conflict_chain = LLMChain(llm=gemini_llm, prompt=ethical_conflict_prompt, verbose=False)
|
504 |
self.ethical_conflict_chain = LLMChain(llm=groq_fallback_llm, prompt=ethical_conflict_prompt, verbose=False)
|
505 |
|
506 |
-
|
507 |
-
|
508 |
-
brand_store = "faiss_brand_store"
|
509 |
wellness_csv = "dataset/AIChatbot.csv"
|
510 |
wellness_store = "faiss_wellness_store"
|
511 |
|
512 |
-
brand_vs
|
513 |
wellness_vs = build_or_load_vectorstore(wellness_csv, wellness_store)
|
514 |
|
515 |
-
# Default LLM & fallback
|
516 |
-
# self.gemini_llm = gemini_llm
|
517 |
self.groq_fallback_llm = groq_fallback_llm
|
518 |
-
|
519 |
-
# self.brand_rag_chain = build_rag_chain2(brand_vs, self.gemini_llm)
|
520 |
-
# self.wellness_rag_chain = build_rag_chain(wellness_vs, self.gemini_llm)
|
521 |
-
self.brand_rag_chain = build_rag_chain2(brand_vs, self.groq_fallback_llm)
|
522 |
self.wellness_rag_chain = build_rag_chain(wellness_vs, self.groq_fallback_llm)
|
523 |
-
# self.brand_rag_chain_fallback = build_rag_chain2(brand_vs, self.groq_fallback_llm)
|
524 |
-
# self.wellness_rag_chain_fallback = build_rag_chain(wellness_vs, self.groq_fallback_llm)
|
525 |
|
526 |
def handle_error(self, error: Exception) -> bool:
|
|
|
527 |
self.error_count += 1
|
528 |
self.metrics.errors += 1
|
529 |
if self.error_count >= MAX_RETRIES:
|
@@ -533,6 +429,7 @@ class PipelineState:
|
|
533 |
return True
|
534 |
|
535 |
def reset(self):
|
|
|
536 |
try:
|
537 |
logger.info("Resetting pipeline state.")
|
538 |
old_metrics = self.metrics
|
@@ -548,6 +445,7 @@ class PipelineState:
|
|
548 |
raise RuntimeError("Failed to reset pipeline.")
|
549 |
|
550 |
def get_metrics(self) -> Dict[str, Any]:
|
|
|
551 |
uptime = (datetime.now() - self.metrics.last_reset).total_seconds() / 3600
|
552 |
return {
|
553 |
"total_requests": self.metrics.total_requests,
|
@@ -558,20 +456,15 @@ class PipelineState:
|
|
558 |
}
|
559 |
|
560 |
def update_metrics(self, start_time: float, is_cache_hit: bool = False):
|
|
|
561 |
duration = time.time() - start_time
|
562 |
self.metrics.update_metrics(duration, is_cache_hit)
|
563 |
|
564 |
pipeline_state = PipelineState()
|
565 |
|
566 |
-
#
|
567 |
-
# Helper checks: detect aggression or ethical conflict
|
568 |
-
# -------------------------------------------------------
|
569 |
-
|
570 |
def is_aggressive_or_harsh(query: str) -> bool:
|
571 |
-
"""
|
572 |
-
Very naive check: If user is insulting AI, complaining about worthless answers, etc.
|
573 |
-
You can refine with better logic or a small LLM classifier.
|
574 |
-
"""
|
575 |
triggers = ["useless", "worthless", "you cannot do anything", "so bad at answering"]
|
576 |
for t in triggers:
|
577 |
if t in query.lower():
|
@@ -579,226 +472,140 @@ def is_aggressive_or_harsh(query: str) -> bool:
|
|
579 |
return False
|
580 |
|
581 |
def is_ethical_conflict(query: str) -> bool:
|
582 |
-
"""
|
583 |
-
Check if user is asking about lying, revenge, or other moral dilemmas.
|
584 |
-
You can expand or refine as needed.
|
585 |
-
"""
|
586 |
ethics_keywords = ["should i lie", "should i cheat", "revenge", "get back at", "hurt them back"]
|
587 |
q_lower = query.lower()
|
588 |
return any(k in q_lower for k in ethics_keywords)
|
589 |
|
590 |
-
|
591 |
-
# -------------------------------------------------------
|
592 |
# Main Pipeline
|
593 |
-
# -------------------------------------------------------
|
594 |
def run_with_chain(query: str) -> str:
|
595 |
-
|
596 |
-
|
597 |
-
|
598 |
-
|
599 |
-
|
600 |
-
|
601 |
-
|
602 |
-
|
603 |
-
|
604 |
-
|
605 |
-
|
606 |
-
|
607 |
-
|
608 |
-
|
609 |
-
if not query or query.strip() == "":
|
610 |
-
return create_error_response("validation", "Empty query.")
|
611 |
-
if len(query.strip()) < 2:
|
612 |
-
return create_error_response("validation", "Too short.")
|
613 |
-
words_in_text = re.findall(r'\b\w+\b', query.lower())
|
614 |
-
if not any(w in english_words for w in words_in_text):
|
615 |
-
return create_error_response("validation", "Unclear words.")
|
616 |
-
if len(query) > 500:
|
617 |
-
return create_error_response("validation", "Too long (>500).")
|
618 |
-
if not handle_rate_limiting(pipeline_state):
|
619 |
-
return create_error_response("rate_limit")
|
620 |
-
# New: Check if the query is a greeting
|
621 |
-
if is_greeting(query):
|
622 |
-
greeting_response = "Hello there!! Welcome to Healthy AI Expert, How may I assist you today?"
|
623 |
-
manage_cache(pipeline_state, query, greeting_response)
|
624 |
-
pipeline_state.update_metrics(start_time)
|
625 |
-
return greeting_response
|
626 |
|
627 |
-
|
628 |
-
|
629 |
-
|
630 |
-
|
631 |
-
|
632 |
-
|
633 |
-
|
634 |
-
|
635 |
-
|
636 |
-
|
637 |
-
|
638 |
-
|
639 |
-
|
640 |
-
|
641 |
-
|
642 |
-
|
643 |
-
|
644 |
-
|
645 |
-
|
646 |
-
manage_cache(pipeline_state, query, final_tailored)
|
647 |
-
pipeline_state.update_metrics(start_time)
|
648 |
-
return final_tailored
|
649 |
-
|
650 |
-
# If hate => refuse
|
651 |
-
if mod_res.categories.get("hate", False):
|
652 |
-
logger.info("Hate content => refusal.")
|
653 |
-
refusal_resp = pipeline_state.refusal_chain.run({"topic": "moderation_flagged"})
|
654 |
-
manage_cache(pipeline_state, query, refusal_resp)
|
655 |
-
pipeline_state.update_metrics(start_time)
|
656 |
-
return refusal_resp
|
657 |
-
|
658 |
-
# If "dangerous" or "violence" is flagged, we might still want to
|
659 |
-
# provide a "non-violent advice" approach (like revenge queries).
|
660 |
-
# So we won't automatically refuse. We'll rely on the
|
661 |
-
# is_ethical_conflict() check below.
|
662 |
-
|
663 |
-
except Exception as e:
|
664 |
-
logger.error(f"Moderation error: {e}")
|
665 |
-
severity = 0.0
|
666 |
-
|
667 |
-
# 3) Check for aggression or ethical conflict
|
668 |
-
if is_aggressive_or_harsh(query):
|
669 |
-
logger.info("Detected harsh/aggressive language => frustration_chain.")
|
670 |
-
frustration_resp = pipeline_state.frustration_chain.run({"query": query})
|
671 |
-
final_tailored = pipeline_state.tailor_chain.run({"response": frustration_resp}).strip()
|
672 |
manage_cache(pipeline_state, query, final_tailored)
|
673 |
pipeline_state.update_metrics(start_time)
|
674 |
return final_tailored
|
675 |
-
|
676 |
-
if
|
677 |
-
logger.info("
|
678 |
-
|
679 |
-
|
680 |
-
manage_cache(pipeline_state, query, final_tailored)
|
681 |
pipeline_state.update_metrics(start_time)
|
682 |
-
return
|
683 |
-
|
684 |
-
|
685 |
-
|
686 |
-
|
687 |
-
|
688 |
-
|
689 |
-
|
690 |
-
|
691 |
-
|
692 |
-
|
693 |
-
|
694 |
-
|
695 |
-
|
696 |
-
|
697 |
-
|
698 |
-
|
699 |
-
|
700 |
-
|
701 |
-
|
702 |
-
|
703 |
-
|
704 |
-
|
705 |
-
|
706 |
-
|
707 |
-
|
708 |
-
|
709 |
-
if
|
710 |
-
|
711 |
-
|
712 |
-
|
713 |
-
|
714 |
-
# rag_chain_fallback = pipeline_state.wellness_rag_chain_fallback
|
715 |
-
|
716 |
-
# RAG with fallback
|
717 |
-
try:
|
718 |
-
try:
|
719 |
-
rag_output = rag_chain_main({"query": query})
|
720 |
-
except Exception as e_main:
|
721 |
-
if "resource exhausted" in str(e_main).lower():
|
722 |
-
logger.warning("Gemini resource exhausted. Falling back to Groq.")
|
723 |
-
# rag_output = rag_chain_fallback({"query": query})
|
724 |
-
else:
|
725 |
-
raise
|
726 |
-
|
727 |
-
if isinstance(rag_output, dict) and "result" in rag_output:
|
728 |
-
csv_ans = rag_output["result"].strip()
|
729 |
-
else:
|
730 |
-
csv_ans = str(rag_output).strip()
|
731 |
-
|
732 |
-
# If not enough => web
|
733 |
-
if "not enough context" in csv_ans.lower() or len(csv_ans) < 40:
|
734 |
-
logger.info("Insufficient RAG => web search.")
|
735 |
-
web_info = do_web_search(query)
|
736 |
-
if web_info:
|
737 |
-
csv_ans += f"\n\nAdditional info:\n{web_info}"
|
738 |
-
except Exception as e:
|
739 |
-
logger.error(f"RAG error: {e}")
|
740 |
-
if not pipeline_state.handle_error(e):
|
741 |
-
return create_error_response("processing", "RAG error.")
|
742 |
-
return create_error_response("processing")
|
743 |
-
|
744 |
-
# Tailor final
|
745 |
try:
|
746 |
-
|
747 |
-
|
748 |
-
|
749 |
-
|
750 |
-
manage_cache(pipeline_state, query, final_tailored)
|
751 |
pipeline_state.update_metrics(start_time)
|
752 |
-
return
|
753 |
except Exception as e:
|
754 |
-
logger.error(f"
|
755 |
if not pipeline_state.handle_error(e):
|
756 |
-
return create_error_response("processing", "
|
757 |
return create_error_response("processing")
|
758 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
759 |
except Exception as e:
|
760 |
-
logger.error(f"
|
761 |
-
pipeline_state.
|
762 |
-
|
763 |
-
|
764 |
-
|
765 |
-
|
766 |
-
|
767 |
-
|
768 |
-
|
769 |
-
# pipeline_state.reset()
|
770 |
-
# return {"status": "success", "message": "Pipeline reset successful"}
|
771 |
-
# except Exception as e:
|
772 |
-
# logger.error(f"Reset pipeline error: {e}")
|
773 |
-
# return {"status": "error", "message": str(e)}
|
774 |
-
|
775 |
-
# def get_pipeline_health() -> Dict[str, Any]:
|
776 |
-
# try:
|
777 |
-
# stats = pipeline_state.get_metrics()
|
778 |
-
# healthy = stats["error_rate"] < 0.1
|
779 |
-
# return {
|
780 |
-
# **stats,
|
781 |
-
# "is_healthy": healthy,
|
782 |
-
# "status": "healthy" if healthy else "degraded"
|
783 |
-
# }
|
784 |
-
# except Exception as e:
|
785 |
-
# logger.error(f"Health check error: {e}")
|
786 |
-
# return {"is_healthy": False, "status": "error", "error": str(e)}
|
787 |
-
|
788 |
-
# def health_check() -> Dict[str, Any]:
|
789 |
-
# try:
|
790 |
-
# _ = run_with_chain("Test query for pipeline health check.")
|
791 |
-
# return {
|
792 |
-
# "status": "ok",
|
793 |
-
# "timestamp": datetime.now().isoformat(),
|
794 |
-
# "metrics": get_pipeline_health()
|
795 |
-
# }
|
796 |
-
# except Exception as e:
|
797 |
-
# return {
|
798 |
-
# "status": "error",
|
799 |
-
# "timestamp": datetime.now().isoformat(),
|
800 |
-
# "error": str(e)
|
801 |
-
# }
|
802 |
-
|
803 |
-
logger.info("Pipeline initialization complete!")
|
804 |
|
|
|
|
10 |
import pandas as pd
|
11 |
from pydantic import BaseModel, Field, ValidationError, validator
|
12 |
|
|
|
13 |
import nltk
|
14 |
from nltk.corpus import words
|
15 |
try:
|
|
|
18 |
nltk.download('words')
|
19 |
english_words = set(words.words())
|
20 |
|
|
|
21 |
from langchain_groq import ChatGroq
|
22 |
from langchain_community.embeddings import HuggingFaceEmbeddings
|
23 |
from langchain_community.vectorstores import FAISS
|
|
|
26 |
from langchain.docstore.document import Document
|
27 |
from langchain_core.caches import BaseCache
|
28 |
from langchain_core.callbacks import Callbacks
|
|
|
|
|
|
|
|
|
29 |
|
|
|
|
|
30 |
from chain.classification_chain import get_classification_chain
|
31 |
from chain.refusal_chain import get_refusal_chain
|
32 |
from chain.tailor_chain import get_tailor_chain
|
33 |
from chain.cleaner_chain import get_cleaner_chain
|
34 |
from chain.tailor_chain_wellnessBrand import get_tailor_chain_wellnessBrand
|
35 |
|
|
|
36 |
from mistralai import Mistral
|
37 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
from smolagents import (
|
39 |
CodeAgent,
|
40 |
DuckDuckGoSearchTool,
|
|
|
43 |
VisitWebpageTool,
|
44 |
)
|
45 |
|
46 |
+
from chain.prompts import selfharm_prompt, frustration_prompt, ethical_conflict_prompt, classification_prompt, refusal_prompt, tailor_prompt, cleaner_prompt
|
|
|
|
|
47 |
|
48 |
logging.basicConfig(level=logging.INFO)
|
49 |
logger = logging.getLogger(__name__)
|
|
|
51 |
from langchain_core.tracers import LangChainTracer
|
52 |
from langsmith import Client
|
53 |
|
54 |
+
os.environ["LANGCHAIN_TRACING_V2"] = "true"
|
55 |
+
os.environ["LANGSMITH_ENDPOINT"] = "https://api.smith.langchain.com"
|
56 |
+
os.environ["LANGCHAIN_API_KEY"] = os.getenv("LANGCHAIN_API_KEY")
|
57 |
+
os.environ["LANGCHAIN_PROJECT"] = os.getenv("LANGCHAIN_PROJECT")
|
58 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
59 |
# Basic Models
|
|
|
60 |
class QueryInput(BaseModel):
|
61 |
query: str = Field(..., min_length=1)
|
62 |
|
|
|
84 |
/ self.total_requests
|
85 |
)
|
86 |
|
|
|
87 |
# Mistral Moderation
|
|
|
88 |
class ModerationResult(BaseModel):
|
89 |
is_safe: bool
|
90 |
categories: Dict[str, bool]
|
|
|
94 |
client = Mistral(api_key=mistral_api_key)
|
95 |
|
96 |
def moderate_text(query: str) -> ModerationResult:
|
97 |
+
"""Moderates text using Mistral to detect unsafe content."""
|
|
|
|
|
98 |
try:
|
99 |
query_input = QueryInput(query=query)
|
100 |
response = client.classifiers.moderate_chat(
|
|
|
126 |
raise RuntimeError(f"Moderation failed: {e}")
|
127 |
|
128 |
def compute_moderation_severity(mresult: ModerationResult) -> float:
|
129 |
+
"""Computes severity score based on moderation flags."""
|
130 |
severity = 0.0
|
131 |
for flag in mresult.categories.values():
|
132 |
if flag:
|
133 |
severity += 0.3
|
134 |
return min(severity, 1.0)
|
135 |
|
|
|
136 |
# Models
|
|
|
137 |
GROQ_MODELS = {
|
138 |
+
"default": "llama3-70b-8192",
|
139 |
"classification": "qwen-qwq-32b",
|
140 |
+
"moderation": "mistral-moderation-latest",
|
141 |
+
"combination": "llama-3.3-70b-versatile"
|
142 |
}
|
143 |
|
144 |
MAX_RETRIES = 3
|
145 |
RATE_LIMIT_REQUESTS = 60
|
146 |
CACHE_SIZE_LIMIT = 1000
|
147 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
148 |
class NoCache(BaseCache):
|
149 |
+
"""No-op cache implementation for ChatGroq."""
|
150 |
def __init__(self):
|
151 |
pass
|
152 |
|
|
|
159 |
def clear(self):
|
160 |
pass
|
161 |
|
|
|
162 |
ChatGroq.model_rebuild()
|
163 |
+
|
164 |
try:
|
165 |
fallback_groq_api_key = os.environ.get("GROQ_API_KEY_FALLBACK", os.environ.get("GROQ_API_KEY"))
|
166 |
if not fallback_groq_api_key:
|
167 |
logger.warning("No Groq API key found for fallback LLM")
|
168 |
groq_fallback_llm = ChatGroq(
|
169 |
+
model=GROQ_MODELS["default"],
|
170 |
temperature=0.7,
|
171 |
groq_api_key=fallback_groq_api_key,
|
172 |
max_tokens=2048,
|
173 |
+
cache=NoCache(),
|
174 |
+
callbacks=[]
|
175 |
)
|
176 |
except Exception as e:
|
177 |
logger.error(f"Failed to initialize fallback Groq LLM: {e}")
|
178 |
raise RuntimeError("ChatGroq initialization failed.") from e
|
179 |
+
|
180 |
# Rate-limit & Cache
|
|
|
181 |
def handle_rate_limiting(state: "PipelineState") -> bool:
|
182 |
+
"""Enforces rate limiting based on request timestamps."""
|
183 |
current_time = time.time()
|
184 |
one_min_ago = current_time - 60
|
185 |
state.request_timestamps = [t for t in state.request_timestamps if t > one_min_ago]
|
|
|
189 |
return True
|
190 |
|
191 |
def manage_cache(state: "PipelineState", query: str, response: str = None) -> Optional[str]:
|
192 |
+
"""Manages cache for query responses."""
|
193 |
cache_key = query.strip().lower()
|
194 |
if response is None:
|
195 |
return state.cache.get(cache_key)
|
|
|
201 |
return None
|
202 |
|
203 |
def create_error_response(error_type: str, details: str = "") -> str:
|
204 |
+
"""Generates standardized error messages."""
|
205 |
templates = {
|
206 |
"validation": "I couldn't process your query: {details}",
|
207 |
"processing": "I encountered an error while processing: {details}",
|
208 |
"rate_limit": "Too many requests. Please try again soon.",
|
209 |
+
"general": "Apologies, but something went wrong."
|
210 |
}
|
211 |
return templates.get(error_type, templates["general"]).format(details=details)
|
212 |
|
|
|
213 |
# Web Search
|
|
|
214 |
web_search_cache: Dict[str, str] = {}
|
215 |
|
216 |
def store_websearch_result(query: str, result: str):
|
|
|
220 |
return web_search_cache.get(query.strip().lower())
|
221 |
|
222 |
def do_web_search(query: str) -> str:
|
223 |
+
"""Performs web search if no cached result exists."""
|
224 |
try:
|
225 |
cached = retrieve_websearch_result(query)
|
226 |
if cached:
|
|
|
228 |
return cached
|
229 |
|
230 |
logger.info("Performing a new web search for: '%s'", query)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
231 |
search_agent = ToolCallingAgent(
|
232 |
+
tools=[DuckDuckGoSearchTool(), VisitWebpageTool()],
|
233 |
+
model=HfApiModel(),
|
234 |
+
name="search_agent",
|
235 |
+
description="This is an agent that can do web search.",
|
236 |
)
|
237 |
|
238 |
manager_agent = CodeAgent(
|
239 |
tools=[],
|
240 |
+
model=HfApiModel(),
|
241 |
+
managed_agents=[search_agent]
|
242 |
)
|
243 |
|
244 |
new_search_result = manager_agent.run(f"Search for information about: {query}")
|
|
|
249 |
return ""
|
250 |
|
251 |
def is_greeting(query: str) -> bool:
|
252 |
+
"""Detects if the query is a greeting."""
|
|
|
|
|
|
|
|
|
|
|
253 |
greetings = {"hello", "hi", "hey", "hii", "hola", "greetings"}
|
|
|
|
|
254 |
cleaned = re.sub(r'[^\w\s]', '', query).strip().lower()
|
|
|
|
|
255 |
words_in_query = set(cleaned.split())
|
|
|
|
|
256 |
return not words_in_query.isdisjoint(greetings)
|
257 |
|
|
|
|
|
258 |
# Vector Stores & RAG
|
|
|
259 |
def build_or_load_vectorstore(csv_path: str, store_dir: str) -> FAISS:
|
260 |
+
"""Builds or loads FAISS vector store from CSV data."""
|
261 |
if os.path.exists(store_dir):
|
262 |
logger.info(f"Loading existing FAISS store from {store_dir}")
|
263 |
embeddings = HuggingFaceEmbeddings(
|
264 |
model_name="sentence-transformers/multi-qa-mpnet-base-dot-v1"
|
265 |
)
|
266 |
+
return FAISS.load_local(store_dir, embeddings, allow_dangerous_deserialization=True)
|
267 |
else:
|
268 |
logger.info(f"Building new FAISS store from {csv_path}")
|
269 |
df = pd.read_csv(csv_path)
|
|
|
290 |
vectorstore = FAISS.from_documents(docs, embedding=embeddings)
|
291 |
vectorstore.save_local(store_dir)
|
292 |
return vectorstore
|
293 |
+
|
294 |
def build_rag_chain(vectorstore: FAISS, llm) -> RetrievalQA:
|
295 |
+
"""Builds RAG chain for wellness queries."""
|
296 |
prompt = PromptTemplate(
|
297 |
template="""
|
298 |
[INST] You are an AI wellness assistant speaking directly to a user who has asked: "{question}"
|
|
|
326 |
}
|
327 |
)
|
328 |
return chain
|
329 |
+
|
330 |
def build_rag_chain2(vectorstore: FAISS, llm) -> RetrievalQA:
|
331 |
+
"""Builds RAG chain for brand strategy queries."""
|
332 |
prompt = PromptTemplate(
|
333 |
template="""
|
334 |
[INST] You are the brand strategy advisor for Healthy AI Expert. A team member has asked: "{question}"
|
|
|
344 |
|
345 |
Remember our key brand pillars: AI-driven personalization, scientific credibility, user-centric design, and innovation leadership.
|
346 |
[/INST]
|
|
|
347 |
""",
|
348 |
input_variables=["context", "question"]
|
349 |
)
|
|
|
362 |
)
|
363 |
return chain
|
364 |
|
|
|
365 |
# PipelineState
|
|
|
366 |
class PipelineState:
|
367 |
_instance = None
|
368 |
|
|
|
378 |
self._initialize()
|
379 |
|
380 |
def _initialize(self):
|
381 |
+
"""Initializes pipeline state and chains."""
|
382 |
try:
|
383 |
self.metrics = ProcessingMetrics()
|
384 |
self.error_count = 0
|
|
|
395 |
raise RuntimeError("Pipeline initialization failed.") from e
|
396 |
|
397 |
def _setup_chains(self):
|
398 |
+
"""Sets up all processing chains and vector stores."""
|
399 |
self.tailor_chainWellnessBrand = get_tailor_chain_wellnessBrand()
|
400 |
self.classification_chain = get_classification_chain()
|
401 |
+
self.refusal_chain = get_refusal_chain()
|
402 |
+
self.tailor_chain = get_tailor_chain()
|
403 |
+
self.cleaner_chain = get_cleaner_chain()
|
404 |
|
|
|
|
|
|
|
|
|
405 |
self.self_harm_chain = LLMChain(llm=groq_fallback_llm, prompt=selfharm_prompt, verbose=False)
|
|
|
|
|
|
|
|
|
|
|
406 |
self.frustration_chain = LLMChain(llm=groq_fallback_llm, prompt=frustration_prompt, verbose=False)
|
|
|
|
|
|
|
|
|
|
|
407 |
self.ethical_conflict_chain = LLMChain(llm=groq_fallback_llm, prompt=ethical_conflict_prompt, verbose=False)
|
408 |
|
409 |
+
brand_csv = "dataset/BrandAI.csv"
|
410 |
+
brand_store = "faiss_brand_store"
|
|
|
411 |
wellness_csv = "dataset/AIChatbot.csv"
|
412 |
wellness_store = "faiss_wellness_store"
|
413 |
|
414 |
+
brand_vs = build_or_load_vectorstore(brand_csv, brand_store)
|
415 |
wellness_vs = build_or_load_vectorstore(wellness_csv, wellness_store)
|
416 |
|
|
|
|
|
417 |
self.groq_fallback_llm = groq_fallback_llm
|
418 |
+
self.brand_rag_chain = build_rag_chain2(brand_vs, self.groq_fallback_llm)
|
|
|
|
|
|
|
419 |
self.wellness_rag_chain = build_rag_chain(wellness_vs, self.groq_fallback_llm)
|
|
|
|
|
420 |
|
421 |
def handle_error(self, error: Exception) -> bool:
|
422 |
+
"""Handles errors and triggers reset if needed."""
|
423 |
self.error_count += 1
|
424 |
self.metrics.errors += 1
|
425 |
if self.error_count >= MAX_RETRIES:
|
|
|
429 |
return True
|
430 |
|
431 |
def reset(self):
|
432 |
+
"""Resets pipeline state while preserving metrics."""
|
433 |
try:
|
434 |
logger.info("Resetting pipeline state.")
|
435 |
old_metrics = self.metrics
|
|
|
445 |
raise RuntimeError("Failed to reset pipeline.")
|
446 |
|
447 |
def get_metrics(self) -> Dict[str, Any]:
|
448 |
+
"""Returns pipeline performance metrics."""
|
449 |
uptime = (datetime.now() - self.metrics.last_reset).total_seconds() / 3600
|
450 |
return {
|
451 |
"total_requests": self.metrics.total_requests,
|
|
|
456 |
}
|
457 |
|
458 |
def update_metrics(self, start_time: float, is_cache_hit: bool = False):
|
459 |
+
"""Updates processing metrics."""
|
460 |
duration = time.time() - start_time
|
461 |
self.metrics.update_metrics(duration, is_cache_hit)
|
462 |
|
463 |
pipeline_state = PipelineState()
|
464 |
|
465 |
+
# Helper Checks
|
|
|
|
|
|
|
466 |
def is_aggressive_or_harsh(query: str) -> bool:
|
467 |
+
"""Detects aggressive or harsh language in query."""
|
|
|
|
|
|
|
468 |
triggers = ["useless", "worthless", "you cannot do anything", "so bad at answering"]
|
469 |
for t in triggers:
|
470 |
if t in query.lower():
|
|
|
472 |
return False
|
473 |
|
474 |
def is_ethical_conflict(query: str) -> bool:
|
475 |
+
"""Detects ethical dilemmas in query."""
|
|
|
|
|
|
|
476 |
ethics_keywords = ["should i lie", "should i cheat", "revenge", "get back at", "hurt them back"]
|
477 |
q_lower = query.lower()
|
478 |
return any(k in q_lower for k in ethics_keywords)
|
479 |
|
|
|
|
|
480 |
# Main Pipeline
|
|
|
481 |
def run_with_chain(query: str) -> str:
|
482 |
+
"""Processes query through validation, moderation, and chains."""
|
483 |
+
start_time = time.time()
|
484 |
+
try:
|
485 |
+
if not query or query.strip() == "":
|
486 |
+
return create_error_response("validation", "Empty query.")
|
487 |
+
if len(query.strip()) < 2:
|
488 |
+
return create_error_response("validation", "Too short.")
|
489 |
+
words_in_text = re.findall(r'\b\w+\b', query.lower())
|
490 |
+
if not any(w in english_words for w in words_in_text):
|
491 |
+
return create_error_response("validation", "Unclear words.")
|
492 |
+
if len(query) > 500:
|
493 |
+
return create_error_response("validation", "Too long (>500).")
|
494 |
+
if not handle_rate_limiting(pipeline_state):
|
495 |
+
return create_error_response("rate_limit")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
496 |
|
497 |
+
if is_greeting(query):
|
498 |
+
greeting_response = "Hello there!! Welcome to Healthy AI Expert, How may I assist you today?"
|
499 |
+
manage_cache(pipeline_state, query, greeting_response)
|
500 |
+
pipeline_state.update_metrics(start_time)
|
501 |
+
return greeting_response
|
502 |
+
|
503 |
+
cached = manage_cache(pipeline_state, query)
|
504 |
+
if cached:
|
505 |
+
pipeline_state.update_metrics(start_time, is_cache_hit=True)
|
506 |
+
return cached
|
507 |
+
|
508 |
+
try:
|
509 |
+
mod_res = moderate_text(query)
|
510 |
+
severity = compute_moderation_severity(mod_res)
|
511 |
+
|
512 |
+
if mod_res.categories.get("selfharm", False):
|
513 |
+
logger.info("Self-harm flagged => providing supportive chain response.")
|
514 |
+
selfharm_resp = pipeline_state.self_harm_chain.run({"query": query})
|
515 |
+
final_tailored = pipeline_state.tailor_chain.run({"response": selfharm_resp}).strip()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
516 |
manage_cache(pipeline_state, query, final_tailored)
|
517 |
pipeline_state.update_metrics(start_time)
|
518 |
return final_tailored
|
519 |
+
|
520 |
+
if mod_res.categories.get("hate", False):
|
521 |
+
logger.info("Hate content => refusal.")
|
522 |
+
refusal_resp = pipeline_state.refusal_chain.run({"topic": "moderation_flagged"})
|
523 |
+
manage_cache(pipeline_state, query, refusal_resp)
|
|
|
524 |
pipeline_state.update_metrics(start_time)
|
525 |
+
return refusal_resp
|
526 |
+
|
527 |
+
except Exception as e:
|
528 |
+
logger.error(f"Moderation error: {e}")
|
529 |
+
severity = 0.0
|
530 |
+
|
531 |
+
if is_aggressive_or_harsh(query):
|
532 |
+
logger.info("Detected harsh/aggressive language => frustration_chain.")
|
533 |
+
frustration_resp = pipeline_state.frustration_chain.run({"query": query})
|
534 |
+
final_tailored = pipeline_state.tailor_chain.run({"response": frustration_resp}).strip()
|
535 |
+
manage_cache(pipeline_state, query, final_tailored)
|
536 |
+
pipeline_state.update_metrics(start_time)
|
537 |
+
return final_tailored
|
538 |
+
|
539 |
+
if is_ethical_conflict(query):
|
540 |
+
logger.info("Detected ethical dilemma => ethical_conflict_chain.")
|
541 |
+
ethical_resp = pipeline_state.ethical_conflict_chain.run({"query": query})
|
542 |
+
final_tailored = pipeline_state.tailor_chain.run({"response": ethical_resp}).strip()
|
543 |
+
manage_cache(pipeline_state, query, final_tailored)
|
544 |
+
pipeline_state.update_metrics(start_time)
|
545 |
+
return final_tailored
|
546 |
+
|
547 |
+
try:
|
548 |
+
class_out = pipeline_state.classification_chain.run({"query": query})
|
549 |
+
classification = class_out.strip().lower()
|
550 |
+
except Exception as e:
|
551 |
+
logger.error(f"Classification error: {e}")
|
552 |
+
if not pipeline_state.handle_error(e):
|
553 |
+
return create_error_response("processing", "Classification error.")
|
554 |
+
return create_error_response("processing")
|
555 |
+
|
556 |
+
if classification in ["outofscope", "out_of_scope"]:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
557 |
try:
|
558 |
+
refusal_text = pipeline_state.refusal_chain.run({"topic": query})
|
559 |
+
tailored_refusal = pipeline_state.tailor_chain.run({"response": refusal_text}).strip()
|
560 |
+
manage_cache(pipeline_state, query, tailored_refusal)
|
|
|
|
|
561 |
pipeline_state.update_metrics(start_time)
|
562 |
+
return tailored_refusal
|
563 |
except Exception as e:
|
564 |
+
logger.error(f"Refusal chain error: {e}")
|
565 |
if not pipeline_state.handle_error(e):
|
566 |
+
return create_error_response("processing", "Refusal error.")
|
567 |
return create_error_response("processing")
|
568 |
+
|
569 |
+
if classification == "brand":
|
570 |
+
rag_chain_main = pipeline_state.brand_rag_chain
|
571 |
+
else:
|
572 |
+
rag_chain_main = pipeline_state.wellness_rag_chain
|
573 |
+
|
574 |
+
try:
|
575 |
+
rag_output = rag_chain_main({"query": query})
|
576 |
+
if isinstance(rag_output, dict) and "result" in rag_output:
|
577 |
+
csv_ans = rag_output["result"].strip()
|
578 |
+
else:
|
579 |
+
csv_ans = str(rag_output).strip()
|
580 |
+
|
581 |
+
if "not enough context" in csv_ans.lower() or len(csv_ans) < 40:
|
582 |
+
logger.info("Insufficient RAG => web search.")
|
583 |
+
web_info = do_web_search(query)
|
584 |
+
if web_info:
|
585 |
+
csv_ans += f"\n\nAdditional info:\n{web_info}"
|
586 |
+
except Exception as e:
|
587 |
+
logger.error(f"RAG error: {e}")
|
588 |
+
if not pipeline_state.handle_error(e):
|
589 |
+
return create_error_response("processing", "RAG error.")
|
590 |
+
return create_error_response("processing")
|
591 |
+
|
592 |
+
try:
|
593 |
+
final_tailored = pipeline_state.tailor_chainWellnessBrand.run({"response": csv_ans}).strip()
|
594 |
+
if severity > 0.5:
|
595 |
+
final_tailored += "\n\n(Please note: This may involve sensitive content.)"
|
596 |
+
|
597 |
+
manage_cache(pipeline_state, query, final_tailored)
|
598 |
+
pipeline_state.update_metrics(start_time)
|
599 |
+
return final_tailored
|
600 |
except Exception as e:
|
601 |
+
logger.error(f"Tailor chain error: {e}")
|
602 |
+
if not pipeline_state.handle_error(e):
|
603 |
+
return create_error_response("processing", "Tailoring error.")
|
604 |
+
return create_error_response("processing")
|
605 |
+
|
606 |
+
except Exception as e:
|
607 |
+
logger.error(f"Critical error in run_with_chain: {e}")
|
608 |
+
pipeline_state.metrics.errors += 1
|
609 |
+
return create_error_response("general")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
610 |
|
611 |
+
logger.info("Pipeline initialization complete!")
|