Spaces:
Sleeping
Sleeping
File size: 18,228 Bytes
a108ce9 | 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 | import os
import json
import time
from typing import Dict, List, Any
from dotenv import load_dotenv
from src.utils.llm import get_fast_llm
load_dotenv()
from src.features.translation import translate_text
from src.features.summarization import summarize_text
from src.features.keyword_extraction import extract_keywords
from src.features.email_writer import write_email, write_cover_letter
from src.graphs.chat_graph import ChatSession
# ββ SAMPLE TEST DATA ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
TRANSLATION_TESTS = [
{
"text": "Artificial intelligence is transforming the world.",
"source": "English", "target": "French",
"reference": "L'intelligence artificielle transforme le monde."
},
{
"text": "Machine learning enables computers to learn from data.",
"source": "English", "target": "Spanish",
"reference": "El aprendizaje automΓ‘tico permite a las computadoras aprender de los datos."
},
{
"text": "Deep learning uses neural networks to solve complex problems.",
"source": "English", "target": "German",
"reference": "Deep Learning verwendet neuronale Netze, um komplexe Probleme zu lΓΆsen."
},
]
SUMMARIZATION_TESTS = [
{"text": """Artificial intelligence (AI) is intelligence demonstrated by machines,
as opposed to the natural intelligence displayed by animals including humans.
AI research has been defined as the field of study of intelligent agents,
which refers to any system that perceives its environment and takes actions
that maximize its chance of achieving its goals. The term "artificial intelligence"
had previously been used to describe machines that mimic and display human cognitive
skills associated with the human mind, such as learning and problem-solving.
This definition has since been rejected by major AI researchers who now describe
AI in terms of rationality and acting rationally, which does not limit how
intelligence can be articulated. AI applications include advanced web search engines,
recommendation systems, understanding human speech, self-driving cars, generative
tools and competing at the highest level in strategic games.""",
"reference": "AI is machine-demonstrated intelligence that perceives environments and takes goal-maximizing actions, with applications ranging from search engines to self-driving cars."
},
{
"text": """LangChain is a framework designed to simplify the creation of applications
using large language models. As a language model integration framework, LangChain's
use-cases largely overlap with those of language models in general, including document
analysis and summarization, chatbots, and code analysis. LangChain was launched in
October 2022 as an open source project. It provides a standard interface for chains,
lots of integrations with other tools, and end-to-end chains for common applications.
The framework allows AI developers to develop applications based on combining LLMs
with other sources of computation or knowledge.""",
"reference": "LangChain is an open-source framework that simplifies building LLM applications through standard interfaces, tool integrations, and pre-built chains."
},
]
KEYWORD_TESTS = [
{
"text": """Natural language processing (NLP) is a subfield of linguistics,
computer science, and artificial intelligence concerned with the interactions
between computers and human language, in particular how to program computers
to process and analyze large amounts of natural language data. The goal is
a computer capable of understanding the contents of documents, including the
contextual nuances of the language within them.""",
"expected_keywords": ["natural language processing", "NLP", "artificial intelligence",
"computer science", "linguistics", "language", "computers"]
},
]
EMAIL_TESTS = [
{
"type": "follow-up",
"context": "Following up on a job application for ML Engineer position submitted last week",
"tone": "professional",
"recipient": "Hiring Manager",
"sender": "Shashank Hegde",
"check_words": ["follow", "application", "ML Engineer", "Shashank"]
},
{
"type": "thank you",
"context": "Thanking the interviewer after a data science interview",
"tone": "professional",
"recipient": "Dr. Smith",
"sender": "Shashank",
"check_words": ["thank", "interview", "opportunity"]
},
]
CHAT_TESTS = [
{"input": "What is machine learning?", "check": ["learning", "data", "model", "algorithm"]},
{"input": "Explain neural networks in simple terms.", "check": ["network", "neuron", "layer", "brain"]},
{"input": "What are the differences between supervised and unsupervised learning?",
"check": ["supervised", "unsupervised", "label", "data"]},
]
COVER_LETTER_TESTS = [
{
"job": "ML Engineer",
"company": "Google DeepMind",
"skills": "Python, PyTorch, LangChain, MLflow",
"experience": "3 years building ML pipelines and deploying models to production",
"name": "Shashank Hegde",
"check_words": ["ML Engineer", "Google", "Python", "Shashank"]
}
]
# ββ UTILITY FUNCTIONS ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
def simple_bleu( hypothesis:str, reference:str) -> float:
"""Simple word-overlap BLEU approximation (no external library needed)."""
hyp_words = set(hypothesis.lower().split())
ref_words = set(reference.lower().split())
if not ref_words:
return 0.0
overlap = hyp_words & ref_words
return len(overlap) / len(ref_words)
def rouge_L(hypothesis:str, reference:str) -> float:
"""Simple ROUGE-L approximation (no external library needed)."""
# This is a simplified version of ROUGE-L, which measures the longest common subsequence
# between the hypothesis and reference.
hyp = list(hypothesis.lower())
ref = list(reference.lower())
if not ref:
return 0.0
common = set(hyp) & set(ref)
precision = len(common) / len(hyp) if hyp else 0
recall = len(common) / len(ref) if ref else 0
if precision + recall == 0:
return 0.0
return 2 * (precision * recall) / (precision + recall)
def keyword_precision(extracted: str, expected: List[str], k: int = 10) -> float:
"""Check how many expected keywords appear in extracted output."""
extracted_lower = extracted.lower()
found = sum(1 for kw in expected if kw.lower() in extracted_lower)
return found / len(expected)
def check_words_present(text: str, check_words: List[str]) -> float:
"""Check what fraction of expected words appear in the output."""
text = text.lower()
found = sum(1 for w in check_words if w.lower() in text)
return found / len(check_words)
def llm_judge(question: str, response: str) -> Dict[str, Any]:
"""Use Groq LLM to evaluate chat response quality."""
from src.utils.llm import get_fast_llm
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import PromptTemplate
prompt = PromptTemplate(
input_variables=["question", "response"],
template="""Evaluate this AI response. Return ONLY a JSON object, no extra text.
Question: {question}
Response: {response}
Return exactly this JSON format:
{{"relevance": <1-10>, "accuracy": <1-10>, "clarity": <1-10>, "overall": <1-10>, "comment": "<one sentence>"}}"""
)
try:
llm = get_fast_llm()
chain = prompt | llm | StrOutputParser()
result = chain.invoke({"question": question, "response": response})
# Extract JSON from response
start = result.find("{")
end = result.rfind("}") + 1
if start != -1 and end > start:
return json.loads(result[start:end])
except Exception as e:
pass
return {"relevance": 0, "accuracy": 0, "clarity": 0, "overall": 0, "comment": "Error occurred while evaluating the response."}
def print_header(title: str):
print(f"\n{'='*60}")
print(f" {title}")
print(f"{'='*60}")
def print_result(label: str, score: float, details: str = ""):
bar = "β" * int(score * 20) + "β" * (20 - int(score * 20))
print(f" {label:<25} [{bar}] {score:.2%} {details}")
# ββ EVALUATION FUNCTIONS βββββββββββββββββββββββββββββββββββββββββββββββββββββ
def evaluate_translation() -> Dict:
print_header("π TRANSLATION EVALUATION")
scores = []
for i, test in enumerate(TRANSLATION_TESTS, 1):
print(f"\n Test {i}: English β {test['target']}")
print(f" Input: {test['text'][:60]}...")
t0 = time.time()
result = translate_text(test["text"], test["source"], test["target"])
latency = time.time() - t0
score = simple_bleu(result, test["reference"])
scores.append(score)
print(f" Output: {result[:80]}...")
print(f" Reference: {test['reference'][:80]}...")
print_result("Word Overlap Score", score, f"| {latency:.2f}s")
avg = sum(scores) / len(scores)
print(f"\n β
Average Score: {avg:.2%}")
return {"feature": "Translation", "avg_score": avg, "scores": scores}
def evaluate_summarization() -> Dict:
print_header("π SUMMARIZATION EVALUATION")
scores = []
for i, test in enumerate(SUMMARIZATION_TESTS, 1):
print(f"\n Test {i}: Document ({len(test['text'].split())} words)")
t0 = time.time()
result = summarize_text(test["text"], "concise")
latency = time.time() - t0
score = rouge_L(result, test["reference"])
scores.append(score)
print(f" Summary: {result[:100]}...")
print(f" Reference: {test['reference'][:100]}...")
print_result("ROUGE-L Score", score, f"| {latency:.2f}s")
avg = sum(scores) / len(scores)
print(f"\n β
Average Score: {avg:.2%}")
return {"feature": "Summarization", "avg_score": avg, "scores": scores}
def evaluate_keywords() -> Dict:
print_header("π KEYWORD EXTRACTION EVALUATION")
scores = []
for i, test in enumerate(KEYWORD_TESTS, 1):
print(f"\n Test {i}: ({len(test['text'].split())} words)")
t0 = time.time()
result = extract_keywords(test["text"], num_keywords=10)
latency = time.time() - t0
score = keyword_precision(result, test["expected_keywords"])
scores.append(score)
print(f" Output (first 150 chars): {result[:150]}...")
print_result("Keyword Precision", score, f"| {latency:.2f}s")
avg = sum(scores) / len(scores)
print(f"\n β
Average Score: {avg:.2%}")
return {"feature": "Keywords", "avg_score": avg, "scores": scores}
def evaluate_email_writer() -> Dict:
print_header("βοΈ EMAIL WRITER EVALUATION")
scores = []
# Email tests
print("\n π§ Email Generation:")
for i, test in enumerate(EMAIL_TESTS, 1):
t0 = time.time()
result = write_email(
test["type"], test["context"],
test["tone"], test["recipient"], test["sender"]
)
latency = time.time() - t0
score = check_words_present(result, test["check_words"])
scores.append(score)
print(f"\n Test {i} ({test['type']}):")
print(f" Output: {result[:120]}...")
print_result("Content Check", score, f"| {latency:.2f}s")
# Cover letter test
print("\n π Cover Letter Generation:")
for test in COVER_LETTER_TESTS:
t0 = time.time()
result = write_cover_letter(
test["job"], test["company"],
test["skills"], test["experience"], test["name"]
)
latency = time.time() - t0
score = check_words_present(result, test["check_words"])
scores.append(score)
print(f" Output: {result[:120]}...")
print_result("Content Check", score, f"| {latency:.2f}s")
avg = sum(scores) / len(scores)
print(f"\n β
Average Score: {avg:.2%}")
return {"feature": "Email Writer", "avg_score": avg, "scores": scores}
def evaluate_chat() -> Dict:
print_header("π¬ CHAT ASSISTANT EVALUATION")
session = ChatSession()
scores = []
judge_scores = []
for i, test in enumerate(CHAT_TESTS, 1):
print(f"\n Test {i}: {test['input']}")
t0 = time.time()
response = session.chat(test["input"])
latency = time.time() - t0
# Word check score
word_score = check_words_present(response, test["check"])
scores.append(word_score)
# LLM Judge score
print(f" Response: {response[:120]}...")
print_result("Word Check Score", word_score, f"| {latency:.2f}s")
print(f" π€ Running LLM Judge...")
judge = llm_judge(test["input"], response)
overall = judge.get("overall", 0) / 10
judge_scores.append(overall)
print(f" LLM Judge β Relevance: {judge.get('relevance')}/10 | "
f"Accuracy: {judge.get('accuracy')}/10 | "
f"Clarity: {judge.get('clarity')}/10 | "
f"Overall: {judge.get('overall')}/10")
print(f" Comment: {judge.get('comment', '')}")
avg_word = sum(scores) / len(scores)
avg_judge = sum(judge_scores) / len(judge_scores) if judge_scores else 0
print(f"\n β
Average Word Check: {avg_word:.2%}")
print(f" β
Average LLM Judge: {avg_judge:.2%}")
return {
"feature": "Chat",
"avg_word_score": avg_word,
"avg_judge_score": avg_judge,
"scores": scores
}
def evaluate_pdf_qa() -> Dict:
"""PDF Q&A evaluation using a sample generated PDF."""
print_header("π PDF Q&A EVALUATION")
# Create a sample text file to simulate PDF content
sample_text = """
LangGraph is a library for building stateful, multi-actor applications with LLMs.
It extends LangChain by adding support for cyclic graphs, which are essential for
creating agent runtimes. LangGraph allows developers to define flows that involve
cycles, enabling more complex agent behaviors.
Key features of LangGraph include:
1. Stateful graphs with persistent memory
2. Support for multi-agent architectures
3. Built-in streaming support
4. Human-in-the-loop capabilities
5. Time travel and debugging features
LangGraph is particularly useful for building chatbots, agents, and other
applications that require maintaining state across multiple interactions.
"""
qa_tests = [
{"question": "What is LangGraph?", "expected_words": ["library", "stateful", "LLM", "graph"]},
{"question": "What are the key features of LangGraph?", "expected_words": ["stateful", "streaming", "memory", "agent"]},
{"question": "What is LangGraph useful for?", "expected_words": ["chatbot", "agent", "state"]},
]
scores = []
try:
from src.features.pdf_qa import build_pdf_qa_graph
for i, test in enumerate(qa_tests, 1):
print(f"\n Test {i}: {test['question']}")
t0 = time.time()
graph = build_pdf_qa_graph()
result = graph.invoke({
"pdf_text": sample_text,
"question": test["question"],
"context": "",
"answer": ""
})
latency = time.time() - t0
answer = result.get("answer", "")
score = check_words_present(answer, test["expected_words"])
scores.append(score)
print(f" Answer: {answer[:120]}...")
print_result("Relevance Score", score, f"| {latency:.2f}s")
except Exception as e:
print(f" β οΈ PDF Q&A evaluation skipped: {e}")
return {"feature": "PDF Q&A", "avg_score": 0, "scores": []}
avg = sum(scores) / len(scores)
print(f"\n β
Average Score: {avg:.2%}")
return {"feature": "PDF Q&A", "avg_score": avg, "scores": scores}
# ββ MAIN RUNNER ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
def run_all_evaluations():
print("\n" + "π¬ " * 20)
print(" GENAI TOOLKIT β FULL EVALUATION REPORT")
print("π¬ " * 20)
results = []
start_time = time.time()
# Run all evaluations
results.append(evaluate_translation())
results.append(evaluate_summarization())
results.append(evaluate_keywords())
results.append(evaluate_email_writer())
results.append(evaluate_chat())
results.append(evaluate_pdf_qa())
total_time = time.time() - start_time
# Final summary
print_header("π FINAL SUMMARY")
print()
overall_scores = []
for r in results:
score = r.get("avg_score") or r.get("avg_judge_score", 0)
overall_scores.append(score)
print_result(r["feature"], score)
overall = sum(overall_scores) / len(overall_scores)
print(f"\n {'β'*55}")
print_result("OVERALL TOOLKIT SCORE", overall)
print(f"\n β±οΈ Total evaluation time: {total_time:.1f}s")
# Save results to JSON
output = {
"overall_score": overall,
"total_time_seconds": total_time,
"features": results
}
os.makedirs("artifacts", exist_ok=True)
with open("artifacts/evaluation_results.json", "w") as f:
json.dump(output, f, indent=2)
print(f"\n πΎ Results saved to artifacts/evaluation_results.json")
print("\n" + "β
" * 20 + "\n")
return output
if __name__ == "__main__":
run_all_evaluations()
|