🧠 Prompt Reasoning Patterns Playground (LAB-05)¶

import os
import sys
import sys
from pathlib import Path

try:
    import google.colab
    IN_COLAB = True
except ImportError:
    IN_COLAB = False

if IN_COLAB:
    print("🤖 Running in Google Colab. Setting up environment...")
    !pip install -q numpy scikit-learn matplotlib loguru
    
    # Clone repository to access custom local source modules
    if not os.path.exists("production-ai-systems"):
        !git clone https://github.com/RogerioLS/production-ai-systems.git
    
    sys.path.append("production-ai-systems")
    print("🎉 Colab environment successfully initialized!")
# Ensure project root is in the path when running locally
if not IN_COLAB:
    # Localiza dinamicamente a raiz do repositório procurando por 'pyproject.toml'
    current_dir = Path.cwd()
    root_dir = current_dir
    for parent in [current_dir] + list(current_dir.parents):
        if (parent / "pyproject.toml").exists() or (parent / "projects").exists():
            root_dir = parent
            break

    if str(root_dir) not in sys.path:
        sys.path.insert(0, str(root_dir))

    print(f"📁 Project root added to sys.path: {root_dir}")

from projects.b_prompt_engineering.src.lab_05_reasoning.templates import PromptTemplate, ChatPromptTemplate
from projects.b_prompt_engineering.src.lab_05_reasoning.types import PromptRole

template = PromptTemplate(
    "Analyze stock {ticker} for fiscal year {year}. Focus on {metric}.",
    role=PromptRole.USER
)

print("Required variables:", template.variables)
formatted = template.format(ticker="PETR4", year="2024", metric="EBITDA Margin")
print("\nFormatted Prompt:")
print(formatted)

from projects.b_prompt_engineering.src.lab_05_reasoning.few_shot import FewShotPromptEngine
from projects.b_prompt_engineering.src.lab_05_reasoning.types import Exemplar

engine = FewShotPromptEngine()
engine.register_exemplar(Exemplar(
    input_text="What is the P/E ratio if Price=100 and EPS=10?",
    output_text="10.0",
    reasoning="P/E = Price / EPS = 100 / 10 = 10.0",
    metadata={"domain": "finance"}
))
engine.register_exemplar(Exemplar(
    input_text="Classify sentiment: 'The quarterly revenue exceeded all guidance.'",
    output_text="Positive",
    reasoning="Words like 'exceeded all guidance' express strong positive performance.",
    metadata={"domain": "nlp"}
))

# Dynamic selector based on keyword
def domain_selector(query, pool, k):
    domain = "finance" if any(w in query.lower() for w in ["ratio", "eps", "price", "ebitda"]) else "nlp"
    return [ex for ex in pool if ex.metadata.get("domain") == domain][:k]

prompt = engine.build_prompt("Calculate P/E with Price=50 and EPS=5.", selector=domain_selector)
print(prompt)

from projects.b_prompt_engineering.src.lab_05_reasoning.chain_of_thought import SelfConsistencyEngine

sc = SelfConsistencyEngine()

# Scenario A: Unanimous agreement (zero entropy)
samples_a = ["$42M", "$42M", "$42M", "$42M", "$42M"]
res_a = sc.evaluate_consensus(samples_a)
print(f"Scenario A -> Winner: {res_a.winning_answer} | Confidence: {res_a.confidence_score*100}% | Entropy: {res_a.shannon_entropy}")

# Scenario B: High uncertainty (divergent paths)
samples_b = ["$42M", "$45M", "$40M", "$42M", "$50M"]
res_b = sc.evaluate_consensus(samples_b)
print(f"Scenario B -> Winner: {res_b.winning_answer} | Confidence: {res_b.confidence_score*100}% | Entropy: {res_b.shannon_entropy}")

from projects.b_prompt_engineering.src.lab_05_reasoning.react import ReActEngine

agent = ReActEngine(max_iterations=5)

# Register financial lookup tool
def financial_database(ticker: str):
    db = {"VALE3": "R$ 280 Billion Market Cap", "ITUB4": "R$ 310 Billion Market Cap"}
    return db.get(ticker.upper(), "Ticker not found")

agent.register_tool(
    name="financial_database",
    description="Returns market cap given a Brazilian stock ticker (e.g. VALE3, ITUB4)",
    handler=financial_database
)

# Mock LLM that reasons and calls the tool
def mock_llm_step(trajectory: str):
    if "Observation: R$ 280 Billion Market Cap" in trajectory:
        return "Thought: I have retrieved the market cap.\nFinal Answer: The market cap of VALE3 is R$ 280 Billion."
    return "Thought: I need to query the database for VALE3.\nAction: financial_database\nAction Input: {\"ticker\": \"VALE3\"}\n"

result = agent.run("What is the market cap of VALE3?", llm_responder=mock_llm_step)
print(f"Success: {result.success}")
print(f"Total Steps: {result.total_steps}")
print(f"Final Answer: {result.final_answer}")
for step in result.steps:
    print(f"  Step {step.step_number}: Tool={step.action_tool} -> Obs={step.observation}")