backend/intelligence/silent_detector.py

"""
PIOE Silent Opportunities Detector - Version 2.0

Detects implicit/hidden opportunities that are never announced clearly.
These appear in blog posts, tweets, Discord messages, research updates.

Examples:
- "We're exploring ideas around..."
- "We're looking for collaborators..."
- "If anyone is interested..."
- "We're building something new..."
"""
import re
from typing import Optional


class SilentOpportunityDetector:
    """
    Detects implicit opportunities from content that doesn't
    explicitly announce them as opportunities.
    """
    
    # Patterns for implicit opportunities
    SIGNAL_PATTERNS = {
        # Pre-hiring signals
        "pre_hiring": [
            r"we(?:'re| are) (?:actively )?(?:looking|searching) for",
            r"we need (?:a |someone|people)",
            r"hiring (?:soon|next|this)",
            r"building (?:a |our |the )?team",
            r"if you(?:'re| are) interested in joining",
            r"open roles? (?:coming|soon)",
            r"dm (?:me|us) if (?:you(?:'re| are)|interested)",
            r"reach out if",
        ],
        
        # Pre-grant signals
        "pre_grant": [
            r"(?:we(?:'re| are)|we will be) (?:funding|supporting|backing)",
            r"grants? (?:coming|opening|soon|next)",
            r"ecosystem fund",
            r"builder(?:s)? program",
            r"retroactive (?:funding|rewards)",
            r"announcing.{0,30}funding",
            r"accepting applications",
        ],
        
        # Collaboration signals
        "collaboration": [
            r"looking for (?:collaborators?|partners?|co-founder)",
            r"seeking (?:collaborat|partner)",
            r"open to (?:collaborat|partner|work)",
            r"anyone (?:want|interested).{0,30}(?:build|work|collaborat)",
            r"let(?:'s| us) (?:build|work|create) together",
            r"who wants to",
            r"exploring.{0,30}partnership",
        ],
        
        # Project/research signals
        "research": [
            r"we(?:'re| are) (?:exploring|researching|investigating)",
            r"new (?:research|project|initiative)",
            r"call for (?:papers?|proposals?|abstracts?)",
            r"(?:research|academic) (?:collaboration|partnership)",
            r"phd (?:position|opportunity|student)",
            r"postdoc",
            r"looking for (?:interns?|students?)",
        ],
        
        # Community/ambassador signals
        "ambassador": [
            r"ambassador program",
            r"community (?:lead|manager|role)",
            r"help (?:us )?(?:grow|build|spread)",
            r"join (?:our|the) (?:community|team|movement)",
            r"early (?:adopter|supporter)",
        ],
        
        # Investment/demo signals
        "investment": [
            r"demo day",
            r"pitch (?:competition|event|day)",
            r"investor (?:meeting|demo|call)",
            r"raising (?:a |our )?(?:seed|round|series)",
            r"open to (?:investment|investors)",
        ],
    }
    
    # Strength indicators (modifiers)
    STRENGTH_BOOSTERS = [
        r"immediately",
        r"urgently",
        r"actively",
        r"now",
        r"today",
        r"this week",
        r"asap",
        r"serious",
        r"exciting",
    ]
    
    # Negative patterns (reduce signal)
    NOISE_PATTERNS = [
        r"not (?:looking|hiring|seeking)",
        r"no longer",
        r"was (?:looking|hiring)",
        r"used to",
        r"back in",
        r"years? ago",
        r"hypothetically",
        r"if only",
    ]
    
    def detect(self, text: str, title: str = "") -> dict:
        """
        Analyze text for silent opportunity signals.
        
        Returns:
        - is_silent_opportunity: bool
        - opportunity_type: str (pre_hiring, pre_grant, etc.)
        - signal_strength: float (0.0 to 1.0)
        - detected_patterns: list
        - recommended_category: str
        """
        full_text = f"{title} {text}".lower()
        
        # Check for noise patterns first
        if self._has_noise(full_text):
            return {
                "is_silent_opportunity": False,
                "opportunity_type": None,
                "signal_strength": 0.0,
                "detected_patterns": [],
                "recommended_category": None,
            }
        
        # Detect patterns
        detected = {}
        for opp_type, patterns in self.SIGNAL_PATTERNS.items():
            matches = self._find_matches(full_text, patterns)
            if matches:
                detected[opp_type] = matches
        
        if not detected:
            return {
                "is_silent_opportunity": False,
                "opportunity_type": None,
                "signal_strength": 0.0,
                "detected_patterns": [],
                "recommended_category": None,
            }
        
        # Find primary opportunity type
        primary_type = max(detected, key=lambda k: len(detected[k]))
        
        # Calculate signal strength
        signal_strength = self._calculate_strength(
            full_text, detected, primary_type
        )
        
        # Map to category
        category_map = {
            "pre_hiring": "pre_hiring_signal",
            "pre_grant": "pre_grant_signal",
            "collaboration": "collaboration",
            "research": "research",
            "ambassador": "ambassador",
            "investment": "pitch_event",
        }
        
        return {
            "is_silent_opportunity": True,
            "opportunity_type": primary_type,
            "signal_strength": round(signal_strength, 3),
            "detected_patterns": detected[primary_type],
            "recommended_category": category_map.get(primary_type, "weak_signal"),
        }
    
    def _find_matches(self, text: str, patterns: list) -> list:
        """Find all matching patterns in text."""
        matches = []
        for pattern in patterns:
            if re.search(pattern, text, re.IGNORECASE):
                # Extract the matching context
                match = re.search(pattern, text, re.IGNORECASE)
                if match:
                    # Get surrounding context
                    start = max(0, match.start() - 20)
                    end = min(len(text), match.end() + 20)
                    context = text[start:end]
                    matches.append(context.strip())
        return matches
    
    def _has_noise(self, text: str) -> bool:
        """Check if text contains noise patterns."""
        for pattern in self.NOISE_PATTERNS:
            if re.search(pattern, text, re.IGNORECASE):
                return True
        return False
    
    def _calculate_strength(
        self, 
        text: str, 
        detected: dict, 
        primary_type: str
    ) -> float:
        """Calculate signal strength."""
        base_strength = 0.5
        
        # More patterns = stronger signal
        pattern_count = len(detected[primary_type])
        base_strength += min(pattern_count * 0.1, 0.3)
        
        # Check for strength boosters
        for booster in self.STRENGTH_BOOSTERS:
            if re.search(booster, text, re.IGNORECASE):
                base_strength += 0.05
        
        # Multiple types of signals = stronger
        if len(detected) > 1:
            base_strength += 0.1
        
        # Cap at 1.0
        return min(base_strength, 1.0)
    
    def reclassify_opportunity(
        self, 
        opportunity: dict
    ) -> tuple[str, float]:
        """
        Re-evaluate an existing opportunity for silent signals.
        
        Returns (new_category, confidence)
        """
        title = opportunity.get("title", "")
        text = opportunity.get("raw_text", "")
        
        result = self.detect(text, title)
        
        if result["is_silent_opportunity"]:
            return (
                result["recommended_category"],
                result["signal_strength"]
            )
        
        return (None, 0.0)


class OpportunityLanguageDetector:
    """
    Detects the urgency, timing, and action language in opportunities.
    """
    
    TIMING_PATTERNS = {
        "early": [
            r"early (?:bird|access|application)",
            r"just (?:launched|announced|opened)",
            r"applications? (?:now )?open",
            r"first (?:round|batch|cohort)",
            r"founding",
            r"new program",
        ],
        "optimal": [
            r"applications? (?:open|accepted)",
            r"deadline (?:is )?(?:soon|approaching)",
            r"apply (?:now|today)",
            r"last call",
            r"extended deadline",
        ],
        "late": [
            r"deadline (?:in )?(?:days?|hours?)",
            r"closes? (?:soon|tomorrow|today)",
            r"final (?:day|hour|chance)",
            r"last (?:day|chance)",
        ],
    }
    
    def detect_timing(self, text: str) -> str:
        """Detect application timing."""
        text = text.lower()
        
        for timing, patterns in self.TIMING_PATTERNS.items():
            for pattern in patterns:
                if re.search(pattern, text, re.IGNORECASE):
                    return timing
        
        return "unknown"
    
    def extract_action_items(self, text: str) -> list:
        """Extract actionable items from text."""
        actions = []
        
        # Common action patterns
        action_patterns = [
            r"apply (?:at|via|through|here)",
            r"visit (?:our|the) (?:website|page|link)",
            r"(?:fill|submit).{0,20}(?:form|application)",
            r"send.{0,20}(?:email|resume|cv|portfolio)",
            r"register (?:at|on|here)",
            r"sign up",
            r"join.{0,20}(?:discord|telegram|slack)",
            r"dm (?:me|us)",
            r"follow.{0,10}on",
        ]
        
        for pattern in action_patterns:
            match = re.search(pattern, text, re.IGNORECASE)
            if match:
                start = max(0, match.start() - 10)
                end = min(len(text), match.end() + 30)
                actions.append(text[start:end].strip())
        
        return actions[:5]  # Limit to 5 actions