1#!/usr/bin/env python3
2# -*- coding: utf-8 -*-
3"""
4Slide Search Core - BM25 search engine for slide design databases
5"""
6 
7import csv
8import re
9from pathlib import Path
10from math import log
11from collections import defaultdict
12 
13# ============ CONFIGURATION ============
14DATA_DIR = Path(__file__).parent.parent / "data"
15MAX_RESULTS = 3
16 
17CSV_CONFIG = {
18    "strategy": {
19        "file": "slide-strategies.csv",
20        "search_cols": ["strategy_name", "keywords", "goal", "audience", "narrative_arc"],
21        "output_cols": ["strategy_name", "keywords", "slide_count", "structure", "goal", "audience", "tone", "narrative_arc", "sources"]
22    },
23    "layout": {
24        "file": "slide-layouts.csv",
25        "search_cols": ["layout_name", "keywords", "use_case", "recommended_for"],
26        "output_cols": ["layout_name", "keywords", "use_case", "content_zones", "visual_weight", "cta_placement", "recommended_for", "avoid_for", "css_structure"]
27    },
28    "copy": {
29        "file": "slide-copy.csv",
30        "search_cols": ["formula_name", "keywords", "use_case", "emotion_trigger", "slide_type"],
31        "output_cols": ["formula_name", "keywords", "components", "use_case", "example_template", "emotion_trigger", "slide_type", "source"]
32    },
33    "chart": {
34        "file": "slide-charts.csv",
35        "search_cols": ["chart_type", "keywords", "best_for", "when_to_use", "slide_context"],
36        "output_cols": ["chart_type", "keywords", "best_for", "data_type", "when_to_use", "when_to_avoid", "max_categories", "slide_context", "css_implementation", "accessibility_notes"]
37    }
38}
39 
40AVAILABLE_DOMAINS = list(CSV_CONFIG.keys())
41 
42 
43# ============ BM25 IMPLEMENTATION ============
44class BM25:
45    """BM25 ranking algorithm for text search"""
46 
47    def __init__(self, k1=1.5, b=0.75):
48        self.k1 = k1
49        self.b = b
50        self.corpus = []
51        self.doc_lengths = []
52        self.avgdl = 0
53        self.idf = {}
54        self.doc_freqs = defaultdict(int)
55        self.N = 0
56 
57    def tokenize(self, text):
58        """Lowercase, split, remove punctuation, filter short words"""
59        text = re.sub(r'[^\w\s]', ' ', str(text).lower())
60        return [w for w in text.split() if len(w) > 2]
61 
62    def fit(self, documents):
63        """Build BM25 index from documents"""
64        self.corpus = [self.tokenize(doc) for doc in documents]
65        self.N = len(self.corpus)
66        if self.N == 0:
67            return
68        self.doc_lengths = [len(doc) for doc in self.corpus]
69        self.avgdl = sum(self.doc_lengths) / self.N
70 
71        for doc in self.corpus:
72            seen = set()
73            for word in doc:
74                if word not in seen:
75                    self.doc_freqs[word] += 1
76                    seen.add(word)
77 
78        for word, freq in self.doc_freqs.items():
79            self.idf[word] = log((self.N - freq + 0.5) / (freq + 0.5) + 1)
80 
81    def score(self, query):
82        """Score all documents against query"""
83        query_tokens = self.tokenize(query)
84        scores = []
85 
86        for idx, doc in enumerate(self.corpus):
87            score = 0
88            doc_len = self.doc_lengths[idx]
89            term_freqs = defaultdict(int)
90            for word in doc:
91                term_freqs[word] += 1
92 
93            for token in query_tokens:
94                if token in self.idf:
95                    tf = term_freqs[token]
96                    idf = self.idf[token]
97                    numerator = tf * (self.k1 + 1)
98                    denominator = tf + self.k1 * (1 - self.b + self.b * doc_len / self.avgdl)
99                    score += idf * numerator / denominator
100 
101            scores.append((idx, score))
102 
103        return sorted(scores, key=lambda x: x[1], reverse=True)
104 
105 
106# ============ SEARCH FUNCTIONS ============
107def _load_csv(filepath):
108    """Load CSV and return list of dicts"""
109    with open(filepath, 'r', encoding='utf-8') as f:
110        return list(csv.DictReader(f))
111 
112 
113def _search_csv(filepath, search_cols, output_cols, query, max_results):
114    """Core search function using BM25"""
115    if not filepath.exists():
116        return []
117 
118    data = _load_csv(filepath)
119 
120    # Build documents from search columns
121    documents = [" ".join(str(row.get(col, "")) for col in search_cols) for row in data]
122 
123    # BM25 search
124    bm25 = BM25()
125    bm25.fit(documents)
126    ranked = bm25.score(query)
127 
128    # Get top results with score > 0
129    results = []
130    for idx, score in ranked[:max_results]:
131        if score > 0:
132            row = data[idx]
133            results.append({col: row.get(col, "") for col in output_cols if col in row})
134 
135    return results
136 
137 
138def detect_domain(query):
139    """Auto-detect the most relevant domain from query"""
140    query_lower = query.lower()
141 
142    domain_keywords = {
143        "strategy": ["pitch", "deck", "investor", "yc", "seed", "series", "demo", "sales", "webinar",
144                     "conference", "board", "qbr", "all-hands", "duarte", "kawasaki", "structure"],
145        "layout": ["slide", "layout", "grid", "column", "title", "hero", "section", "cta",
146                   "screenshot", "quote", "timeline", "comparison", "pricing", "team"],
147        "copy": ["headline", "copy", "formula", "aida", "pas", "hook", "cta", "benefit",
148                 "objection", "proof", "testimonial", "urgency", "scarcity"],
149        "chart": ["chart", "graph", "bar", "line", "pie", "funnel", "metrics", "data",
150                  "visualization", "kpi", "trend", "comparison", "heatmap", "gauge"]
151    }
152 
153    scores = {domain: sum(1 for kw in keywords if kw in query_lower) for domain, keywords in domain_keywords.items()}
154    best = max(scores, key=scores.get)
155    return best if scores[best] > 0 else "strategy"
156 
157 
158def search(query, domain=None, max_results=MAX_RESULTS):
159    """Main search function with auto-domain detection"""
160    if domain is None:
161        domain = detect_domain(query)
162 
163    config = CSV_CONFIG.get(domain, CSV_CONFIG["strategy"])
164    filepath = DATA_DIR / config["file"]
165 
166    if not filepath.exists():
167        return {"error": f"File not found: {filepath}", "domain": domain}
168 
169    results = _search_csv(filepath, config["search_cols"], config["output_cols"], query, max_results)
170 
171    return {
172        "domain": domain,
173        "query": query,
174        "file": config["file"],
175        "count": len(results),
176        "results": results
177    }
178 
179 
180def search_all(query, max_results=2):
181    """Search across all domains for comprehensive results"""
182    all_results = {}
183 
184    for domain in AVAILABLE_DOMAINS:
185        result = search(query, domain, max_results)
186        if result.get("count", 0) > 0:
187            all_results[domain] = result
188 
189    return all_results
190 
191 
192# ============ CONTEXTUAL SEARCH (Premium Slide System) ============
193 
194# New CSV configurations for decision system
195DECISION_CSV_CONFIG = {
196    "layout-logic": {
197        "file": "slide-layout-logic.csv",
198        "key_col": "goal"
199    },
200    "typography": {
201        "file": "slide-typography.csv",
202        "key_col": "content_type"
203    },
204    "color-logic": {
205        "file": "slide-color-logic.csv",
206        "key_col": "emotion"
207    },
208    "backgrounds": {
209        "file": "slide-backgrounds.csv",
210        "key_col": "slide_type"
211    }
212}
213 
214 
215def _load_decision_csv(csv_type):
216    """Load a decision CSV and return as dict keyed by primary column."""
217    config = DECISION_CSV_CONFIG.get(csv_type)
218    if not config:
219        return {}
220 
221    filepath = DATA_DIR / config["file"]
222    if not filepath.exists():
223        return {}
224 
225    data = _load_csv(filepath)
226    return {row[config["key_col"]]: row for row in data if config["key_col"] in row}
227 
228 
229def get_layout_for_goal(goal, previous_emotion=None):
230    """
231    Get layout recommendation based on slide goal.
232    Uses slide-layout-logic.csv for decision.
233    """
234    layouts = _load_decision_csv("layout-logic")
235    row = layouts.get(goal, layouts.get("features", {}))
236 
237    result = dict(row) if row else {}
238 
239    # Apply pattern-breaking logic
240    if result.get("break_pattern") == "true" and previous_emotion:
241        result["_pattern_break"] = True
242        result["_contrast_with"] = previous_emotion
243 
244    return result
245 
246 
247def get_typography_for_slide(slide_type, has_metrics=False, has_quote=False):
248    """
249    Get typography recommendation based on slide content.
250    Uses slide-typography.csv for decision.
251    """
252    typography = _load_decision_csv("typography")
253 
254    if has_metrics:
255        return typography.get("metric-callout", {})
256    if has_quote:
257        return typography.get("quote-block", {})
258 
259    # Map slide types to typography
260    type_map = {
261        "hero": "hero-statement",
262        "hook": "hero-statement",
263        "title": "title-only",
264        "problem": "subtitle-heavy",
265        "agitation": "metric-callout",
266        "solution": "subtitle-heavy",
267        "features": "feature-grid",
268        "proof": "metric-callout",
269        "traction": "data-insight",
270        "social": "quote-block",
271        "testimonial": "testimonial",
272        "pricing": "pricing",
273        "team": "team",
274        "cta": "cta-action",
275        "comparison": "comparison",
276        "timeline": "timeline",
277    }
278 
279    content_type = type_map.get(slide_type, "feature-grid")
280    return typography.get(content_type, {})
281 
282 
283def get_color_for_emotion(emotion):
284    """
285    Get color treatment based on emotional beat.
286    Uses slide-color-logic.csv for decision.
287    """
288    colors = _load_decision_csv("color-logic")
289    return colors.get(emotion, colors.get("clarity", {}))
290 
291 
292def get_background_config(slide_type):
293    """
294    Get background image configuration.
295    Uses slide-backgrounds.csv for decision.
296    """
297    backgrounds = _load_decision_csv("backgrounds")
298    return backgrounds.get(slide_type, {})
299 
300 
301def should_use_full_bleed(slide_index, total_slides, emotion):
302    """
303    Determine if slide should use full-bleed background.
304    Premium decks use 2-3 full-bleed slides strategically.
305 
306    Rules:
307    1. Never consecutive full-bleed
308    2. One in first third, one in middle, one at end
309    3. Reserved for high-emotion beats (hope, urgency, fear)
310    """
311    high_emotion_beats = ["hope", "urgency", "fear", "curiosity"]
312 
313    if emotion not in high_emotion_beats:
314        return False
315 
316    if total_slides < 3:
317        return False
318 
319    third = total_slides // 3
320    strategic_positions = [1, third, third * 2, total_slides - 1]
321 
322    return slide_index in strategic_positions
323 
324 
325def calculate_pattern_break(slide_index, total_slides, previous_emotion=None):
326    """
327    Determine if this slide should break the visual pattern.
328    Used for emotional contrast (Duarte Sparkline technique).
329    """
330    # Pattern breaks at strategic positions
331    if total_slides < 5:
332        return False
333 
334    # Break at 1/3 and 2/3 points
335    third = total_slides // 3
336    if slide_index in [third, third * 2]:
337        return True
338 
339    # Break when switching between frustration and hope
340    contrasting_emotions = {
341        "frustration": ["hope", "relief"],
342        "hope": ["frustration", "fear"],
343        "fear": ["hope", "relief"],
344    }
345 
346    if previous_emotion in contrasting_emotions:
347        return True
348 
349    return False
350 
351 
352def search_with_context(query, slide_position=1, total_slides=9, previous_emotion=None):
353    """
354    Enhanced search that considers deck context.
355 
356    Args:
357        query: Search query
358        slide_position: Current slide index (1-based)
359        total_slides: Total slides in deck
360        previous_emotion: Emotion of previous slide (for contrast)
361 
362    Returns:
363        Search results enriched with contextual recommendations
364    """
365    # Get base results from existing BM25 search
366    base_results = search_all(query, max_results=2)
367 
368    # Detect likely slide goal from query
369    goal = detect_domain(query.lower())
370    if "problem" in query.lower():
371        goal = "problem"
372    elif "solution" in query.lower():
373        goal = "solution"
374    elif "cta" in query.lower() or "call to action" in query.lower():
375        goal = "cta"
376    elif "hook" in query.lower() or "title" in query.lower():
377        goal = "hook"
378    elif "traction" in query.lower() or "metric" in query.lower():
379        goal = "traction"
380 
381    # Enrich with contextual recommendations
382    context = {
383        "slide_position": slide_position,
384        "total_slides": total_slides,
385        "previous_emotion": previous_emotion,
386        "inferred_goal": goal,
387    }
388 
389    # Get layout recommendation
390    layout = get_layout_for_goal(goal, previous_emotion)
391    if layout:
392        context["recommended_layout"] = layout.get("layout_pattern")
393        context["layout_direction"] = layout.get("direction")
394        context["visual_weight"] = layout.get("visual_weight")
395        context["use_background_image"] = layout.get("use_bg_image") == "true"
396 
397    # Get typography recommendation
398    typography = get_typography_for_slide(goal)
399    if typography:
400        context["typography"] = {
401            "primary_size": typography.get("primary_size"),
402            "secondary_size": typography.get("secondary_size"),
403            "weight_contrast": typography.get("weight_contrast"),
404        }
405 
406    # Get color treatment
407    emotion = layout.get("emotion", "clarity") if layout else "clarity"
408    color = get_color_for_emotion(emotion)
409    if color:
410        context["color_treatment"] = {
411            "background": color.get("background"),
412            "text_color": color.get("text_color"),
413            "accent_usage": color.get("accent_usage"),
414            "card_style": color.get("card_style"),
415        }
416 
417    # Calculate pattern breaking
418    context["should_break_pattern"] = calculate_pattern_break(
419        slide_position, total_slides, previous_emotion
420    )
421    context["should_use_full_bleed"] = should_use_full_bleed(
422        slide_position, total_slides, emotion
423    )
424 
425    # Get background config if needed
426    if context.get("use_background_image"):
427        bg_config = get_background_config(goal)
428        if bg_config:
429            context["background"] = {
430                "image_category": bg_config.get("image_category"),
431                "overlay_style": bg_config.get("overlay_style"),
432                "search_keywords": bg_config.get("search_keywords"),
433            }
434 
435    # Suggested animation classes
436    animation_map = {
437        "hook": "animate-fade-up",
438        "problem": "animate-fade-up",
439        "agitation": "animate-count animate-stagger",
440        "solution": "animate-scale",
441        "features": "animate-stagger",
442        "traction": "animate-chart animate-count",
443        "proof": "animate-stagger-scale",
444        "social": "animate-fade-up",
445        "cta": "animate-pulse",
446    }
447    context["animation_class"] = animation_map.get(goal, "animate-fade-up")
448 
449    return {
450        "query": query,
451        "context": context,
452        "base_results": base_results,
453    }
454