/**
 * Vision Context Detector
 *
 * Deterministically evaluates whether a user prompt requests analysis, explanation, or interpretation
 * of an existing image/diagram/screenshot/chart in the current chat.
 *
 * Excludes:
 * 1. Fresh/unnamed chats (isFreshChat === true) where searched images cannot yet exist.
 * 2. Search Images queries ("Explain X with images", "Show me images of Y").
 * 3. Pure image lookups lacking analysis/explanation intent ("Find the second image").
 * 4. General conceptual questions mentioning image/diagram without specific reference.
 */

const ANALYSIS_INTENT_PATTERN =
  /\b(?:analyze|analyzing|analysis|explain|explaining|understand|understanding|describe|describing|interpret|interpreting|read|reading|compare|comparing|look\s+at|looking\s+at|extract|extracting|find\s+(?:the\s+)?errors?|find\s+(?:the\s+)?bugs?|find\s+what(?:\s+is|'s)?\s+wrong|what(?:'s|s|\s+is|\s+'s)?\s+wrong(?:\s+with)?|what\s+can\s+you\s+tell\s+me|tell\s+me\s+about|what\s+does|what\s+is|whats)\b/i;

const IMAGE_NOUNS_SOURCE =
  "(?:images?|pics?|pictures?|photos?|diagrams?|figures?|illustrations?|screenshots?|charts?|flowcharts?|schematics?|drawings?)";

const ORDINALS_SOURCE =
  "(?:\\d+(?:st|nd|rd|th)|first|second|third|fourth|fifth|sixth|seventh|eighth|ninth|tenth|eleventh|twelfth|thirteenth|fourteenth|fifteenth|sixteenth|seventeenth|eighteenth|nineteenth|twentieth)";

const DEMONSTRATIVES_PATTERN = new RegExp(
  `\\b(?:this|that|these|those|both|all)\\s+${IMAGE_NOUNS_SOURCE}\\b`,
  "i"
);

const ORDINAL_REF_PATTERN = new RegExp(
  `\\b(?:the\\s+)?${ORDINALS_SOURCE}\\s+${IMAGE_NOUNS_SOURCE}\\b`,
  "i"
);

const ORDINAL_JOINED_REF_PATTERN = new RegExp(
  `\\b(?:the\\s+)?${ORDINALS_SOURCE}\\s+(?:and|or)\\s+${ORDINALS_SOURCE}\\s+${IMAGE_NOUNS_SOURCE}\\b`,
  "i"
);

const INDEXED_REF_PATTERN = new RegExp(
  `\\b${IMAGE_NOUNS_SOURCE}\\s+(?:#?\\d+|[0-9]+)\\b`,
  "i"
);

const SPATIAL_ABOVE_PATTERN = new RegExp(
  `\\b(?:the\\s+)?(?:above\\s+${IMAGE_NOUNS_SOURCE}|${IMAGE_NOUNS_SOURCE}\\s+above)\\b`,
  "i"
);

const PURE_LOOKUP_PATTERN = new RegExp(
  `^(?:find|fetch|get|show|display|look\\s+for|search\\s+for)\\s+(?:the\\s+)?(?:${ORDINALS_SOURCE}\\s+)?${IMAGE_NOUNS_SOURCE}(?:\\s+#?\\d+)?$`,
  "i"
);

export interface VisionContextOptions {
  isFreshChat?: boolean;
  requestId?: string;
}

function matchImageReference(text: string): string | null {
  const domMatch = DEMONSTRATIVES_PATTERN.exec(text);
  if (domMatch) return domMatch[0];

  const ordJoinedMatch = ORDINAL_JOINED_REF_PATTERN.exec(text);
  if (ordJoinedMatch) return ordJoinedMatch[0];

  const ordMatch = ORDINAL_REF_PATTERN.exec(text);
  if (ordMatch) return ordMatch[0];

  const idxMatch = INDEXED_REF_PATTERN.exec(text);
  if (idxMatch) return idxMatch[0];

  const spatialMatch = SPATIAL_ABOVE_PATTERN.exec(text);
  if (spatialMatch) return spatialMatch[0];

  return null;
}

export function requiresVisionContext(
  userText: string,
  options?: VisionContextOptions
): boolean {
  const text = typeof userText === "string" ? userText.trim() : "";
  const isFreshChat = options?.isFreshChat === true;

  const analysisMatch = text ? ANALYSIS_INTENT_PATTERN.exec(text) : null;
  const hasAnalysisIntent = analysisMatch !== null;

  const matchedRef = text ? matchImageReference(text) : null;
  const hasImageReference = matchedRef !== null;

  // Pure lookup check: e.g. "Find the second image" lacking explicit analysis/explanation keywords
  const isPureLookup =
    text.length > 0 &&
    PURE_LOOKUP_PATTERN.test(text) &&
    !/\b(?:explain|analyze|describe|interpret|error|bug|wrong|meaning|show|showing|tell)\b/i.test(text);

  const finalResult =
    text.length > 0 &&
    !isFreshChat &&
    hasAnalysisIntent &&
    hasImageReference &&
    !isPureLookup;

  return finalResult;
}

/**
 * Vision Context Detector
 *
 * Deterministically evaluates whether a user prompt requests analysis, explanation, or interpretation
 * of an existing image/diagram/screenshot/chart in the current chat.
 *
 * Excludes:
 * 1. Fresh/unnamed chats (isFreshChat === true) where searched images cannot yet exist.
 * 2. Search Images queries ("Explain X with images", "Show me images of Y").
 * 3. Pure image lookups lacking analysis/explanation intent ("Find the second image").
 * 4. General conceptual questions mentioning image/diagram without specific reference.
 */

const ANALYSIS_INTENT_PATTERN =
  /\b(?:analyze|analyzing|analysis|explain|explaining|understand|understanding|describe|describing|interpret|interpreting|read|reading|compare|comparing|look\s+at|looking\s+at|extract|extracting|find\s+(?:the\s+)?errors?|find\s+(?:the\s+)?bugs?|find\s+what(?:\s+is|'s)?\s+wrong|what(?:'s|s|\s+is|\s+'s)?\s+wrong(?:\s+with)?|what\s+can\s+you\s+tell\s+me|tell\s+me\s+about|what\s+does|what\s+is|whats)\b/i;

const IMAGE_NOUNS_SOURCE =
  "(?:images?|pics?|pictures?|photos?|diagrams?|figures?|illustrations?|screenshots?|charts?|flowcharts?|schematics?|drawings?)";

const ORDINALS_SOURCE =
  "(?:\\d+(?:st|nd|rd|th)|first|second|third|fourth|fifth|sixth|seventh|eighth|ninth|tenth|eleventh|twelfth|thirteenth|fourteenth|fifteenth|sixteenth|seventeenth|eighteenth|nineteenth|twentieth)";

const DEMONSTRATIVES_PATTERN = new RegExp(
  `\\b(?:this|that|these|those|both|all)\\s+${IMAGE_NOUNS_SOURCE}\\b`,
  "i"
);

const ORDINAL_REF_PATTERN = new RegExp(
  `\\b(?:the\\s+)?${ORDINALS_SOURCE}\\s+${IMAGE_NOUNS_SOURCE}\\b`,
  "i"
);

const ORDINAL_JOINED_REF_PATTERN = new RegExp(
  `\\b(?:the\\s+)?${ORDINALS_SOURCE}\\s+(?:and|or)\\s+${ORDINALS_SOURCE}\\s+${IMAGE_NOUNS_SOURCE}\\b`,
  "i"
);

const INDEXED_REF_PATTERN = new RegExp(
  `\\b${IMAGE_NOUNS_SOURCE}\\s+(?:#?\\d+|[0-9]+)\\b`,
  "i"
);

const SPATIAL_ABOVE_PATTERN = new RegExp(
  `\\b(?:the\\s+)?(?:above\\s+${IMAGE_NOUNS_SOURCE}|${IMAGE_NOUNS_SOURCE}\\s+above)\\b`,
  "i"
);

const PURE_LOOKUP_PATTERN = new RegExp(
  `^(?:find|fetch|get|show|display|look\\s+for|search\\s+for)\\s+(?:the\\s+)?(?:${ORDINALS_SOURCE}\\s+)?${IMAGE_NOUNS_SOURCE}(?:\\s+#?\\d+)?$`,
  "i"
);

export interface VisionContextOptions {
  isFreshChat?: boolean;
  requestId?: string;
}

function matchImageReference(text: string): string | null {
  const domMatch = DEMONSTRATIVES_PATTERN.exec(text);
  if (domMatch) return domMatch[0];

  const ordJoinedMatch = ORDINAL_JOINED_REF_PATTERN.exec(text);
  if (ordJoinedMatch) return ordJoinedMatch[0];

  const ordMatch = ORDINAL_REF_PATTERN.exec(text);
  if (ordMatch) return ordMatch[0];

  const idxMatch = INDEXED_REF_PATTERN.exec(text);
  if (idxMatch) return idxMatch[0];

  const spatialMatch = SPATIAL_ABOVE_PATTERN.exec(text);
  if (spatialMatch) return spatialMatch[0];

  return null;
}

export function requiresVisionContext(
  userText: string,
  options?: VisionContextOptions
): boolean {
  const text = typeof userText === "string" ? userText.trim() : "";
  const isFreshChat = options?.isFreshChat === true;

  const analysisMatch = text ? ANALYSIS_INTENT_PATTERN.exec(text) : null;
  const hasAnalysisIntent = analysisMatch !== null;

  const matchedRef = text ? matchImageReference(text) : null;
  const hasImageReference = matchedRef !== null;

  // Pure lookup check: e.g. "Find the second image" lacking explicit analysis/explanation keywords
  const isPureLookup =
    text.length > 0 &&
    PURE_LOOKUP_PATTERN.test(text) &&
    !/\b(?:explain|analyze|describe|interpret|error|bug|wrong|meaning|show|showing|tell)\b/i.test(text);

  const finalResult =
    text.length > 0 &&
    !isFreshChat &&
    hasAnalysisIntent &&
    hasImageReference &&
    !isPureLookup;

  return finalResult;
}