feat(audit): add pricing signature guards and reporting

Add snapshot, signature, and drift guard support for Vertex AI, Cloudflare Workers AI, and Perplexity API, backed by a queryable audit table and recent-window view.

This commit also wires the audit query layer into daily signal materialization and report generation so structure drift becomes a first-class signal instead of a log-only artifact.

scripts/perplexity_pricing_lib.go

@@ -0,0 +1,150 @@
+//go:build llm_script
+
+package main
+
+import (
+	"fmt"
+	"regexp"
+	"strings"
+)
+
+const (
+	defaultPerplexityPricingFetchURL  = "https://docs.perplexity.ai/docs/agent-api/models.md"
+	defaultPerplexityPricingSourceURL = "https://docs.perplexity.ai/docs/agent-api/models"
+)
+
+var markdownLinkPattern = regexp.MustCompile(`\[(.*?)\]\((https://[^)]+)\)`)
+
+func parsePerplexityPricingCatalog(raw string) ([]officialPricingRecord, error) {
+	lines := strings.Split(raw, "\n")
+	records := make([]officialPricingRecord, 0)
+	header := []string(nil)
+	modelIndex := -1
+	inputIndex := -1
+	outputIndex := -1
+	docIndex := -1
+	for _, line := range lines {
+		line = strings.TrimSpace(line)
+		if !strings.HasPrefix(line, "|") {
+			continue
+		}
+
+		parts := splitMarkdownTableRow(line)
+		if len(parts) == 0 {
+			continue
+		}
+		if header == nil {
+			header = parts
+			modelIndex, inputIndex, outputIndex, docIndex = detectPerplexityTableColumns(parts)
+			continue
+		}
+		if isMarkdownTableSeparator(parts) {
+			continue
+		}
+		if modelIndex < 0 || inputIndex < 0 || outputIndex < 0 || modelIndex >= len(parts) || inputIndex >= len(parts) || outputIndex >= len(parts) {
+			continue
+		}
+
+		modelPath := strings.Trim(parts[modelIndex], "`")
+		inputCell := parts[inputIndex]
+		outputCell := parts[outputIndex]
+		inputPrice, ok := firstDollarPrice(inputCell)
+		if !ok {
+			continue
+		}
+		outputPrice, ok := firstDollarPrice(outputCell)
+		if !ok {
+			continue
+		}
+		sourceURL := defaultPerplexityPricingSourceURL
+		if docIndex >= 0 && docIndex < len(parts) {
+			if matches := markdownLinkPattern.FindStringSubmatch(parts[docIndex]); len(matches) == 3 {
+				sourceURL = matches[2]
+			}
+		}
+		providerName := providerFromModelPath(modelPath)
+		providerNameCn, providerCountry, providerWebsite := providerMetadata(providerName)
+		record := officialPricingRecord{
+			ModelID:         normalizeExternalID("perplexity", modelPath),
+			ModelName:       modelPath,
+			ProviderName:    providerName,
+			ProviderNameCn:  providerNameCn,
+			ProviderCountry: providerCountry,
+			ProviderWebsite: providerWebsite,
+			OperatorName:    "Perplexity API",
+			OperatorNameCn:  "Perplexity API",
+			OperatorCountry: "US",
+			OperatorWebsite: "https://docs.perplexity.ai",
+			OperatorType:    "relay",
+			Region:          "global",
+			Currency:        "USD",
+			InputPrice:      inputPrice,
+			OutputPrice:     outputPrice,
+			SourceURL:       defaultPerplexityPricingSourceURL,
+			ModelSourceURL:  sourceURL,
+			DateConfidence:  "unknown",
+			DateSourceKind:  "official_pricing",
+			Modality:        detectModality(modelPath),
+		}
+		record.IsFree = record.InputPrice == 0 && record.OutputPrice == 0
+		records = append(records, record)
+	}
+	if len(records) == 0 {
+		return nil, fmt.Errorf("unexpected perplexity pricing content")
+	}
+	return records, nil
+}
+
+func splitMarkdownTableRow(line string) []string {
+	trimmed := strings.TrimSpace(line)
+	trimmed = strings.TrimPrefix(trimmed, "|")
+	trimmed = strings.TrimSuffix(trimmed, "|")
+	if trimmed == "" {
+		return nil
+	}
+	parts := strings.Split(trimmed, "|")
+	result := make([]string, 0, len(parts))
+	for _, part := range parts {
+		result = append(result, strings.TrimSpace(part))
+	}
+	return result
+}
+
+func detectPerplexityTableColumns(header []string) (int, int, int, int) {
+	modelIndex := -1
+	inputIndex := -1
+	outputIndex := -1
+	docIndex := -1
+	for i, col := range header {
+		lower := strings.ToLower(strings.TrimSpace(col))
+		switch {
+		case strings.Contains(lower, "model") && modelIndex == -1:
+			modelIndex = i
+		case strings.Contains(lower, "input") && strings.Contains(lower, "price") && inputIndex == -1:
+			inputIndex = i
+		case strings.Contains(lower, "output") && strings.Contains(lower, "price") && outputIndex == -1:
+			outputIndex = i
+		case (strings.Contains(lower, "documentation") || strings.Contains(lower, "docs")) && docIndex == -1:
+			docIndex = i
+		}
+	}
+	return modelIndex, inputIndex, outputIndex, docIndex
+}
+
+func isMarkdownTableSeparator(parts []string) bool {
+	if len(parts) == 0 {
+		return false
+	}
+	for _, part := range parts {
+		trimmed := strings.TrimSpace(part)
+		if trimmed == "" {
+			return false
+		}
+		for _, ch := range trimmed {
+			if ch != '-' && ch != ':' {
+				return false
+			}
+		}
+	}
+	return true
+}