- Category enhance prompts (Sync A1 / seed-a1 / repair) now use the same three role sections as the rest of the platform: Title / Description / Meta. The retired "--- Attributes ---" role section is removed from the canonical template, the legacy-split overlay, and the web prompt editor; Category/Attrs stay as plain product-context lines (attribute extraction remains pipeline-level via AppendAttributeConstraints). - Stored prompts still carrying an attributes section are detected by CategoryEnhancePromptNeedsRepair and rewritten on the next Sync. - Legacy wp_product_categories.sql splits now also seed a default Slovenian metaTitle rule (dumps only carried <metaDescription>), so every A1 category gets all four prompt areas: title formula, description sections, meta title, meta description. - Role-sectioned prompts no longer imply the A1 SEO-omit cohort (CompanyOmitsSEOMeta): sectioned prompts are the canonical prompt-editor output for every tenant, and the sniff silently disabled SEO meta for any company that saved a category prompt. - Verified locally: repair-category-prompts -apply updated 238 A1 + Platform Demo categories from scripts/seed/wp_product_categories.sql (216 legacy splits), idempotent on re-run, zero attributes sections left in DB. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
176 lines
6.8 KiB
Go
176 lines
6.8 KiB
Go
package aiprompts
|
|
|
|
import (
|
|
"regexp"
|
|
"strings"
|
|
)
|
|
|
|
var (
|
|
reLegacyNameTag = regexp.MustCompile(`(?is)<\s*name\s*>\s*\{?\s*(.*?)\s*\}?\s*<\s*/\s*name\s*>`)
|
|
reLegacyMetaTag = regexp.MustCompile(`(?is)<\s*metaDescription\s*>\s*\{?\s*(.*?)\s*\}?\s*<\s*/\s*metaDescription\s*>`)
|
|
reLegacyDescPH = regexp.MustCompile(`(?i)\{\s*""?\s*OPIS\s+IZDELKA\s*""?\s*\}`)
|
|
reLegacyNamePH = regexp.MustCompile(`(?i)\{\s*""?\s*STARO\s+IME\s+IZDELKA\s*""?\s*\}`)
|
|
reDoubledQuotes = regexp.MustCompile(`"{2,}`)
|
|
)
|
|
|
|
// LegacyEnhanceParts holds extracted role content from a combined PHP/A1 category Prompt.
|
|
type LegacyEnhanceParts struct {
|
|
TitleRules string
|
|
DescriptionRules string
|
|
MetaRules string
|
|
WasLegacy bool
|
|
}
|
|
|
|
// IsLegacyCombinedEnhancePrompt reports whether prompt looks like the old A1/PHP
|
|
// combined name+description(+meta) marketing blob (<name>/<metaDescription>/GPT predloga).
|
|
func IsLegacyCombinedEnhancePrompt(prompt string) bool {
|
|
p := strings.TrimSpace(prompt)
|
|
if p == "" {
|
|
return false
|
|
}
|
|
// Require real tag wrappers — not instructional mentions like
|
|
// `do NOT … <name>/<metaDescription> tags` in CategoryEnhanceUserTemplate.
|
|
hasNameTag := reLegacyNameTag.MatchString(p)
|
|
hasMetaTag := reLegacyMetaTag.MatchString(p)
|
|
if hasNameTag && (hasMetaTag || strings.Contains(strings.ToLower(p), "gpt predloga")) {
|
|
return true
|
|
}
|
|
if hasNameTag && (strings.Contains(p, "STARO IME IZDELKA") || strings.Contains(p, "OPIS IZDELKA")) {
|
|
return true
|
|
}
|
|
if strings.Contains(p, "STARO IME IZDELKA") && strings.Contains(p, "OPIS IZDELKA") && strings.Contains(strings.ToLower(p), "gpt predloga") {
|
|
return true
|
|
}
|
|
return false
|
|
}
|
|
|
|
// ParseLegacyCombinedEnhancePrompt extracts title / description / meta instruction
|
|
// blobs from a legacy combined prompt. Placeholders are modernized to {{name}} /
|
|
// {{description}}. WasLegacy is false when no combined markers are found.
|
|
func ParseLegacyCombinedEnhancePrompt(prompt string) LegacyEnhanceParts {
|
|
raw := strings.TrimSpace(prompt)
|
|
out := LegacyEnhanceParts{}
|
|
if raw == "" || !IsLegacyCombinedEnhancePrompt(raw) {
|
|
return out
|
|
}
|
|
out.WasLegacy = true
|
|
|
|
if m := reLegacyNameTag.FindStringSubmatch(raw); len(m) == 2 {
|
|
out.TitleRules = cleanLegacyInstruction(m[1])
|
|
}
|
|
if m := reLegacyMetaTag.FindStringSubmatch(raw); len(m) == 2 {
|
|
out.MetaRules = cleanLegacyInstruction(m[1])
|
|
}
|
|
|
|
rest := reLegacyNameTag.ReplaceAllString(raw, "")
|
|
rest = reLegacyMetaTag.ReplaceAllString(rest, "")
|
|
if idx := strings.Index(strings.ToLower(rest), "gpt predloga:"); idx >= 0 {
|
|
rest = rest[idx+len("gpt predloga:"):]
|
|
}
|
|
// Drop Slovenian boilerplate intro lines that are not body structure.
|
|
rest = stripLegacyIntroNoise(rest)
|
|
out.DescriptionRules = cleanLegacyInstruction(rest)
|
|
return out
|
|
}
|
|
|
|
// SplitLegacyCombinedEnhancePrompt rewrites a legacy combined name+description(+meta)
|
|
// prompt into CategoryEnhanceUserTemplate role sections, preserving Slovenian
|
|
// naming / HTML / meta intent as category-specific rules under Title / Description / Meta.
|
|
// Non-legacy input returns CategoryEnhanceUserTemplate (canonical shared overlay).
|
|
func SplitLegacyCombinedEnhancePrompt(prompt string) string {
|
|
parts := ParseLegacyCombinedEnhancePrompt(prompt)
|
|
if !parts.WasLegacy {
|
|
return strings.TrimSpace(CategoryEnhanceUserTemplate)
|
|
}
|
|
return BuildCategoryEnhanceOverlay(parts)
|
|
}
|
|
|
|
// BuildCategoryEnhanceOverlay composes a role-sectioned enhance USER overlay from
|
|
// extracted legacy parts (or empty parts → canonical template text with no extra rules).
|
|
func BuildCategoryEnhanceOverlay(parts LegacyEnhanceParts) string {
|
|
var b strings.Builder
|
|
b.WriteString(`Your reply is parsed as JSON {"name":"string","description":"string","meta_title":"string","meta_description":"string","attrs":{}} only (system schema). Write all string fields in {{language}} (do not hardcode a language).`)
|
|
b.WriteString("\n\n")
|
|
|
|
b.WriteString(SectionTitleStart)
|
|
b.WriteByte('\n')
|
|
b.WriteString(TitleRoleInstruction)
|
|
b.WriteByte('\n')
|
|
if rules := strings.TrimSpace(parts.TitleRules); rules != "" {
|
|
b.WriteString("Category naming rules (preserve intent): ")
|
|
b.WriteString(rules)
|
|
b.WriteByte('\n')
|
|
}
|
|
b.WriteString("Name: {{name}}\n")
|
|
b.WriteString(SectionTitleEnd)
|
|
b.WriteString("\n\n")
|
|
|
|
b.WriteString(SectionDescriptionStart)
|
|
b.WriteString("\nRole: description. Build JSON \"description\": product body HTML only (not SEO meta). When a Description formula follows, emit ONE HTML string covering each section in order (tags matching type: h1/h2/h3/h4, p, ul); otherwise prefer 1-3 factual paragraphs as ONE string with limited HTML (<h2><p><ul><li>) — do NOT emit a competing full HTML document or wrap the whole reply in <name>/<metaDescription> tags.\n")
|
|
if rules := strings.TrimSpace(parts.DescriptionRules); rules != "" {
|
|
b.WriteString("Category HTML structure (preserve intent; emit as ONE description HTML string, not tagged name/meta blocks):\n")
|
|
b.WriteString(rules)
|
|
b.WriteByte('\n')
|
|
}
|
|
b.WriteString("Description: {{description}}\n")
|
|
b.WriteString(SectionDescriptionEnd)
|
|
b.WriteString("\n\n")
|
|
|
|
b.WriteString(SectionMetaStart)
|
|
b.WriteString("\nRole: meta. Build JSON \"meta_title\" and \"meta_description\" as plain SEO text (never HTML). meta_title: 50-60 chars; meta_description: 120-155 chars; follow any SEO meta formula that follows; never copy the full description HTML into meta_description.\n")
|
|
if rules := strings.TrimSpace(parts.MetaRules); rules != "" {
|
|
b.WriteString("Category SEO rules (preserve intent; plain text only): ")
|
|
b.WriteString(rules)
|
|
b.WriteByte('\n')
|
|
}
|
|
b.WriteString(SectionMetaEnd)
|
|
b.WriteString("\n\n")
|
|
|
|
b.WriteString("Product context:\n")
|
|
b.WriteString("Category: {{category}}\n")
|
|
b.WriteString("Attrs: {{attrs}}")
|
|
return strings.TrimSpace(b.String())
|
|
}
|
|
|
|
func cleanLegacyInstruction(s string) string {
|
|
s = modernizeLegacyPlaceholders(s)
|
|
s = reDoubledQuotes.ReplaceAllString(s, `"`)
|
|
s = strings.ReplaceAll(s, "\r\n", "\n")
|
|
s = strings.TrimSpace(s)
|
|
// Collapse excessive blank lines.
|
|
for strings.Contains(s, "\n\n\n") {
|
|
s = strings.ReplaceAll(s, "\n\n\n", "\n\n")
|
|
}
|
|
return s
|
|
}
|
|
|
|
func modernizeLegacyPlaceholders(prompt string) string {
|
|
prompt = reLegacyDescPH.ReplaceAllString(prompt, "{{description}}")
|
|
prompt = reLegacyNamePH.ReplaceAllString(prompt, "{{name}}")
|
|
return prompt
|
|
}
|
|
|
|
func stripLegacyIntroNoise(s string) string {
|
|
lines := strings.Split(s, "\n")
|
|
out := make([]string, 0, len(lines))
|
|
for _, line := range lines {
|
|
trim := strings.TrimSpace(line)
|
|
lower := strings.ToLower(trim)
|
|
switch {
|
|
case trim == "":
|
|
if len(out) > 0 {
|
|
out = append(out, "")
|
|
}
|
|
case strings.HasPrefix(lower, "ustvari nov opis"),
|
|
strings.HasPrefix(lower, "star_opis_izdelka"),
|
|
strings.HasPrefix(lower, "staro_ime_izdelka"),
|
|
strings.HasPrefix(lower, "uporabi spodnjo gpt"),
|
|
strings.HasPrefix(lower, "sledi tej gpt"):
|
|
continue
|
|
default:
|
|
out = append(out, trim)
|
|
}
|
|
}
|
|
return strings.TrimSpace(strings.Join(out, "\n"))
|
|
}
|