Sangeetha-Grantha

Metadata	Value
Status	Archived
Version	1.0.0
Last Updated	2026-09-10
Author	Sangeetha Grantha Team
Document Type	Archive

Update Summary: Guru-Guha Blogspot Source Correction

---

[!NOTE] Historical evidence: results, counts, commands, and observations below belong to the original work described here. The editorial update date is not a new test or corpus verification. For present behavior, use current ingestion guide.

---

Date: 2026-01-19

Changes Made

Source Characterization Update

Previous Understanding:

• guru-guha.blogspot.com was characterized as unstructured blog posts about Murugan/Subramanya compositions
• Assumed it would require AI/LLM extraction due to prose format
• Estimated 300-800 krithis focused on Murugan deity

Corrected Understanding:

• guru-guha.blogspot.com hosts alphabetical lists for specific composers
• Specific URL identified: https://guru-guha.blogspot.com/2009/04/dikshitar-kritis-alphabetical-list.html
• Focus: Muthuswami Dikshitar compositions (not Murugan-specific)
• Structure: List-based format similar to Thyagaraja and Syama Krishna blogspot sources
• Can use regex pattern matching instead of AI extraction
• Estimated 300-700 Dikshitar krithis

Technical Implementation Impact

Scraper Implementation:

• Changed from GeminiService-based AI extraction to pattern matching
• Removed unstructured blog post parsing logic
• Added list parsing with multiple regex patterns:
  • Pattern 1: Name - Raga format
  • Pattern 2: Name (Raga) format
• Composer is implicit (determined from URL/page title)
• Returns batch entries rather than single krithi per page

Configuration Changes:

// OLD
extractionStrategy = ExtractionStrategy.AI_EXTRACTION
requiresAIExtraction = true
implicitDeity = "Murugan"

// NEW
extractionStrategy = ExtractionStrategy.PATTERN_MATCHING
requiresAIExtraction = false
implicitComposer = "Muthuswami Dikshitar"

Document Updates Made

1. Data Source Comparison Matrix (Section 2.1)
   • Updated structure: “List-based” (was “Unstructured blog posts”)
   • Updated challenge: “Composer implicit (Dikshitar), parsing list format”
   • Updated volume: 300-700 krithis
2. Scraper Implementation (Section 7.5)
   • Complete rewrite of GuruGuhaBlogScraper class
   • Removed Gemini/AI extraction logic
   • Added parseListEntries() method with regex patterns
   • Added extractComposerFromTitle() helper
3. Koog Integration Strategy (Section 8)
   • Updated use cases: removed “unstructured blog extraction”
   • Changed to “semi-structured list extraction” (if needed)
   • Reduced emphasis on AI extraction for this source
4. Configuration Examples (Appendix A)
   • Updated guruGuhaBlogConfig with correct parameters

Implications

Positive Impacts:

✅ Simpler implementation - no AI extraction needed ✅ Lower costs - no LLM API calls for this source ✅ Faster processing - regex is faster than AI inference ✅ More reliable - deterministic parsing vs probabilistic AI ✅ Consistency - all three blogspot sources now use similar list-based approach

Considerations:

⚠️ May have other composer lists - the blog might have additional pages for other composers ⚠️ Pattern variations - list format may vary between pages ⚠️ Less metadata - no lyrics, deity, or temple information in list format

Recommendations

1. Discovery Phase: Crawl guru-guha.blogspot.com to find other composer-specific list pages
2. Pattern Testing: Test regex patterns on actual page content to ensure accurate parsing
3. Fallback Strategy: If list parsing fails, can still fall back to AI extraction for specific pages
4. Volume Validation: Verify actual number of Dikshitar krithis on the page

Next Steps

• Implement URL discovery to find additional composer lists on guru-guha.blogspot.com
• Test regex patterns against actual HTML
• Update entity resolution to handle “Dikshitar” → “Muthuswami Dikshitar” mapping
• Consider this source as Medium priority (after karnatik.com, similar to other blogspot sources)

---

Documentation home · Feature status