OrgPulse was first presented at Albania Dreamin' 2026 — the premier Salesforce community conference in the Balkans — on April 25, 2026, at the Pyramid of Tirana, Albania.

The session demonstrated how AI-assisted screenshot analysis combined with a structured prioritization matrix can help architects and developers diagnose performance issues in mature Salesforce orgs.

OrgPulse turns your Scale Center Org Performance screenshots into actionable optimization recommendations. Upload a screenshot, and OrgPulse identifies performance hotspots, validates them against known Scale Center metrics, and maps them to a proven 9-field prioritization matrix (Impact vs. Effort).

OrgPulse is a performance diagnostic tool. It reads Scale Center screenshots and maps observed performance signals (CPU time, request latency, error counters, concurrency issues) to a prioritized remediation matrix.

What OrgPulse does:

• Extract counter values via in-browser OCR
• Interpret chart patterns via Vision LLM (optional, BYOK)
• Match observed signals to concrete Salesforce performance remediation steps
• Highlight the highest-impact actions based on severity scoring

What OrgPulse does NOT do:

• Customer experience strategy or unified profile recommendations
• Agentforce readiness assessment
• Data Cloud adoption strategy (except where relevant for performance offloading)
• Marketing, Commerce, or Service Cloud architecture guidance
• Multi-cloud integration planning

• Runs entirely in your browser using Tesseract.js OCR
• Extracts the six counter values from the Scale Center top bar (Logins, Errors, etc.)
• No data leaves your machine — no account or API key required

• Uses Anthropic Claude Vision to interpret charts, detect spikes, and identify correlations
• Constrained to a known Scale Center metric vocabulary — prevents hallucinated metric names
• Separates findings (anomalies that need attention) from clearances (confirmed healthy areas)
• Returns AI-generated contextual insights with remediation hints specific to your screenshot
• BYOK (Bring Your Own Key) — your API key and screenshots never touch our servers

• Threshold-based severity scoring: each detected signal earns points based on graduated thresholds (info / warning / critical)
• Three visual severity levels on the matrix: low (subtle glow), medium (pulse animation), high (strong pulse + scale)
• Priority Ranking list sorted by cell score, showing signal sources (OCR / Deep Analysis)

• Recommendations are filtered by the signals actually detected — not just by which matrix cell was triggered
• Root cause type matching (compute / data / concurrency / integration / configuration) for better recommendation relevance
• "Show all recommendations" toggle to access the full set when needed

• Deep Analysis automatically runs OCR first for counter extraction, then cross-references with Vision values
• Confidence-aware preference: OCR values trusted when confidence is above 50%, Vision preferred when OCR is unreliable
• Disagreements are transparently displayed with source attribution

• All Vision findings are validated against a whitelist of known Scale Center metric identifiers
• Hallucinated or invented metric names are rejected before they enter the scoring pipeline
• Rejected observations are logged and optionally displayed for transparency

1. Upload a Scale Center Org Performance screenshot (PNG/JPG)
2. Basic Mode extracts counter values using in-browser OCR (Tesseract.js)
3. Deep Mode sends the screenshot to Claude Vision (your API key, direct from your browser) for chart pattern analysis
4. Validation filters all results against known Scale Center metrics — no hallucinated metric names
5. Scoring evaluates findings against threshold-based severity rules and calculates a score per matrix cell
6. Prioritization highlights the most critical cells, shows filtered recommendations matched to detected signals, and displays AI-generated contextual insights

OrgPulse has two build outputs from a single codebase:

• Screenshot-based Scale Center diagnostics only
• Deployed to https://rammc.github.io/orgpulse
• Command: npm run build or npm run dev

• Screenshot diagnostics plus metadata analysis of SFDX projects
• Runs on localhost via npm run dev:local
• Metadata analyzer scans Apex classes, triggers, and Flows for patterns correlating with Scale Center signals
• Not deployed — for architects working with real Salesforce metadata
• Command: npm run dev:local

git clone https://github.com/rammc/orgpulse.git
cd orgpulse
npm install
npm run dev

npm run dev:local

Open http://localhost:5173 in Chrome or Edge (File System Access API required for metadata analysis).

OrgPulse is completely free to use in Basic Mode.

Optional Deep Analysis Mode requires your own Anthropic API key:

• ~$0.02 per screenshot analysis (Claude Sonnet with Vision)
• Your key stays in your browser's LocalStorage
• Screenshots are sent directly from your browser to Anthropic's API
• We never see your screenshots or your key

The maintainer of this project does not pay any per-user costs. Hosting is free via GitHub Pages.

We welcome contributions — especially from Salesforce architects and developers who can improve the recommendation library based on real-world experience.

1. Add Recommendations (no coding required): Edit src/data/recommendations.json — add new entries with title, description, relevant signals, and root cause types. See Contributing Recommendations for the schema guide.

2. Improve Detection Accuracy: Better OCR preprocessing, Vision prompt refinements, validation rules.

3. Report Issues: Found a wrong recommendation? A false positive? A metric that should be a clearance? Open an issue.

See CONTRIBUTING.md for setup instructions and coding standards.

If you're troubleshooting OCR counter extraction, see docs/OCR_DEBUG.md for the diagnostic workflow.

This project was built with significant assistance from AI tools — specifically Claude by Anthropic (Claude Opus and Claude Code). AI was used for:

• Code generation: Initial project scaffolding, UI components, OCR integration, and Vision API wiring
• Architecture design: Data flow, scoring system, recommendation matching logic
• Content creation: Recommendation texts, documentation drafts, prompt engineering for the Vision analysis

All AI-generated code and content has been reviewed, tested, and validated by the maintainer — a Salesforce Certified Technical Architect (CTA) and Salesforce MVP with hands-on experience in enterprise Salesforce performance engineering.

AI is a tool in this project, not a replacement for domain expertise. Every recommendation in the matrix, every threshold value, and every architectural decision reflects real-world Salesforce platform knowledge. The AI helped build it faster — the human ensured it's correct.

OrgPulse is an independent open-source project and is not affiliated with, endorsed by, or sponsored by Salesforce, Inc. Salesforce, Scale Center, Apex, Lightning Web Components, and related marks are trademarks of Salesforce, Inc.

MIT — see LICENSE

Christopher Ramm — Salesforce CTA, Salesforce MVP (Class of 2025), DCX CTO Germany at Capgemini. cramm.dev · LinkedIn · GitHub

Sebastiano Schwarz — Salesforce CTO Capgemini Germany. LinkedIn · GitHub