Calculate and deduct credit usage from token consumption

[afrinxnahar]

🎯 Problem Statement
Credits are not currently properly calculated/charged based on actual token usage (recently implemented), causing inaccurate accounting for usage-based features.

💡 Proposed Solution

• Define 1 credit = 1000 tokens.
• Instrument server-side request flows to capture token usage.
• Calculate credits_used = ceil(tokens_used / tokens_per_credit).
• Deduct credits_used from the user’s balance atomically.
• Store usage records in credit_usage table.
• Pre-check: if insufficient credits, reject request.

📊 Impact Assessment
User Impact: High (accurate billing).
Development Effort: Medium.
Priority: High.

🧪 Acceptance Criteria

• Token usage captured for all LLM requests.
• Credits deducted based on conversion rule.
• Users see per-request credit deduction on each feature before generating response
• Safe behavior when credits insufficient.

🔗 Related Features
Billing, invoices, credit awarding/referrals, usage dashboard.

📚 Additional Context
Ensure the model's API returns token counts or compute tokens manually.

📋 Checklist

• Token capture for all LLM calls
• Credit conversion and deduction logic
• DB table for usage records
• UX for insufficient credits
• Tests for calculations and concurrency