Billing Architecture Alignment

[1amKhush]

Summary

Addresses the billing audit in #1 by routing prepaid billing through CEP-8, fixing free-model output capping, and making the regular CLI payment-aware.

Changes

CEP-8 billing

• Added CEP-8 resolvePrice handling:
  • Funded chat.complete calls return waivePrice().
  • Unfunded calls receive an immediate -32000 policy rejection.
  • balance.topup returns a Lightning quotePrice(sats).
• Credits the prepaid USD ledger only after payment verification.
• Added NWC Lightning processor configuration.

Payment-aware client and CLI

• Wrapped MuxllClient with CEP-8 client payment middleware.
• Fixed streamed requests waiting indefinitely when authorization fails before the stream opens.
• Added typed client methods for reading and topping up balances.
• Added CLI commands:
  • muxll balance
  • muxll topup --usd <amount>
• Added manual BOLT11 invoice display and payment confirmation.
• Added MUXLL_CLIENT_PRIVATE_KEY for maintaining the same prepaid identity across CLI invocations.
• Added an actionable top-up hint when chat is rejected for insufficient balance.

Pricing

• Fixed free-output models defaulting to a one-token output cap when no explicit user or model limit is provided.
• Free models now fall back to their context window.

Tests and documentation

• Added regression coverage for:
  • CEP-8 policy rejection.
  • Unfunded streamed requests.
  • Lightning top-up and ledger crediting.
  • Balance retrieval using the same client identity.
  • Free-model output limits.
  • Actionable CLI billing errors.
• Documented billing, NWC, persistent client identity, balance, and top-up configuration.

Verification

• Typecheck passes.
• Lint passes.
• Formatting checks pass.
• Full test suite: 79 passed, 2 skipped, 0 failed.
• Live Lightning smoke test completed successfully:
  • BOLT11 invoice generated.
  • Payment accepted.
  • Ledger credited.
  • Chat completed.
  • Actual usage debited from the prepaid balance.

Refs #1

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.