Skip to content

datasette/datasette-llm-accountant

Repository files navigation

datasette-llm-accountant

PyPI Changelog Tests License

Budget management and cost tracking for LLM usage in Datasette.

Installation

Install this plugin in the same environment as Datasette:

datasette install datasette-llm-accountant

This plugin works alongside datasette-llm to provide automatic cost tracking and budget enforcement for LLM prompts.

Overview

This plugin provides:

  • Automatic cost calculation based on token usage and model pricing
  • Reserve/settle pattern for budget enforcement
  • Accountant plugin system for custom spending trackers
  • Custom pricing providers with access to the full response object
  • Hook integration with datasette-llm for transparent accounting

When installed, all prompts made through datasette-llm are automatically wrapped with accounting logic. Accountants can enforce spending limits, log usage, and track costs.

How It Works

  1. When a prompt is made via datasette-llm, this plugin's hooks intercept the call
  2. A reservation is made with all registered accountants for the estimated cost
  3. The prompt executes
  4. The actual cost is calculated via PricingProvider.calculate_cost_from_response()
  5. Accountants are settled with the real cost (refunding any unused reservation)

Configuration

Configure reservation amounts in datasette.yaml:

plugins:
  datasette-llm-accountant:
    # Default reservation for single prompts
    auto_reservation_usd: 0.10

    # Default reservation for grouped prompts
    default_reservation_usd: 0.50

    # Purpose-specific reservations
    purposes:
      enrichments:
        reservation_usd: 5.00
      query-assistant:
        reservation_usd: 0.25

Nanocents

All monetary amounts use the Nanocents type — an int subclass that makes units explicit and prevents accidentally passing raw dollar amounts where nanocents are expected.

  • 1 nanocent = 1/1,000,000,000 of a cent
  • 1 USD = 100 cents = 100,000,000,000 nanocents
  • This allows tracking costs down to fractions of a cent without floating-point errors
from datasette_llm_accountant import Nanocents

# Create from USD or cents
cost = Nanocents.from_usd(1.50)   # 150,000,000,000
cost = Nanocents.from_cents(50)   # 50,000,000,000

# Convert back
cost.to_usd()    # 1.5
cost.to_cents()  # 150.0

# Works like a regular int for arithmetic
total = cost + Nanocents.from_usd(0.50)

Creating an Accountant Plugin

Accountants track and enforce LLM spending. Create a plugin that implements the register_llm_accountants hook:

from datasette import hookimpl
from datasette_llm_accountant import Accountant, Tx, Nanocents, InsufficientBalanceError

class MyAccountant(Accountant):
    """Custom accountant that tracks spending."""

    def __init__(self, datasette):
        self.datasette = datasette

    async def reserve(
        self,
        nanocents: Nanocents,
        model_id: str = None,
        purpose: str = None,
        actor_id: str = None,
    ) -> Tx:
        """Reserve the specified amount. Raise InsufficientBalanceError to block."""
        if not await self.has_sufficient_balance(nanocents):
            raise InsufficientBalanceError("Insufficient balance")
        tx_id = await self.create_reservation(nanocents, model_id, purpose)
        return Tx(tx_id)

    async def settle(
        self,
        tx: Tx,
        nanocents: Nanocents,
        model_id: str = None,
        purpose: str = None,
        actor_id: str = None,
    ):
        """Settle a transaction for the actual amount spent."""
        await self.record_settlement(tx, nanocents, model_id, purpose)

    async def rollback(self, tx: Tx):
        """Optional: Release a reservation without charging."""
        await self.settle(tx, Nanocents(0))

@hookimpl
def register_llm_accountants(datasette):
    return [MyAccountant(datasette)]

See datasette-llm-allowance for a complete implementation that uses Datasette's internal database to track a spending allowance.

Multiple Accountants

Multiple accountants can be registered. When a reservation is made:

  1. All accountants are called in sequence to reserve the amount
  2. If any accountant fails (e.g., InsufficientBalanceError), previous reservations are rolled back
  3. When the prompt completes, all accountants are settled with the actual cost

This enables layered accounting (per-user limits, per-project budgets, global caps, etc.).

Custom Pricing Providers

The default pricing provider fetches model prices from llm-prices.com. You can register a custom provider to control how costs are calculated:

from datasette import hookimpl
from datasette_llm_accountant import PricingProvider, Nanocents

class MyPricingProvider(PricingProvider):
    async def calculate_cost_from_response(self, model_id, usage, response):
        """
        Calculate cost from a completed response.

        Args:
            model_id: The model identifier
            usage: An llm.Usage object with input/output token counts
            response: The llm.AsyncResponse object — use this to access
                      provider-specific metadata (e.g., generation IDs
                      for exact cost lookups from the provider's API)
        Returns:
            Cost as a Nanocents value
        """
        input_tokens = usage.input or 0
        output_tokens = usage.output or 0
        cost_usd = input_tokens * 0.01 / 1_000_000 + output_tokens * 0.03 / 1_000_000
        return Nanocents.from_usd(cost_usd)

    async def supported_models(self):
        """Return the set of model IDs this provider can price,
        or None to indicate all models are supported.

        Used to filter the model list — models not in this set
        won't be available when accountants are registered.
        """
        return self.known_models

@hookimpl
def register_llm_accountant_pricing(datasette):
    return MyPricingProvider()

API Reference

Accountant Base Class

class Accountant(ABC):
    @abstractmethod
    async def reserve(
        self,
        nanocents: Nanocents,
        model_id: str = None,
        purpose: str = None,
        actor_id: str = None,
    ) -> Tx:
        """Reserve an amount, return transaction ID."""

    @abstractmethod
    async def settle(
        self,
        tx: Tx,
        nanocents: Nanocents,
        model_id: str = None,
        purpose: str = None,
        actor_id: str = None,
    ):
        """Settle a transaction for the actual amount."""

    async def rollback(self, tx: Tx):
        """Release a reservation (default: settle for 0)."""
        await self.settle(tx, Nanocents(0))

PricingProvider Base Class

class PricingProvider(ABC):
    @abstractmethod
    async def calculate_cost_from_response(
        self, model_id: str, usage: Usage, response: AsyncResponse
    ) -> Nanocents:
        """Calculate cost from a completed response."""

    async def supported_models(self) -> Optional[set[str]]:
        """Return set of supported model IDs, or None for all. Default: None."""
        return None

Exceptions

  • InsufficientBalanceError - Raised when an accountant cannot reserve the requested amount
  • ReservationExceededError - Raised when actual cost exceeds the reserved amount
  • ModelPricingNotFoundError - Raised when pricing data is not available for a model

Development

cd datasette-llm-accountant
pip install -e '.[dev]'
pytest

About

LLM accounting for Datasette

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors