Extending Intelligence Programs with Cost Tracking

When writing .lang programs that use LLMs:

When writing .lang programs that use LLMs:

Import the cost tracking API:

from intelligence.monitor.cost_tracker import CostTracker
from intelligence.monitor.budget_policy import BudgetPolicy

At the start of the program (each run), generate a run_id (UUID) and call BudgetPolicy().check_and_enforce(run_id).
After obtaining LLMResponse.usage, call CostTracker().add_cost(...).
Errors during LLM calls should still log cost if partial usage is available.

See docs/MONITORING_OPERATIONS.md for details.

When writing .lang programs that use LLMs:

When writing .lang programs that use LLMs:

Import the cost tracking API:

from intelligence.monitor.cost_tracker import CostTracker
from intelligence.monitor.budget_policy import BudgetPolicy

At the start of the program (each run), generate a run_id (UUID) and call BudgetPolicy().check_and_enforce(run_id).
After obtaining LLMResponse.usage, call CostTracker().add_cost(...).
Errors during LLM calls should still log cost if partial usage is available.

See docs/MONITORING_OPERATIONS.md for details.

When writing .lang programs that use LLMs:

When writing .lang programs that use LLMs:

Import the cost tracking API:

from intelligence.monitor.cost_tracker import CostTracker
from intelligence.monitor.budget_policy import BudgetPolicy

At the start of the program (each run), generate a run_id (UUID) and call BudgetPolicy().check_and_enforce(run_id).
After obtaining LLMResponse.usage, call CostTracker().add_cost(...).
Errors during LLM calls should still log cost if partial usage is available.

See docs/MONITORING_OPERATIONS.md for details.