Decisions & products

Python SDK reference for decision memory and product intelligence — log_decision, record_outcome, execute_decision, plus product_decision, product_optimize, and product_forecast.

These methods record decisions in decision memory and drive product-level intelligence (decision, optimize, retrieve, forecast). They are available on both AlgentaClient and AsyncAlgentaClient (await the async calls). The product methods take a request dict; the decision-memory methods take a decision id plus typed keyword arguments.

Method reference

Method	Purpose
product_decision(request)	Make a product-level decision (ProductDecisionResult). Posts to /v1/decision.
product_agent_run(request)	Run a product agent workflow (ProductAgentRunResult). Posts to /v1/agent/run.
product_optimize(request)	Optimize a product objective (ProductOptimizeResult). Posts to /v1/optimize.
product_retrieve(request)	Retrieve ranked product results for a query (ProductRetrieveResult). Posts to /v1/retrieve.
product_forecast(request)	Forecast product metrics over future periods (ProductForecastResult). Posts to /v1/forecast.
log_decision(request)	Record a decision in decision memory (DecisionLogResult). Posts to /v1/decisions.
list_decisions(*, page=None, limit=None, page_size=None, with_outcome_only=None)	List recorded decisions (DecisionListResult). Reads /v1/decisions.
get_decision(decision_id)	Fetch a single recorded decision (DecisionLogResult). Reads /v1/decisions/{decision_id}.
record_outcome(decision_id, *, actual_outcome, outcome_notes=None)	Attach the realized outcome to a decision (DecisionLogResult). Posts to /v1/decisions/{decision_id}/outcome.
execute_decision(decision_id, *, webhook_url, timeout_seconds=10.0, force=False, override_safety=False, metadata=None)	Execute a recorded decision via a webhook and return a receipt (ExecutionReceiptResult). Posts to /v1/decisions/{decision_id}/execute.
delete_decision(decision_id)	Delete a recorded decision. Sends DELETE /v1/decisions/{decision_id}.

Example: log a decision and record its outcome

import os
from decision_engine import AlgentaClient

client = AlgentaClient(api_key=os.environ["ALGENTA_API_KEY"])

# Record a decision.
logged = client.log_decision(
    {
        "chosen_action": "Scale checkout service",
        "expected_value": 42000,
        "context": "Checkout p95 latency at 820ms",
    }
)

# Later, attach what actually happened.
updated = client.record_outcome(
    logged.id,
    actual_outcome=39500,
    outcome_notes="Latency improved; revenue slightly below forecast.",
)

# Review decisions that already have an outcome.
history = client.list_decisions(with_outcome_only=True, limit=20)
for decision in history.decisions:
    print(decision.id, decision.chosen_action)

Example: forecast product metrics

import os
from decision_engine import AlgentaClient

client = AlgentaClient(api_key=os.environ["ALGENTA_API_KEY"])

forecast = client.product_forecast(
    {
        "product_id": "sku-1042",
        "metric": "units_sold",
        "horizon": 6,
    }
)
for period in forecast.periods:
    print(period.period, period.forecast)

Related pages

Log decisions & record outcomesDecision memorySimulations & queries