ledgerwerk

pyjev

Reusable, inspectable Jev decision contracts for Python applications and automation.

pyjev is the Python application layer above TypeSafe’s official typesafe-sdk. It adds typed synchronous and native asynchronous clients, version-controlled .pyjev.toml decision contracts, offline validation and compilation, uncertainty-preserving results, credential ergonomics, and confidence-aware application policy.

Why pyjev?

  • Embed Jev in Python applications with dependency injection and typed results.

  • Declare stable named decisions and bundles once, then reuse them from Python, shell, CI, or an agent workflow.

  • Inspect and compile requests without credentials, client construction, or network access.

  • Preserve probabilities, confidence, usage, request IDs, and raw metadata.

  • Gate automation explicitly without consuming a low-confidence value by mistake.

  • Let the official SDK own transport, retries, API models, authentication, and API error semantics.

60-second workflow

cat > .pyjev.toml <<'EOF'
[pyjev]
schema = 1

[decision.ticket-route]
type = "choice"
question = "Which team should handle this support request?"

[decision.ticket-route.options]
billing = "Billing or refunds"
engineering = "Technical problem or product bug"
sales = "Purchasing or procurement"
EOF

pyjev decision validate
pyjev decision compile ticket-route --state "Stripe checkout fails"
echo "Stripe checkout fails" | pyjev decide ticket-route --json

The same contract is available to Python callers:

from pyjev import Jev

with Jev() as jev:
    result = jev.decide("ticket-route", state="Stripe checkout fails")

print(result.value, result.confidence, result.probabilities)

Direct dynamic decisions

For criteria that depend on runtime deterministic logic, use Jev/AsyncJev or the direct pyjev choice, pyjev noul, and pyjev score commands. The CLI is a useful shell and CI interface to pyjev’s application layer, not its only abstraction.

Documentation