jdsl | Behavior-Tree Agents

No codegen.

Determinism lives in the tree. The tree structure is the call structure. No object-literal DSLs parsing strings in the dark.

Typed by signatures.

The model enters only at predict leaves, guided cleanly by DSPy-style signatures like "message -> category".

Drive your tools.

A react leaf hands the model your native Python @tools and lets it pick, chain, and execute them until it answers.

Compile to small models.

Use the jdsl harness to capture frontier-model traces and compile proven paths into portable policies for smaller, frozen models.

from jdsl import root, seq, sel, act, check, predict

# The tree structure is the call structure.
skill = (
    root("Triage")
    .model("deepseek-chat")
    .do(seq(
        predict("message -> category"),
        sel(
            seq(check("category", "billing"), act(route_to_billing)),
            seq(check("category", "support"), act(route_to_support)),
            act(route_to_human)
        )
    ))
)