LangState01
Give uncertainty a structure.
Keep candidate values, confidence, and the latest inference together. Inspect what the system believes before deciding what to accept.
Explore interpretive state ↗ Open source · Source-only alpha ↗Language is fluid.
Language is fluid.
Your state should
be explicit.
A Python framework for the space between what a user says and what your application knows.
Represent uncertainty. Resolve it on your terms.
Build around state you can inspect.
Start from a pinned source checkout. Installation guide ↗
FROM LANGUAGE TO STATELS / 001
interpretation
registrant.name2 candidatesAda Lovelace0.96
Ada0.42
LATEST INFERENCE
The user supplied a full name after a short name.
Possibilities first. Decisions when you’re ready.
BUILT ON EXPLICIT PRIMITIVES
Python OpenAPI schemas Pydantic models Field-level state
01 / THE APPROACH
A conversation is context.
State is something you can work with.
Natural language is partial, ambiguous, and sometimes contradictory. Keep those possibilities visible before they become application decisions.
02
Make resolved state explicit.
Store authoritative values at stable field paths. Give downstream validation and business logic a clear representation to work with.
Explore canonical state ↗03
Evaluate the state change.
Compare expected and actual interpretive states field by field. Test the effect of a mutator beyond the text of its response.
Explore evaluation ↗02 / THE ARCHITECTURE
Separate the belief
from the decision.
LangState gives interpretation and authoritative data different homes. Your application owns the policy that connects them.
Understand the two-state model ↗01{…}
Interpretive state
Candidates · Confidence · Latest inference
Your resolution policyYOU DEFINE
02✓
Canonical state
Resolved values · Stable field paths
Projector and orchestration contracts are abstract.
Their implementations belong to your application.
03 / START WITH THE PRIMITIVES
Small building blocks.
Room for your architecture.
Read an OpenAPI schema, create canonical state, and derive interpretive state. Start with concrete APIs exercised by the documented source tests.
Walk through the quickstart →Early by design. Clear about the boundaries.Source-only alpha. No supported PyPI release; the PyPI name belongs to an unrelated project. See implementation status ↗
quickstart.pyPYTHON
# Run from the documented source checkout
from core.state.factory.state_factory import StateFactory
# schema comes from OpenAPIReader
factory = StateFactory()
canonical = factory.create_canonical_state(schema)
canonical.set_field(
"registrant.name", "Ada Lovelace"
)
interpretive = factory.create_interpretive_state(
canonical
)
# A candidate you can inspect
best = interpretive.get_best_value(
"registrant.name"
) Concrete state APIs Import reference ↗
MAKE THE IMPLICIT EXPLICIT