What we are exploring
In systems with incomplete observations, probabilistic calculations, heuristics, and LLM explanations can become mixed together, making results difficult to audit and use safely.
Who it is for
Teams building AI/ML products with domain states, event sequences, multiple valid trajectories, and auditability requirements.
What is validated
PathWeaver 0.1.1 is packaged as a Python wheel, has passed strict quality gates, and runs in the Gadalnya production backend. It does not yet have a public PyPI release or production-ready library status.
04 · DATA AND PROCESS
Structural calculations stay separate from LLM interpretation
PATHWEAVER 0.1.1 · ALPHA/RC · FIRST PRODUCTION INTEGRATION
- DomainPack + observationsDomain states and observations
- Markov / belief / beam / ViterbiStructural trajectory calculations
- Validated LLM rolesProposer, judge, and generator
- Gadalnya Reading v2First production integration
- Typed Python 3.12 package
- Wheel and strict quality gates
- Live Gadalnya health confirms PathWeaver 0.1.1
- Caption
- Auditable flow: domain data → structural calculations → validated LLM roles → product integration.
- Provenance
- PathWeaver 0.1.1 and the live Gadalnya production-backend health endpoint, verified 13 August 2026.
- Caveat
- PathWeaver remains an alpha/RC without a public PyPI release or general production-ready library status. The models are not yet calibrated; scenarios are not forecasts, and path or LLM scores are not probabilities.
ALPHA FRAMEWORK · PRODUCTION INTEGRATION
See PathWeaver in a working product
The link opens Reading v2 in Gadalnya, where PathWeaver is used inside the product. It is not a standalone public release of the library.