> ## Documentation Index
> Fetch the complete documentation index at: https://docs.langstate.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Implementation status

> Current release, packaging, test, and API readiness at the pinned source commit.

This site is pinned to [`langstate/langstate@df40bc49bc8e3a51b2bb695b30ff93ef81b2cc4b`](https://github.com/langstate/langstate/commit/df40bc49bc8e3a51b2bb695b30ff93ef81b2cc4b), with source metadata version **0.1.0 source alpha**.

## Release status

* **Source-only alpha:** the repository has no project tags and no GitHub releases at the documented baseline.
* **No supported PyPI install:** the PyPI name `langstate` is owned by an unrelated project.
* **Source imports:** working examples use `core.*` while the current directory is `packages/langstate`.
* **Packaging incomplete:** the Poetry project metadata does not yet produce the intended public import surface.

## Verified workflow

```bash theme={null}
cd packages/langstate
poetry install --no-root --with dev
poetry run python -m pytest -q
```

At `df40bc49bc8e3a51b2bb695b30ff93ef81b2cc4b`, this workflow passes **628 tests**. This count describes the pinned baseline; it is not a compatibility guarantee for later commits.

## Capability matrix

| Capability                               | Status             | Notes                                                                           |
| ---------------------------------------- | ------------------ | ------------------------------------------------------------------------------- |
| `CanonicalState` and `InterpretiveState` | Concrete           | Path-addressed state with DAH-backed storage                                    |
| `OpenAPIReader`                          | Concrete           | Reads local YAML or JSON and selects a named root schema                        |
| `StateFactory`                           | Concrete           | Creates canonical state and derives interpretive state                          |
| `LLMMutator`                             | Concrete adapter   | Requires a caller-provided `BaseLLMClient` implementation                       |
| `LLMEvaluator`                           | Concrete           | Executes a mutator and compares expected and actual interpretive states         |
| In-memory repository and snapshots       | Concrete           | Separate utilities; not wired into `LangState` automatically                    |
| Canonical and UI projectors              | Abstract contracts | Applications provide implementations                                            |
| `LangState`                              | Abstract contract  | `initialize`, `invoke`, state accessors, and `reset` remain application-defined |
| Installed `langstate.*` imports          | Not release-ready  | Use the documented checkout and `core.*` imports                                |

## Contract corrections

The source defines:

* `ActionResultData` for a completed orchestrator flow
* `get_current_state()` for interpretive state access
* `InterpretiveStateSchema` and `CanonicalStateSchema` as transport models
* one latest `inference` per interpretive field, plus multiple candidate `values`

Older names such as `StateResultData`, `get_state()`, or a generic `StateSchema` are not part of the pinned implementation.

## Adopter guidance

Evaluate the concrete state and schema primitives first. If you need a full agent loop, stable installed imports, a prebuilt UI projector, or a production persistence integration, plan to implement and maintain those extension points in your application until the project publishes a release-ready package.

For the intended generated developer workflow and end-to-end architecture, see [Library target](/design-notes/library-target) and [System design](/design-notes/system-design). Both are explicitly labeled design targets rather than current package behavior.