Limitations & Roadmap¶
Early development
Lathe is in early development. APIs, the YAML schema, and runtime behaviour are all subject to change between releases.
Current limitations¶
- Graphs must be acyclic. There's no support yet for cycles or retry loops within a pipeline. There's no control-flow or conditional node type yet to break out of a cycle, so a loop would just run forever.
- Only one tool kind.
LLMNodes can call tools (see Concepts § Tools), butHttpRequestis the only tool kind implemented so far, and every declared param is required — there's no optional-param support yet. - No multi-turn conversation support. Each
lathe runinvocation or/invokerequest is a single pass through the graph starting from fresh state. There's no built-in way to continue a conversation across calls, though anLLMNodewith tools can call them repeatedly within its own turn. - Only
graph_version: V1is supported. - Two providers only:
OpenAIandLMStudio.
None of these are permanent restrictions; they reflect where the project is today. Check GitHub issues for current status before relying on functionality not listed here.
Roadmap¶
Reflects the upstream project's own roadmap and may change without notice; treat it as direction, not a committed schedule.
- Web server: serve a pipeline as an HTTP API endpoint
- Tool call support: let
LLMNodes invoke tools/functions (currentlyHttpRequestonly) - Multi-turn conversation support
- Graph visualizer: render a pipeline's node/edge structure for inspection
- Additional node types (branching, etc.) and tool kinds
- Cyclic graph support for retry loops and iterative agents
- Visual UI: local graph builder and debugger for authoring and stepping through pipelines
Getting help / following development¶
- Source code: github.com/BBloggsbott/lathe
- Releases: github.com/BBloggsbott/lathe/releases
- Issues: github.com/BBloggsbott/lathe/issues