mirror of
https://github.com/hanzoai/flow.git
synced 2026-10-17 01:29:18 +00:00
zeekay
d650dd60f0
Docs corpus (Langflow->Hanzo Flow, upstream URLs langflow-ai/langflow->hanzoai/ flow, docs.langflow.org->docs.hanzo.ai), starter-project template help-URLs, CI workflows, and test files. Remaining 'langflow' repo-wide is now dominated by LANGFLOW_* env-var names + langflow_ settings fields — the config-boundary layer deliberately preserved pending a coordinated FLOW_/HANZO_ alias-with-fallback (renaming them blind breaks every deployment).
26 lines
1.4 KiB
Go
26 lines
1.4 KiB
Go
// Package flow is the ONE step-DAG engine the whole platform composes on.
|
|
//
|
|
// There is exactly one scheduler. Everything — LLM chains, media/diffusion
|
|
// pipelines, IFTTT automations, agent tool-graphs — is this engine applied to
|
|
// two parameters:
|
|
//
|
|
// Value type — what travels on an edge (a Message, a Tensor, connector I/O).
|
|
// The engine NEVER inspects it; a Step interprets it per its domain.
|
|
// Mode — execution locality/durability (InProcess | GPUResident | Durable).
|
|
// A PARAMETER, never baked in: a media flow keeps latents pinned in
|
|
// VRAM and serializes nothing; an automation checkpoints every step
|
|
// on the durable backend. Same scheduler, different Backend.
|
|
//
|
|
// So a product is just flow curried:
|
|
//
|
|
// studio = Run(GPUResident, Graph-of-Tensor-steps) // comfy nodes
|
|
// automations = Run(Durable, Graph-of-ConnectorIO) // triggers -> actions
|
|
// llmFlow = Run(InProcess, Graph-of-Message-steps) // flow (lfx) nodes
|
|
//
|
|
// A domain supplies Steps (typed input ports -> typed output ports + a Run fn) and,
|
|
// if it has special execution needs, a Backend for a Mode. The engine only wires
|
|
// matching ports and schedules them. This package is stdlib-only and holds no
|
|
// domain knowledge — it is the seam every node-pack and every execution backend
|
|
// plugs into.
|
|
package flow
|