basemode-loom¶
basemode-loom is a persistent branching exploration tool for LLM text generation. It lets you generate multiple continuations from any point in a text, navigate the resulting tree, and build up complex branching narratives or explorations — all saved to a local SQLite database.
What it does¶
When you're exploring what an LLM might write, you often want to see several possibilities at once and follow the most interesting one further. basemode-loom makes that workflow first-class:
- Generate N parallel continuations from any node in a tree
- Navigate the tree interactively — parent, child, siblings
- Bookmark nodes, edit text, switch models mid-tree
- Persist everything locally; resume any tree at any time
- Analyze tree shape and generation quality with built-in stats
Interfaces¶
| Interface | Use case |
|---|---|
| CLI Reference | Scripting, quick generation, export/import |
| TUI Guide | Interactive exploration (basemode-loom view) |
| Python API | Embedding in your own tools |
| Web Server | Headless server + REST/WebSocket API |
Quick example¶
# Generate 3 continuations from a prompt
basemode-loom run "The last human on Earth sat alone in a room" -n 3 -m gpt-4o-mini
# Open interactive TUI to explore the tree
basemode-loom view
See Quickstart to get running in 5 minutes.