-
Notifications
You must be signed in to change notification settings - Fork 56
Interface Guide
English · 简体中文
Three columns: sessions on the left, the conversation in the middle, and a right pane you assemble yourself out of tabs.
The box you type in. Four shortcuts do most of the work:
| Type | To |
|---|---|
@ |
Reference an artifact |
# |
Reference another session |
/ |
Invoke a skill by name |
⌘K / Ctrl+K
|
Search |
⌘P / Ctrl+P
|
Action palette |
Enter sends, Shift+Enter makes a newline. You can drag the top edge to resize the composer, and drop files onto it to attach them.
The agent-options menu (the slider icon) collects the per-conversation controls in one place:
- Auto-review — a bounded automatic reviewer pass over the agent's work
- Reviewer model — which model does that review
- Memory — whether the project's memory is in play
- Specialist — run as a saved agent persona
- Compute — which registered servers this conversation may use (see Execution Contexts)
The right pane is a tab strip you build. The + button opens the full list; tabs can be dragged to reorder and closed individually. Four are open by default (Artifacts, Agents, Files, Environment).
Everything the conversation generated — figures, tables, CSVs, HTML — grouped and previewable inline. Images get a viewer with prev/next. There is a list and a grid view.
Every tool call in the conversation, with its input and output. This is where you check what actually ran rather than what the model said it ran.
The conversation's executed code cells, collected into notebook form.
A file browser over the project — and over any SSH context selected for the conversation, so you can look at a remote directory without a terminal.
Execution contexts, runtimes and Runs. Add and probe hosts, watch running jobs, open the in-memory environment table for a Python or R runtime.
Agent workflows and delegation configuration — the sub-agents available for
delegate_tasks, and their contracts.
Text you have starred out of the conversation (划线). Click one to jump back to where it came from.
Decisions, papers and data assets the agent recorded for this project, grouped by kind, with each item's links printed inline:
DECISIONS 1
Use DESeq2 over edgeR for the DE call
cites → Love et al. 2014, Moderated estimation…
applies to → counts.tsv
The agent writes these through the research_graph tool as it works. The panel
refetches when you open the tab.
A second conversation alongside the main one, for asking something without derailing the thread.
Starred items — code, figures, text — from across every project, reachable from the sidebar or the ⌘P palette.
What makes it more than a bookmark list:
- Immutable version history. Editing a starred snippet appends a new version; it never rewrites the old one. v1 is always the original.
- Re-run a version. Any version can be dropped into the composer as a "re-run this" request, carrying its item id and version number so the reply traces back to the exact version.
- Figures keep the code that generated them, with its own version history.
The left sidebar holds sessions, groupable into folders and pinnable. Sessions can be dragged between folders.
Projects are the top level — each has its own directory, database rows, Python/R environment and research graph. The projects landing screen shows every project with session counts and status, plus recent sessions across all of them.
Wisp can import conversations from other agents — Codex and Claude sessions are scanned, previewed and imported with de-duplication, so prior work in another tool becomes browsable here.
- Terminal — a real terminal dock, local or on a selected SSH host.
- Capabilities — what is installed and available: skills, MCP servers, memory.
- Settings — models and providers, environments, connections (per-tool allow/ask/deny), credentials, channels, appearance.
- Notifications — a completed background turn notifies you; only the window that owns the session arms click-to-open.
Light and dark themes are both supported; the light palette derives from the project's DNA-helix icon.
Wisp Science
English
- Home
- Getting Started
- Built-in Tools
- Skills
- Bioinformatics Databases
- Execution Contexts
- Interface Guide
简体中文