How-tos
Task-oriented guides — pick the thing you want to do, follow the steps. For what something is and how it works internally, see the Specs section.
This section is being filled in. Some pages are live (linked below); the rest are planned coverage and land as they are written. To suggest or contribute a guide, open an issue on
vance-docs.
Feature tours
Screenshot walkthroughs of the surfaces and building blocks you actually work in. Short, visual, “here’s how X works.”
- Cortex — chat, document and execute on one surface
- Canvas — think spatially: nodes, links and groups
- Workbook — notes and pages in a block editor
- Kanban & Issues — plan work and track it through
- Sheet — spreadsheets with functions
- Finance — structure money decisions
- Compose — run scripts (Shell, Python, LaTeX, an agent) right in a page
- Images — generate and refine pictures from a description
Use-case stories
The narrative path from a request to a result, across several features at once — what people actually get done with Vancetope.
- Analyse a dataset, step by step — pull public data, sample it, compute, chart, and ask an agent — five compose blocks in a row
- Stress-test an idea with a panel of judges — stand up a reusable council of AI personas, get one synthesised verdict, keep it as a document
- Size up a stock with live data — pull a stock’s prices off the web, chart them as candlesticks, and get a grounded assessment
- Turn research into a slideshow — research a topic, fetch an image per point, and get a presentable slides document
- Research a topic, end to end — deep-research with Marvin, reshape it, illustrate it, compile a document, export a PDF
- Coding — an agent does real, sustained engineering on a repo (planned)
- Research — a question becomes a sourced, verified synthesis (planned)
- Automation — work that runs on a timer or an event, no chatting (planned)
Planned guides
Getting productive
- Your first Vancetope session — install, log in, run a Recipe, read the result
- vance-foot CLI reference — commands, flags, slash-commands, a typical session
- Picking a model — what Vancetope asks of an LLM and how the recommended models compare under that load
Building on Vancetope
- Write your first Recipe — a
default-style YAML recipe end-to-end, with the LLM you already configured - Build a Project Kit — pack skills, recipes, tools and settings into a Git repo and install it into a project
- Add a server-side tool — register a
vance-sharedtool, wire it into a recipe, call it from a session - Anbind an MCP server — point Vancetope at an MCP endpoint, expose its tools through the recipe layer
Operations
- Configure an LLM provider — beyond the setup wizard: aliases, fallbacks, per-tier model bindings
- Add a webhook / event trigger — fire a recipe or workflow from an external HTTP call
- Schedule recurring work — Ursa Scheduler from a YAML doc to a running job
- Pin a release — set
IMAGE_TAGin the Docker stack so upgrades are explicit - Back up MongoDB — what to dump, how to restore, encryption-key implications
Table of contents
- Cortex — chat, document, execute
- vance-foot CLI
- Canvas — think spatially
- Workbook — notes and pages
- Kanban — plan work, track it
- Sheet — a spreadsheet with functions
- Finance — structure money decisions
- Compose — run scripts in a page
- Images — generate and refine
- Analyse a dataset, step by step
- Stress-test an idea with a panel of judges
- Size up a stock with live data
- Turn research into a slideshow
- Research a topic, end to end