Zoost - workbench for Zoho Analytics

Zoost mirrors a Zoho Analytics workspace into plain local files, then draws you the map that Zoho Analytics never shows you. A workspace that has been alive for a few years is hundreds of views. Some are tables, some are query tables with SQL somebody wrote at 2am, most are reports built on top of reports. Zoho Analytics will tell you what a single view depends on if you ask it, one view at a time. It will not tell you the shape of the whole thing, it will not tell you what changed last month, and it will not tell you which of those hundreds of views nothing reads any more. WHAT IT DOES - A window of its own. Zoost opens from the toolbar icon, or with Alt+Shift+A, as an ordinary resizable window you can move to a second monitor and leave beside Zoho Analytics. It comes back the size and in the place you left it. The ER diagram, the lineage and the settings are views inside that one window, not extra browser windows. - Judge it before you give it access to anything. "+ Sample" writes a workspace of invented data into your working folder - tables with their columns and foreign keys, query tables with their SQL, reports and dashboards - so you can open the census, the ER diagram, the lineage and the exports without a Zoho Analytics tab and without an account. It is generated, never fetched, everything that would talk to the platform is disabled for it, and it is deleted like any other workspace. - Local mirror, your Git. One pull writes the whole workspace to disk: the view census, the column structure of every table and query table, the relations, and one .sql file per query table. Zoho Analytics keeps no history of anything: change a query table's SQL, a column or a relation and the previous state is gone. Commit the mirror and git diff answers "what changed in this workspace last month". - Every view, in one list. All of them with type, folder, owner, column count and dates, filtered by type and searched by name, folder or column name. Search a column name to find which tables carry it, before you go looking for where the data lives. - The structure of anything, including reports. Tables and query tables carry their own columns with Zoho's data types. A report or a pivot has none of its own - so Zoost follows what it is built on and shows you that structure instead, saying whose it is rather than pretending it is the report's. - Foreign keys on every column, and they are links. Each column shows what it points at and what points at it, both directions, because a foreign key is not symmetric. Click one and you are looking at the other table. Taken from the ER model Zoho Analytics itself draws, with the join written exactly as Zoho writes it. - ER diagram that stays readable. Tables as boxes, relations as arrows, the join on the arc. Focus one table, adjust the depth, walk the graph by clicking. Click an arc to isolate one relation and fade the rest. Boxes can be dragged: the automatic layout is a starting point, and what you arrange is what the PDF prints. Live sliders for spacing, spread, label gap and label size, and Save PDF for wall-size prints. - The SQL of your query tables. Read in the panel, written to disk one file each, and searched across all of them at once - the "in: SQL" switch beside the search box looks inside every query in the workspace, literally or as a regular expression (with named patterns you save and manage yourself), and shows, per query, the first matching line and how many times the pattern occurs. Each one also records which source tables it reads and which of their columns it actually involves. - What nothing depends on. Ask Zoho Analytics its own dependency question for every view, and see which ones nothing in the workspace reads. Stated as candidates, never as a verdict: a shared link, a scheduled export, an embedded report or an API consumer is invisible to that graph, and Zoost says so next to the number. - Design and data, told apart. Zoho Analytics tracks when a view's design last changed separately from when its data last refreshed. A report whose data refreshed nightly for two years but whose design nobody has touched since 2019 is a very different object from one redesigned last week, and the list shows both. - What Zoho put there, and what you built. The ER model flags system tables - the ones synced from a connected source rather than built by you. The view list marks them; the health view counts them. - Health audit. Views nothing depends on, tables in no relation at all, views with no description, items that could not be read. Counts and lists, with no thresholds and no score, and every figure stating what it does not cover. - Exports you can hand over. The whole workspace as one self-contained HTML file, or as Markdown shaped as context for an external AI tool - the Markdown carries the constraints of Zoho Analytics' own SQL dialect too, so an assistant that has never seen Zoho Analytics does not write queries that cannot run. A dialog decides section by section what the file may contain; the SQL is opt-in and flagged every time. - Optional AI assistant, bring your own key. Anthropic (Claude) or OpenAI (ChatGPT). With Anthropic it runs as an agent with read-only tools over your local mirror - it can read a structure, follow a foreign key, open a query's SQL, search columns, and say what depends on what. It is told what Zoho documents about query-table SQL, so what it writes is a draft you can paste rather than plausible SQL that will not run. - Your key, your machine, your choice. The API key is stored locally and sent only to the provider you pick. You can protect it with a passphrase you choose - it is then stored encrypted and asked for once per browser session. Off by default, because on a personal machine it buys little and on a shared one a lot, and the trade-off is explained where the choice is made. - Built for more than one workspace. The workspace does not have to be yours: one working folder holds a subfolder per work…