SELO
How-To Guide
Open the app →
SELO
How-To Guide
Open the app →
Selo turns a prompt into governed, production-ready dashboards and data models in minutes. No coding, no clicks — just describe what you need.
Watch the 5-min demoHow Selo works
Quickstart walkthroughs
Each quickstart form is built to be filled in under ONE minute — give it a try, you won’t regret it.
Connecting Selo
Selo authenticates to your Looker instance in two ways — pick one in Settings → Authentication Method. OAuth (PKCE, recommended): save the settings, click the Authorize Selo banner above the chat and sign in on the Looker page that opens; every action then runs with exactly your Looker permissions (a custom instance needs a one-time OAuth app registration by a Looker admin). API keys: in Looker, Admin → Users → your profile → API Keys, then paste the Client ID and Secret into Settings. Tip: in an Incognito window (or with third-party cookies blocked) the dashboard and explore previews show Looker’s login page — allow third-party cookies for your Looker instance and reload.
Features list
Everything Selo can do in Looker mode, grouped by surface. The Quickstart forms themselves are covered in the walkthroughs above.
The dashboard title on the preview toolbar is a searchable dropdown of every dashboard (and explore) you can open here.
Jump between the dashboards of a conversation — or any recent one — without leaving the panel or losing the chat.
Click the title → type to search → pick a dashboard. Recent dashboards and the ones created in this chat are listed first. When Linked mode is on, the chat now targets the dashboard you switched to.
Returns from a drill-down or a linked dashboard to where you came from, inside the preview.
Drill into a tile, follow a home-page card, then come back in one click — no reload, no lost filters.
Back appears once you have navigated inside the embedded dashboard (drill, link, tab card). Click it to step back one hop; the dropdown arrow next to it shows the full trail so you can jump straight to any earlier step.
Point Selo at exactly the tile(s) you want changed, instead of describing them.
Surgical edits: one tile changes, the rest of the dashboard stays untouched — no regeneration, no surprises.
1. Click Edit tile on the toolbar; the tiles become selectable. 2. Tick one or more tiles ("This tab only" or "All tabs" controls the scope of the picker). 3. Click Done, then type the change in chat: "make this a line chart", "rename it to Net revenue", "add a comparison to last year". Selo applies the change with an in-place element update and records it for Undo.
An automated review pass over the dashboard as it is actually rendered on your screen.
Catches what a query-level check cannot: truncated labels, overlapping legends, empty or erroring tiles, inconsistent formats, off-theme colours — and fixes them with you.
Click Quality → multi-select what to check (layout, labels, data quality, empty tiles, theme consistency, accessibility…) → Run. Selo screenshots the panel, reviews it and replies in chat with findings and proposed fixes; say "apply all" or pick individual ones. Works on the current tab; switch tabs and run again for the others.
Every edit Selo makes to a dashboard is journaled server-side and reversible, even after a reload.
Experiment freely: a wrong suggestion is one click away from being reverted, and redo brings it back.
The Undo and Redo buttons on the toolbar open a dropdown listing each edit (tile, what changed, when). Click an entry to revert up to that point. Redo re-applies reverted edits in order. The journal is per dashboard, so it is still there tomorrow.
Re-renders the embedded dashboard in place.
See an edit, a new theme or fresh data immediately without losing your place in the chat.
Click the circular-arrow button. Filters and the active tab are preserved where Looker allows it.
The "#id • Folder" chip under the dashboard title shows where the dashboard lives and lets you move it.
Put a freshly built dashboard where your team will find it, without opening Looker's folder browser.
Click the chip → pick a folder from the list → the move happens via the API and the chip updates. Only folders you can edit are offered.
Expand the panel over the whole app, or collapse it into a pill at the top right.
Present a dashboard on a call, or get the chat back — nothing is lost either way.
The expand icon hides the sidebar and chat (click it again to restore the previous split). The X collapses the panel into a pill; click the pill to reopen exactly where you were. Drag the divider to resize; double-click it to reset.
Design bespoke chart types beyond Looker's built-ins and manage instance themes from a guided flow.
The exact visual your stakeholder asked for — validated in a sandbox before it ever reaches a dashboard.
Account panel → Custom Visualizations → "+ Create Viz" (or the Dashboard Quickstart's "Design custom viz"). Describe the visual; Selo builds, validates and previews it, then registers it for use on dashboards. Themes: Account panel → Themes → "+ create", or "Create a new theme" inside the Dashboard form.
Ask questions in plain language and get answers with charts, straight from your Looker explores.
No SQL, no explore-hunting: Selo verifies real field names before every query, so the numbers come from your semantic layer — never invented.
Type a question in the chat bar (any supported language, or use the mic). Selo picks the right explore from your Context selection, runs the query, and answers with the data, a chart link and suggested follow-ups. If it is less than 80% sure what you mean, it asks one short clarifying question first.
A live, interactive explore opens beside the chat whenever a query benefits from visual exploration.
Full Looker interactivity — drill, filter, change the viz — while Selo stays aware of what is on your screen and can refine it.
Ask a data question and click the explore link, or say "show me that as a chart". Chat instructions like "colour the bars by category" update the preview in place. "Save this view" on an in-chat explore embed stores it as a Look.
Every explore opened in this chat gets a tab in the panel.
Compare several analyses side by side and return to any of them without re-running anything.
Drag tabs to reorder, click × to close one, use the collapse arrow to hide the strip, and "Clear all" to close every explore at once.
A dropdown of every state an explore went through (fields, filters, viz) in this tab.
Retrace an analysis step by step — the "how did I get here?" answer, one click away.
Open the history dropdown on an explore tab, select an earlier state and confirm "Go to"; the explore reloads to it. Explore Undo / Redo step through the same history one move at a time.
Controls whether your chat messages target what is on screen.
Linked: "add a filter on region" edits the open dashboard or explore. Unlinked: the chat is free-form and the panel is just a viewer.
Toggle the link icon on the toolbar. The dropdown next to it lists the linked explores and lets you link or unlink each one. The badge shows how many are linked.
Asks Selo to read the linked explore's result set and write the key takeaways.
From a table of numbers to three sentences a stakeholder can act on — grounded in the rows on screen, not a guess.
With an explore linked, click Compute insights. Selo runs the explore's query, summarizes trends, outliers and drivers, and suggests follow-up questions in chat.
Jumps from an explore to the LookML files that define it.
Close the loop between "why is this metric wrong?" and the line that defines it.
Open the Go to LookML dropdown on an explore tab → pick a view or model file → it opens in the LookML editor panel, ready for "fix this" requests in chat.
Every LookML file Selo writes appears as a card in the answer (path, project, line count).
Read what was generated without leaving the conversation; open the editor only when you want to change it.
Click a card to preview the file inline; click its expand icon to open that file in the LookML editor panel. Re-opening an older chat shows the cards again without opening any panel.
Project/file browsing, inline editing, validation, save and deploy — in a panel beside the chat.
Generated code streams into the editor, Validate runs Looker's validator, and every save is a versioned step you can undo.
The editor opens during Model/Prototype work or from a file card. Files dropdown: open any file (chips show the open ones). Toolbar: open the explores a file defines, Quality (AI review of the open file and CI checks on the project), Validate, Save (Ctrl/Cmd+S), refresh files, Undo/Redo across saved versions, and Deploy. Diagnostics under the editor link to the line; "Ask Selo to fix this" sends the error to chat.
Every open LookML file is a tab above the editor; the one you are reading is highlighted.
Work across a model, its views and a dashboard file without losing your place.
Open files from the Files dropdown (or let a build open them). Click a tab to switch, × to close it, drag to reorder. Tabs persist across reloads. The “link” icon on a tab makes that file the one Selo edits when you say “fix this” in chat.
Collapse views, explores, joins or dimension blocks to read a long file as an outline.
A 600-line view becomes a 20-line map; expand only the block you are changing.
Click the chevron in the editor gutter next to any block to fold or unfold it. Folding is per tab and remembered while the file stays open.
Every Save is a version; Undo/Redo step through them.
Experiment on a view knowing you can walk back any save — including ones Selo made.
Toolbar → Undo / Redo arrows (the dropdown lists versions with timestamps). Validate before Save (Ctrl/Cmd+S); Deploy publishes the whole project to production when you are done.
Rebuild an existing Tableau or Power BI asset in Looker from its file.
The data model, calculations and dashboard layout come across in one run instead of being re-authored by hand.
Prototype form → Objective “Migrate to Looker” → drop the .twb / .twbx / .pbit in the Tableau / Power BI File field. Selo lists the tables, relationships and sheets it found (Extracted Model); confirm the project and connection and what to migrate, then Build. The editor shows the generated LookML as it lands and the rebuilt dashboard opens at the end. Review differences in chat (“what did you not migrate?”).
Auto opens the right Quickstart form when your prompt calls for one, and says so in a banner.
You never have to remember which helper to use — and you can always decline it.
Leave Auto on (Quickstarts section of the sidebar; it locks on when no helper is on). When a form opens automatically the banner offers Dismiss (send the prompt as plain chat), "Not quite? Open instead" chips for the other helpers, and Quick answers to clarify in one click. Toggle a specific Quickstart off to stop it from auto-opening.
Choose which explores/models Selo may query; the active context shows as pills above the message box.
Precision and governance: Selo only answers from the data you pointed it at.
Click the Context button in the chat bar → multi-select explores → Done. Pills: click × to remove one, "Clear all context" to reset, and the chevron to collapse/expand the bar. The selector is disabled while only Model/Prototype helpers are on.
Drop CSV/Excel/JSON, images or BI files into the conversation.
Ground a build in real inputs: a spreadsheet becomes a dataset, a screenshot becomes a layout brief, a .twb becomes a migration.
Click the paperclip (or drag a file onto the chat). Large files prompt for how to handle them (sample, full upload). Images can be pasted directly; Selo reads them for layout, colours and numbers.
Dictate messages and answer in any of the supported languages.
Hands-free briefing and multilingual teams on the same chat.
Click the mic to dictate (click again to stop). The language selector next to it sets both transcription and reply language; it is remembered per user.
Add instructions while Selo is still working, or stop it.
No waiting for a wrong answer to finish: redirect mid-stream.
"Add to current response" appends a note the agent reads at its next step (e.g. "use monthly granularity"). "Stop generating" ends the turn immediately; what was produced so far stays.
AI suggestions under the box, and per-message actions.
Faster next steps, and a feedback loop that improves Selo for your instance.
"Show / Hide AI suggestions" toggles prompt ideas; "Edit before sending" lets you tweak one. On messages: copy the prompt, 👍 / 👎 rate the answer, expand the thinking / tool trace, open Recommended Actions and Suggested Next Steps. A budget hold notice appears if AI consumption is paused for the period.
Looker, Databases and Knowledge Catalog are three modes of the same assistant; the active one shows as a pill next to the sidebar menu button.
One interface across the stack — and the Quickstarts, context picker and these guides follow the mode you pick.
Click the mode pill (or open Settings) and pick a mode. In Knowledge Catalog mode you can also attach a BigQuery or Looker backend so Selo can act on the cataloged assets.
Full-width search, filters (activity, objective, industry, asset), grouping and AI hover-recaps for every chat.
Pinpoint the one conversation about a specific asset in seconds.
Use the search bar above RECENT; the sliders icon opens filters — the Asset tab filters chats by the dashboards, explores, datasets or catalog entries they used. Group by activity or industry, collapse/expand groups, and hover any chat for its AI recap.
Every chat row has a ⋮ menu for housekeeping.
Pinned chats stay on top in the order you choose; archived ones leave the list but keep their history.
⋮ → Pin/Unpin, Rename, Archive, Move up/down (pinned order), or Bulk edit to act on many chats at once. New Chat starts a fresh conversation. Files opens the attachments browser.
Turn any chat into a recurring email: scheduled reports, threshold alerts, anomaly checks, or an AI-written digest.
The insight you asked for once keeps arriving on a cadence — with real data summaries, not just static screenshots.
Chat ⋮ menu → Monitor. Pick the tab (Schedule / Alert / Digest), confirm the suggested data context and visuals from the chat, set the cadence and recipients, and save. "Run once now" tests it immediately. Manage them later in Account → Schedules (edit, run, pause, delete).
Share a chat with teammates, fork a copy, print to PDF, or send an AI summary by email.
The analysis travels: colleagues open a live copy of the conversation instead of screenshots.
Chat ⋮ menu → Share (owner; recipients see a banner with their role), Fork a copy (your own editable duplicate), Print (transcript export), or Summary → copy / email.
Persona, language, and your activity: schedules, AI consumption, agent instructions, custom visualizations, themes.
One place for everything that is yours — with projections so AI spend is never a surprise.
Click your avatar. Preferences: persona (how Selo addresses you and what it assumes), language. Your activity: Schedules (+ create), AI consumption chart (projection; admins see instance scope), Agent instructions (+ Create Agent, revoke/restore), Custom Visualizations (+ Create Viz), Themes (+ create).
Model, mode, Looker connection and appearance.
Switch models per task, fix a broken authorization in one click, and pick the look that suits you.
Settings → Model (which Gemini model answers), Mode, Looker instance, authentication method & OAuth client (re-authorize from the banner above the chat), default connection, Dark mode. Admin-managed fields are shown read-only.
This guide, quick tips, bug/idea reports with history, and a support chat.
Feedback lands with the team that builds Selo, with the context needed to act on it.
Sidebar → Help & Feedback: Help tab (tips + this guide), Submit (bug / idea / question / praise with severity and mood), My Reports (status of what you sent), Admin (for admins). The floating support widget answers how-to questions (curated or AI mode) and keeps its own history.
How Selo works
Quickstart walkthroughs
Each quickstart form is built to be filled in under ONE minute — give it a try, you won’t regret it.
Connecting Selo
In Settings, pick the Databases mode and enter the Google Cloud project ID where your BigQuery data lives. Authenticate with a service-account key (JSON) with BigQuery read access, or the Google sign-in option the settings form offers. Then open the Context picker in the chat bar and tick the datasets (or tables) Selo may query.
Features list
Everything Selo can do in Databases mode, grouped by surface. The Quickstart forms themselves are covered in the walkthroughs above.
Ask questions in plain language; Selo writes and runs the SQL against the datasets in your Context.
No SQL to write, no schema-hunting: Selo reads the real table schemas before every query, so column names are never invented.
Type a question in the chat bar (any supported language, or use the mic). Selo picks the tables from your Context selection, writes the query, runs it and answers with the result table, a chart and suggested follow-ups. Ask "show me the SQL" to see the query.
Every query answer renders its rows and a chart inline in the chat.
See the numbers and the picture together, then refine in the same thread.
Ask follow-ups ("break it down by region", "as a line chart", "only the top 10"). Shared chats redact BigQuery result rows for recipients who do not own the chat.
Selo designs the schema, creates the tables and views in BigQuery and loads dummy data — from an industry template or your own brief.
A working dataset in minutes: schema, tables, views and sample data designed together, partitioned and clustered the way BigQuery wants them.
Model Quickstart → "Create a new dataset" → pick a template or describe the use case, optionally name the dataset, and tick what to Include (Tables & schema, Views, Dummy data, Partitioning & clustering, Materialized views, BQML model). Run AI Screening to have the brief checked first; the chat reports everything created.
Point Selo at a dataset and a focus area; it proposes and applies the improvements.
Schema cleanup, performance and documentation work that usually stays on the backlog, done conversationally.
Model Quickstart → "Refine an existing dataset" → pick the dataset and a Focus On (New tables or views, Schema cleanup, Partitioning & clustering, Performance, Data quality checks, Documentation & descriptions) — or just describe the change in chat.
Beyond the form: create derived tables, forecast with BQML and run data health checks straight from the conversation.
The warehouse-engineering loop stays in one thread — build, check and iterate without leaving the chat.
Ask in chat: "create a view of monthly revenue by store", "train a BQML model to forecast orders", or "run a health check on the sales dataset" (Freshness, Nulls, Duplicates, Volume anomalies).
Plan and prototype a database-to-database migration onto BigQuery (or another GCP target).
A representative schema, the mapping and the first migrated tables — before the migration project starts.
Prototype form → Objective "Migrate to GCP" → Current Database (Oracle, SQL Server, …), Target on GCP, paste the Current Schema (or leave blank and Selo proposes one), tick what to migrate, then Migrate to GCP.
Auto opens the right Quickstart form when your prompt calls for one, and says so in a banner.
You never have to remember which helper to use — and you can always decline it.
Leave Auto on (Quickstarts section of the sidebar; it locks on when no helper is on). When a form opens automatically the banner offers Dismiss (send the prompt as plain chat), "Not quite? Open instead" chips for the other helpers, and Quick answers to clarify in one click. Toggle a specific Quickstart off to stop it from auto-opening.
Choose which project datasets (or tables) Selo may query; the active context shows as pills above the message box.
Precision and governance: Selo only answers from the data you pointed it at.
Click the Context button in the chat bar → tick datasets (or "All datasets") and, inside one, specific tables → Done. Pills: click × to remove one, "Clear all context" to reset, and the chevron to collapse/expand the bar.
Drop CSV/Excel/JSON files or images into the conversation.
Ground a build in real inputs: a spreadsheet becomes a table, a screenshot becomes a brief.
Click the paperclip (or drag a file onto the chat). Large files prompt for how to handle them (sample, full upload). Images can be pasted directly; Selo reads them for layout, colours and numbers.
Dictate messages and answer in any of the supported languages.
Hands-free briefing and multilingual teams on the same chat.
Click the mic to dictate (click again to stop). The language selector next to it sets both transcription and reply language; it is remembered per user.
Add instructions while Selo is still working, or stop it.
No waiting for a wrong answer to finish: redirect mid-stream.
"Add to current response" appends a note the agent reads at its next step (e.g. "use monthly granularity"). "Stop generating" ends the turn immediately; what was produced so far stays.
AI suggestions under the box, and per-message actions.
Faster next steps, and a feedback loop that improves Selo for your instance.
"Show / Hide AI suggestions" toggles prompt ideas; "Edit before sending" lets you tweak one. On messages: copy the prompt, 👍 / 👎 rate the answer, expand the thinking / tool trace, open Recommended Actions and Suggested Next Steps. A budget hold notice appears if AI consumption is paused for the period.
Looker, Databases and Knowledge Catalog are three modes of the same assistant; the active one shows as a pill next to the sidebar menu button.
One interface across the stack — and the Quickstarts, context picker and these guides follow the mode you pick.
Click the mode pill (or open Settings) and pick a mode. In Knowledge Catalog mode you can also attach a BigQuery or Looker backend so Selo can act on the cataloged assets.
Full-width search, filters (activity, objective, industry, asset), grouping and AI hover-recaps for every chat.
Pinpoint the one conversation about a specific asset in seconds.
Use the search bar above RECENT; the sliders icon opens filters — the Asset tab filters chats by the dashboards, explores, datasets or catalog entries they used. Group by activity or industry, collapse/expand groups, and hover any chat for its AI recap.
Every chat row has a ⋮ menu for housekeeping.
Pinned chats stay on top in the order you choose; archived ones leave the list but keep their history.
⋮ → Pin/Unpin, Rename, Archive, Move up/down (pinned order), or Bulk edit to act on many chats at once. New Chat starts a fresh conversation. Files opens the attachments browser.
Turn any chat into a recurring email: scheduled reports, threshold alerts, anomaly checks, or an AI-written digest.
The insight you asked for once keeps arriving on a cadence — with real data summaries, not just static screenshots.
Chat ⋮ menu → Monitor. Pick the tab (Schedule / Alert / Digest), confirm the suggested data context and visuals from the chat, set the cadence and recipients, and save. "Run once now" tests it immediately. Manage them later in Account → Schedules (edit, run, pause, delete).
Share a chat with teammates, fork a copy, print to PDF, or send an AI summary by email.
The analysis travels: colleagues open a live copy of the conversation instead of screenshots.
Chat ⋮ menu → Share (owner; recipients see a banner with their role), Fork a copy (your own editable duplicate), Print (transcript export), or Summary → copy / email.
Persona, language, and your activity: schedules, AI consumption, agent instructions.
One place for everything that is yours — with projections so AI spend is never a surprise.
Click your avatar. Preferences: persona (how Selo addresses you and what it assumes), language. Your activity: Schedules (+ create), AI consumption chart (projection; admins see instance scope), Agent instructions (+ Create Agent, revoke/restore).
Model, mode, Google Cloud project / credentials and appearance.
Switch models per task, fix a broken authorization in one click, and pick the look that suits you.
Settings → Model (which Gemini model answers), Mode, Google Cloud project and service-account / sign-in credentials, Dark mode. Admin-managed fields are shown read-only.
This guide, quick tips, bug/idea reports with history, and a support chat.
Feedback lands with the team that builds Selo, with the context needed to act on it.
Sidebar → Help & Feedback: Help tab (tips + this guide), Submit (bug / idea / question / praise with severity and mood), My Reports (status of what you sent), Admin (for admins). The floating support widget answers how-to questions (curated or AI mode) and keeps its own history.
How Selo works
Quickstart walkthroughs
Each quickstart form is built to be filled in under ONE minute — give it a try, you won’t regret it.
Connecting Selo
In Settings, pick the Knowledge Catalog mode and enter the Google Cloud project whose Dataplex catalog Selo should read. Choose how Selo authenticates (service-account key or application default credentials) and, optionally, attach a BigQuery or Looker backend so Selo can act on the cataloged assets, not only describe them. Then tick the catalog entries to ground on in the Context picker.
Features list
Everything Selo can do in Knowledge Catalog mode, grouped by surface. The Quickstart forms themselves are covered in the walkthroughs above.
Ask where data lives, what an asset means, who owns it and how fresh it is.
Answers grounded in the catalog metadata — descriptions, owners, labels, freshness — not guesses.
Type a question in the chat bar ("which table holds daily order revenue, and how fresh is it?"). Selo searches the entries in your Context (or the whole catalog) and answers with the matching assets and their metadata. With a BigQuery or Looker backend attached, it can query the asset too.
An audit of the selected catalog entries for the metadata gaps that hurt discoverability.
Know which assets lack descriptions, owners, labels or recent updates before your users find out.
Explore form → Objective "Metadata health check" → pick the entries in Data Context → Checks to Run: Missing descriptions, Ownership, Labels & tags, Freshness → Run Metadata Check. Findings come back per asset; hand them to Context engineering to fix.
Selo proposes descriptions, labels, ownership notes and glossary links for many assets at once.
Better metadata makes every other answer better — for Selo and for everyone searching the catalog.
Context engineering Quickstart → pick the Assets and a Template (Catalog-wide sweep, Domain spotlight, Onboarding pack, Governance audit or Custom), tick what to Include (Descriptions, Labels & tags, Ownership, PII flags, Business glossary links, Freshness documentation) and run — the proposals arrive in chat.
Approved proposals can be applied to the catalog directly instead of staying ideas.
The audit-to-fix loop closes in one place: findings become applied metadata, not a backlog document.
Set "Apply as" to Write back to KC once the proposals look right — or keep "Ideas in chat" to review first. "Refine an asset's metadata" does the same for a single entry.
Auto opens the right Quickstart form when your prompt calls for one, and says so in a banner.
You never have to remember which helper to use — and you can always decline it.
Leave Auto on (Quickstarts section of the sidebar; it locks on when no helper is on). When a form opens automatically the banner offers Dismiss (send the prompt as plain chat), "Not quite? Open instead" chips for the other helpers, and Quick answers to clarify in one click. Toggle a specific Quickstart off to stop it from auto-opening.
Choose which catalog entries Selo grounds on; the active context shows as pills above the message box.
Precision and governance: Selo only answers from the data you pointed it at.
Click the Context button in the chat bar → search the catalog and tick the entries (tables, datasets, explores, dashboards) → Done. Pills: click × to remove one, "Clear all context" to reset, and the chevron to collapse/expand the bar.
Drop CSV/Excel/JSON files or images into the conversation.
Ground a build in real inputs: a spreadsheet becomes a table, a screenshot becomes a brief.
Click the paperclip (or drag a file onto the chat). Large files prompt for how to handle them (sample, full upload). Images can be pasted directly; Selo reads them for layout, colours and numbers.
Dictate messages and answer in any of the supported languages.
Hands-free briefing and multilingual teams on the same chat.
Click the mic to dictate (click again to stop). The language selector next to it sets both transcription and reply language; it is remembered per user.
Add instructions while Selo is still working, or stop it.
No waiting for a wrong answer to finish: redirect mid-stream.
"Add to current response" appends a note the agent reads at its next step (e.g. "use monthly granularity"). "Stop generating" ends the turn immediately; what was produced so far stays.
AI suggestions under the box, and per-message actions.
Faster next steps, and a feedback loop that improves Selo for your instance.
"Show / Hide AI suggestions" toggles prompt ideas; "Edit before sending" lets you tweak one. On messages: copy the prompt, 👍 / 👎 rate the answer, expand the thinking / tool trace, open Recommended Actions and Suggested Next Steps. A budget hold notice appears if AI consumption is paused for the period.
Looker, Databases and Knowledge Catalog are three modes of the same assistant; the active one shows as a pill next to the sidebar menu button.
One interface across the stack — and the Quickstarts, context picker and these guides follow the mode you pick.
Click the mode pill (or open Settings) and pick a mode. In Knowledge Catalog mode you can also attach a BigQuery or Looker backend so Selo can act on the cataloged assets.
Full-width search, filters (activity, objective, industry, asset), grouping and AI hover-recaps for every chat.
Pinpoint the one conversation about a specific asset in seconds.
Use the search bar above RECENT; the sliders icon opens filters — the Asset tab filters chats by the dashboards, explores, datasets or catalog entries they used. Group by activity or industry, collapse/expand groups, and hover any chat for its AI recap.
Every chat row has a ⋮ menu for housekeeping.
Pinned chats stay on top in the order you choose; archived ones leave the list but keep their history.
⋮ → Pin/Unpin, Rename, Archive, Move up/down (pinned order), or Bulk edit to act on many chats at once. New Chat starts a fresh conversation. Files opens the attachments browser.
Turn any chat into a recurring email: scheduled reports, threshold alerts, anomaly checks, or an AI-written digest.
The insight you asked for once keeps arriving on a cadence — with real data summaries, not just static screenshots.
Chat ⋮ menu → Monitor. Pick the tab (Schedule / Alert / Digest), confirm the suggested data context and visuals from the chat, set the cadence and recipients, and save. "Run once now" tests it immediately. Manage them later in Account → Schedules (edit, run, pause, delete).
Share a chat with teammates, fork a copy, print to PDF, or send an AI summary by email.
The analysis travels: colleagues open a live copy of the conversation instead of screenshots.
Chat ⋮ menu → Share (owner; recipients see a banner with their role), Fork a copy (your own editable duplicate), Print (transcript export), or Summary → copy / email.
Persona, language, and your activity: schedules, AI consumption, agent instructions.
One place for everything that is yours — with projections so AI spend is never a surprise.
Click your avatar. Preferences: persona (how Selo addresses you and what it assumes), language. Your activity: Schedules (+ create), AI consumption chart (projection; admins see instance scope), Agent instructions (+ Create Agent, revoke/restore).
Model, mode, Google Cloud project / credentials and appearance.
Switch models per task, fix a broken authorization in one click, and pick the look that suits you.
Settings → Model (which Gemini model answers), Mode, Google Cloud project and service-account / sign-in credentials, Dark mode. Admin-managed fields are shown read-only.
This guide, quick tips, bug/idea reports with history, and a support chat.
Feedback lands with the team that builds Selo, with the context needed to act on it.
Sidebar → Help & Feedback: Help tab (tips + this guide), Submit (bug / idea / question / praise with severity and mood), My Reports (status of what you sent), Admin (for admins). The floating support widget answers how-to questions (curated or AI mode) and keeps its own history.