Safeheron Insight
1. Product Overview
Safeheron Insight is an AI data assistant built into the Web Console. Instead of exporting data manually and processing it in a spreadsheet, you simply ask in plain language — the AI queries your transactions, balances, fees, policies, and whitelists, answers with a written conclusion plus a data table, and supports one-click Excel export.
Safeheron Insight is strictly read-only. It never initiates transfers, approvals, or any write operation.
Key benefits:
Natural language queries: replace menu clicks and filter configuration with everyday language — just ask "how much did we send out yesterday?"
Zero install, ready to use: once enabled, open it right from the Web Console navigation bar — no plugin or client to install
Take the data with you: export any result table to Excel with one click; export runs asynchronously in the background and never interrupts your next question
Strictly permission-bound: you only ever see data you already have permission to view in Safeheron
What you can do:
Transaction aggregation (outflow totals, transaction counts, inflow vs. outflow, daily trends)
Transaction detail queries (filter and sort by time, wallet, coin, direction, status, amount)
Fee statistics (total, average, highest — split by chain / coin / wallet)
Current balances and point-in-time historical balances
Approval policy and whitelist queries
2. Prerequisites
Before you start, make sure the following are in place:
Safeheron team
An active, configured Safeheron account
Insight feature
Off by default; enable it under Web Console > Apps > Safeheron Lab
Who can enable
Team Creator only
Browser
A modern browser (latest Chrome / Edge / Safari)
3. Enable and Use
Open the assistant
Once enabled, a persistent Insight icon appears on the right side of the Web Console navigation bar. Click it to open the chat window.

Insight offers two window modes with identical functionality:
Floating window (default)
Compact; session history and the conversation view are toggled separately
Click the navigation bar icon
Full-screen mode
New Chat + session history list on the left, conversation area on the right
Click the expand button inside the floating window


4. Core Features in Action
4.1 Transaction Aggregation
Scenario: get a number or a trend to quickly grasp the overall transaction picture.
Example prompt:
Expected result: Insight returns last month's total transaction count and total amount (USD), broken down by direction into inflow and outflow.

Advanced prompt:
Expected result: Insight returns a table grouped by day with counts and amounts. When the result exceeds 10 rows, it shows the first 10 with the total count and a summary above the table.

4.2 Transaction Detail Queries
Scenario: review transactions one by one, investigate anomalies, or reconcile records.
Example prompt 1: largest outgoing transfers
Expected result: returns each transaction's created time, coin, amount, amount (USD), direction, source name, destination name, destination type, status, and creator. Status is shown in readable words ("Completed", "In Approval", "Failed") — no status codes to look up.

Example prompt 2: filter by conditions

Example prompt 3: add or remove table columns
Expected result: Insight adjusts the table columns on top of the original query and re-displays it. Columns available on request include: on-chain hash, fee, source address, destination address, memo, completion time, and transaction ID.

4.3 Fee Queries
Scenario: track and analyze your team's fee spending.
Example prompt 1: split by chain
Expected result: Insight returns the total fee (USD) and average fee per transaction (USD), grouped by chain.

4.4 Assets & Balance Queries
Scenario: understand your team's current asset position, or look back at holdings at a past point in time.
Example prompt 1: current total assets
Expected result: Insight returns the team's total asset value in USD, plus each asset's wallet name, coin, amount, amount (USD), and share of total team assets.

Example prompt 2: filter by balance threshold

Example prompt 3: point-in-time holdings
Expected result: Insight returns a snapshot of each wallet's USDT balance at the specified point in time.

4.5 Policy & Whitelist Queries
Scenario: understand your team's current approval rules and whitelist configuration.
Example prompt 1: view approval policies
Expected result: Insight returns each policy's name, initiator scope, source wallet scope, destination scope, applicable assets, and per-transaction limit.
Example prompt 2: view whitelists
Expected result: returns each whitelist entry's name, address, and network.
4.6 Tables and Excel Export
Structured query results are shown as tables, following these display rules:
10 rows or fewer: the table shows everything
More than 10 rows: the table shows only the first 10, with the total count and a summary above it (e.g. "50 transactions, total $2,300,000; first 10 below")
Very large match sets (result cap reached): Insight proactively tells you the data wasn't fully returned, and suggests narrowing the time range, adding filters, or exporting to Excel for the complete set
Every table has an Export to Excel button beneath it. Export is asynchronous: the button switches to "Exporting…", and on completion your browser downloads the file automatically with an "Export successful" toast. You can keep chatting the whole time.
5. Best Practices
5.1 Prompting Tips
Describe what you need in specific, concrete natural language:
Show me transactions
Show all failed USDT outgoing transfers in the last 7 days
What are the fees
Total and average fee per transaction last month, split by chain
Balance status
Which wallets have a balance below 1,000 USDT
5.2 Multi-Turn Follow-Ups
You don't have to restate your conditions each time. Ask on top of the previous result, and Insight automatically carries the earlier query conditions forward:
The first two follow-ups layer a filter and a sort onto the original query. The last one sums the full set — Insight switches to an aggregation query for an accurate result, rather than adding up the 10 rows on screen.
5.3 Adjusting Table Columns
Every query type comes with a set of default columns, and you can add or remove them anytime in natural language:
5.4 Export First for Large Data Sets
When a query matches hundreds or thousands of rows, the conversation only shows the first few. For row-by-row reconciliation or offline analysis, click Export to Excel for the complete data — faster than paging through the conversation.
6. Security Notes
Read-only queries: Insight only queries data. It never initiates transfers, approvals, or configuration changes.
Permission isolation: Insight strictly follows the existing permission system — you can only query data you have permission to access. When you query data outside your scope, Insight tells you the permission is missing and suggests contacting your team admin.
Verify Insight output: a fixed disclaimer sits at the bottom of the chat box "AI can make mistakes. Please double-check responses." Cross-check against the source data in the Web Console before any funds decision.
Available to all members: once enabled, all members of the team can use it — make sure your team understands the relevant security practices. Each member's data range stays bound by their own permissions.
7. FAQ
Last updated



