Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.affzero.com/llms.txt

Use this file to discover all available pages before exploring further.

​
Overview

The instruction is the plain-English description you give the AI to explain what transformation you want. The AI uses your instruction along with the column names and sample rows from your sources to build a step-by-step plan. The clearer your instruction, the more accurate the plan.

​
Good instruction patterns

​
Filtering rows

Describe exactly which values to keep or exclude. Name the column and list the values. 
Keep only rows where "Sub ID 2" is one of: partner_a, partner_b, partner_c

Remove rows where "Status" is "Pending" or "Cancelled"

Keep only rows where Revenue is greater than 100

​
Grouping and aggregating

Name the grouping column and the aggregation you want. 
Group by "Affiliate Name" and sum "Revenue" and "Payout" for each group

Group by "Offer" and count the number of conversions

​
Joining two sources

Name both sources and describe the shared column. 
Join "Affise Stats" and "Sub IDs" on "Affiliate ID", keeping all columns from both

Match rows in "Report A" and "Report B" where "Sub ID" matches, then combine their columns

​
Sorting

Sort by Revenue descending

Group by Affiliate, sum Payout, then sort by Payout descending and keep the top 10

​
Selecting columns

Keep only the "Affiliate Name", "Conversions", and "Revenue" columns

Rename "aff_name" to "Affiliate Name" and "rev" to "Revenue"

​
Tips for better results

  • Name specific columns exactly as they appear in the header row — the AI matches by column name
  • For OR logic, list all values in one instruction: “where Sub ID is one of [a, b, c]” rather than three separate conditions
  • Be explicit about aggregation: say “sum Revenue” not “total the Revenue” if you want numeric addition
  • Describe the desired output shape: mention which columns you want in the result and in what order
  • Mention the source name when joining or filtering a specific file: “in the Affise export, keep rows where…”

​
What the AI cannot do

  • It cannot read values that are not in the sample rows or column headers during plan generation
  • It cannot look up external data
  • It cannot apply business logic it has no way to infer — be explicit