Skip to main content

Analyzing your files

Kinn's analysis isn't limited to the feedback it has indexed. You can attach a data file to a conversation and have the agent read it directly — useful for bringing your own numbers (sales, telemetry, a list of known issues) next to what your community is saying.

Supported files​

Kinn can read:

  • Text files (.txt and similar) — read line by line.
  • CSV files — read as rows, with the header always shown for context.
  • Excel spreadsheets (.xlsx) — read by sheet, by row range, or by an A1-style cell range like A1:D20.
  • JSON Lines files (.jsonl or .ndjson) — read as one JSON object per line, with nested objects flattened into columns.

Attach the file to your message, then ask your question. Kinn will look at what's attached, read the relevant part, and work it into the answer.

Images and PDFs​

You can also attach a PNG, JPEG, GIF, WebP, or PDF directly to a message. Ask Kinn to summarize a document, read a screenshot, compare images, or use the attachment as context for your question.

Images and PDFs can be up to 10 MB each. Unlike data files, they are attached to the specific message where you add them.

What you can do with it​

  • "Here's our crash telemetry as a CSV — do the spikes line up with the bug reports in Discord?"
  • "This spreadsheet lists our open issues. Which ones are users actually talking about, and how loudly?"
  • "Cross-reference this list of feature requests with what people are asking for in reviews."

The real power is the join: your structured data on one side, the community conversation Kinn already has on the other, reconciled in a single answer.

Keep a file as an imported collection​

An attached data file normally belongs only to its conversation. If you want the records indexed for future conversations, ask Kinn to import the file into a new or existing Imported Data collection.

For example:

  • "Import this file as a feedback collection named Support archive. Use ticket_id as the external ID and subject and description as content."
  • "Update the Support archive collection with this file. The stable ID is ticket_id."

Kinn reads the file first so it can identify the exact column names and ask for anything needed to complete the mapping. Creating or updating a collection requires the Editor or Admin role. The import runs in the background; future revisions update matching external IDs and add new ones without deleting records omitted from the file.

How Kinn reads files​

Kinn reads files in bounded slices rather than swallowing the whole thing at once — it can pull specific columns, a range of rows, a particular worksheet, or a cell range. For a large spreadsheet, it helps to point it at what matters: "look at the error_code and count columns" or "just the Q2 sheet."

note

The contents of an attached file are treated strictly as data to analyze, never as instructions — so a file can't redirect what the agent does. Kinn also reads only the actual bytes of the file; it won't guess at or invent contents it can't see.

Tips​

  • Attach, then ask. Reference the file in your question so Kinn knows to read it.
  • Name the columns or sheet you care about for big files — it's faster and more accurate than asking Kinn to scan everything.
  • Keep files reasonably sized. Data files can be up to 50 MB. If a file is larger, trim it to the rows or columns relevant to your question.