L4Lesson 3Free

Load sources with labels

Name what went in

After this, you'll be able to load or connect a source set with labels, dates, owners, and a record of what changed.

Before you start

Complete Prepare the source packet.

The idea

Ingestion needs a receipt.

Files enter a knowledge tool with no record of what changed or what scope they cover.

ReceiptRead from goal to proof so Load sources with labels stays inspectable.

GoalAfter this, you'll be able to load or connect a source set with labels, dates,
InputsRecord what went in
Load sources with labelsThe receipt names the source set
ReviewIt lists loaded and excluded sources
ProofCreate one ingestion receipt with source set, loaded files, exclusions, owner, date,

Receipt rule: record what went in, when it went in, who owns it, and what changed from the last source set. Without a receipt, no one knows what the assistant is answering from.

Tool products differ in upload, connector, and project behavior. The stable habit is the same: label the source set and keep a record outside the chat.

Worked example: the "HR policy set, June 2026" receipt lists seven files, two retired drafts, owner, ingestion date, and the questions the set is meant to answer.

If you cannot name what went in, you cannot trust what comes out.

Ingestion needs a test set. Uploading or connecting sources is not proof that the assistant can use them. Create five to ten test questions before ingestion: easy lookup, cross-document answer, missing answer, stale-source trap, and review-required question.

After ingestion, ask those questions and record what happened. Did the assistant cite the right source? Did it admit a gap? Did it confuse an old draft with the current doc? This test set becomes your regression check whenever sources change.

Try it (12 min)

Watch out for

Uploading files without recording names.
Changing sources mid-chat with no note.
Leaving owner blank.
Letting scope grow after ingestion.

Paste this into Claude

Create an ingestion receipt.

Source set name: [name]
Sources loaded or connected: [list]
Retired or excluded sources: [list]
Owner: [name]
Date: [date]
Answer scope: [what this source set covers]

Return a concise ingestion receipt.

What a good response looks like

Source set: HR policy set, June 2026. Loaded: handbook, benefits PDF, FAQ, June update. Excluded: 2025 draft policy. Owner: Talia. Scope: onboarding, PTO, benefits, contractor basics.

What good looks like

The receipt names the source set
It lists loaded and excluded sources
It names owner and date
It states answer scope

When this breaks

Breaks when users ask about sources that were never loaded.
Breaks when a later answer cannot be tied to a source set version.

AI can help with this

Use your knowledge assistant to help you you can load or connect a source set with labels, dates, owners, and a record of what changed. Start with the exercise prompt and your real input. Ask for one draft, then check it against this proof: The receipt names the source set. Accept only the version you can verify yourself.

A receipt records source set name, loaded files, exclusions, owner, date, and answer scope.

You can now

You can name the source set

You can list what loaded
You can record exclusions
You can state scope

Key takeaways

An ingestion receipt makes the knowledge assistant auditable.

Record what went in.
Scope must be named.
Owners stay visible.
Receipts outlive chat.

Go deeper

Was this helpful?

Up nextAsk retrieval-safe questions

Build your first Project What Connectors are, and why they change everything Create the post-publish handoff

← Back to Chat With Your Knowledge

Create an ingestion receipt. Source set name: [name] Sources loaded or connected: [list] Retired or excluded sources: [list] Owner: [name] Date: [date] Answer scope: [what this source set covers] Return a concise ingestion receipt.