harnyGlossary
Fact extraction
Fact extraction is the step where an AI system turns raw text, such as a chat message or a document, into short, standalone statements it can store and search later. Each statement captures one fact, like a preference, a deadline or a budget.
What fact extraction is
Fact extraction reads text and pulls out the claims that matter. A long conversation becomes a set of short statements, each one readable on its own.
The output is usually plain sentences with a date or source attached, so a later step can find them and check them.
Why it matters for agents
- Agents cannot reread every past conversation. Short statements fit into a prompt.
- Extraction can drop or distort details. Keeping a link back to the source helps you check what was said.
How it works in Harny
- You send text to Harny as a data point with
POST /v1/memory/ingest. Harny builds memories on that data point. - Each recall result lists its
sources, withsourceId,occurredAtandlabel, so you can trace a result back to the data point it came from.
Does fact extraction keep the original text?
Each recall result carries sources with the sourceId of the data point it came from, so the source stays traceable.
Keep reading
Go on. Get Harny.
The first 500 monthly-active end users are free for good. Create an app key and make your first call in minutes.