Knowledge & analysis

Give a character a curated body of information and use conversation analysis to produce structured summaries or profile updates.

Knowledge ingestion

A typical workflow creates a knowledge collection, imports approved files or URLs, extracts text, optionally rewrites it for retrieval, and indexes the processed content. Ingestion and indexing can be asynchronous.

Track processing states before allowing a collection to influence production conversations. A successfully uploaded file is not necessarily ready for retrieval.

Retrieval context

Retrieve a bounded set of relevant passages for the current interaction. Preserve source identifiers so answers can be traced back to their context. Scope collections to the correct character and customer workspace.

Conversation analysis

Analysis can produce a conversation summary, sentiment feedback, custom metrics or proposed profile updates. Define the schema around the product decision the output will support.

{
  "summary": "A short factual summary of the exchange",
  "profile_updates": {
    "preferred_name": null,
    "favorite_topics": null
  }
}

This is an example application output shape, not an Amora endpoint response contract.

Budgets and retention

Agree ingestion limits, processing budgets, supported files and output retention before launch. Copy any result that your application must retain into your own approved storage rather than assuming temporary job output is permanent.