Conversation analysis reads an exchange between you and someone else and returns structured signal: tone, reciprocity, investment, consistency, and any safety-category flags.
Supported input: text you paste or type, or a single screenshot. Conversations are entered as text — you paste or type the exchange and label who said what. You can also upload one screenshot of a conversation: Perscivo reads the text out of it and hands it to the same review-and-label step, where you correct anything it misread before saving. Screenshot import is a shortcut into that review step, not a separate automatic pipeline. Bulk platform import and document/file import are not implemented.
Size. There is no message-count cap and no limit on how many conversations a single profile can have. Longer exchanges generally produce a stronger read because sufficiency rises with volume.
Editing. A conversation can be edited after it is created. Editing does not retroactively change an existing analysis — analyses are snapshots of the material as it stood when they ran.
Re-analysis is manual and versioned. Nothing re-runs automatically. When you trigger a new analysis, the previous one is preserved rather than overwritten, so you can see how the read changed as the conversation developed. Each analysis records the engine version that produced it.
What it does not do. It does not reply for you, it does not send anything anywhere, and it does not look the person up. Every conclusion comes from the text you provided.