Conversational 0.7.19 is out: read the changelog

Solutions

Keep a changed decision inspectable

A project review moves from Tuesday to Friday. The earlier note should still be openable, and the answer should show the correction.

Starting material

Use the tiny demo file project-notes.md, or two notes you already have: one dated statement and one later correction. Import them into Conversational and wait until the source shows Ready.

Question

Ask: When is the project review?

Expected output

The answer should report Friday. Open Why this answer, or the equivalent evidence control, and find the 12 March correction. The 10 March Tuesday note remains visible as the earlier record.

Setup

Install Conversational, activate, connect the AI you already use, then import one source. The Conversational guide covers that sequence.

Evidence inspection

A cited source makes the answer easier to check. It is not a guarantee that every answer is correct. If the evidence panel is empty, the product is not ready to treat the answer as complete.

Limits

You connect your own AI. Nomic selects text evidence. Optional ColQwen2 and ColFlor select pages. They are not the answering model. Product Standard mode uses 8,192 tokens of evidence. Deep uses 24,000. R454 research used a 7,000-token evidence budget. That is not a product mode. Research rows that name Qwen 3.8 27B Q4 are evaluation readers, not a model shipped in the product. This page describes a first useful example, not a measured customer result.