Research Timeline Rendering

The system SHALL render research results in a channel-specific section order. PDF reports SHALL follow the order: title, summary, references, cited paragraphs, research timeline. Email responses SHALL follow: title, summary, references, cited paragraphs, workflow signature. Chat responses SHALL follow: summary, references, cited paragraphs, then a collapsible timeline. The system SHALL NOT place the research timeline or cited paragraphs between the summary and references sections.

The system SHALL generate a contextual title of 3-8 words for PDF research reports using AI, rather than reusing the email subject or first sentence. The generated title SHALL be used as both the PDF heading and the sanitized filename. Chat sidebar titles SHALL remain derived from the first words of the user message.

The system SHALL render a compact workflow signature in email responses instead of the full research timeline. The signature SHALL include a completion indicator, processing metadata, a summary of completed tasks, and a pointer to the full timeline in the attached PDF.

The system SHALL render a references section as a numbered bibliography with sequential numbering matching inline citations. Each entry SHALL contain the source title in short form. The references section SHALL appear before the cited paragraphs section.

The system SHALL render a cited paragraphs section containing full source details for each reference: reference number, source title, source type indicator, text excerpt, and URL or file identifier. The cited paragraphs section SHALL appear after the references section.

The system SHALL display each cited insight as a paraphrased finding followed by citation numbers in brackets, with supporting quotes rendered in italics. The system SHALL NOT display internal confidence scores to end users.

The system SHALL deliver complete cited insight data to the chat interface, including the paraphrased finding, source reference, and supporting quotes. The system SHALL NOT expose internal confidence scores in the delivered data.