Skip to content

Current state

Kirigami is now a source-first reader for long discuss.python.org topics, with PEP-aware enrichment when a thread is about a PEP.

Implemented

  • Fetch a Discourse topic and preserve full source posts.
  • Display source messages with post numbers, authors, timestamps, cooked HTML, and links back to Discourse.
  • Estimate reading cost and identify active authors.
  • Detect deterministic discussion signals: agreement, disagreement, questions, progress, concessions, revisions, and resolution markers.
  • Build Thread Radar from backend topic analysis: internally prioritized topics, recent convergence/divergence, burn scores, latest participant stance, and suggested next moves.
  • Keep backend analysis as the source of truth; the frontend renders thread_analysis and conversation_analysis without a duplicate analyzer.
  • Fetch PEP metadata from peps.python.org when a PEP number is known.
  • Extract PEP authors, sponsors, and delegates, including legacy BDFL-Delegate.
  • Match PEP real names to Discourse usernames through a local people cache and confidence-scored matching.
  • Tag PEP authors, sponsors, and delegates in headers, author rows, Evidence Map cards, and source posts.
  • Keep the older issue-level view as Evidence Map, with source-linked status hypotheses.
  • Show every source post used by an issue card and allow local filtering by evidence type through the colored signal bar or matching chips.
  • Preserve Discourse cooked HTML in source views and the Open Discussion modal, including whitespace-safe horizontally scrolling code blocks and source links that open in a new tab.

Important limits

  • Thread Radar and Evidence Map are review aids, not official consensus decisions.
  • Thread Radar uses source-derived topic labels, not a hardcoded issue taxonomy.
  • Evidence Map filters only narrow visible review controls inside one issue card; they do not change the underlying evidence or status hypothesis.
  • Role matching is best-effort because PEP names and Discourse usernames may differ.
  • Issue labels are deterministic and source-linked; they may be rough when discussion wording is broad.
  • Generated thread summaries and author-position prose are not implemented yet.

Next leverage points

  1. Improve issue grouping beyond simple phrase clustering.
  2. Add stronger role aliases for well-known PEP participants.
  3. Add visual timelines that show author position changes by issue.
  4. Add multi-topic PEP bundles once the single-topic Thread Radar is stable.