VisualEther Release Notes

The headline features in each VisualEther release line, newest first.

9.1 — Annotate the flow

The AI-annotation release: extract a flow, explain it message by message, and render it anywhere — no capture needed at render time.

• AI-annotated sequence diagrams — overlay a per-message remarks file onto a flow (apply-remarks), so every arrow carries its own explanation. Exposed over the MCP server as a deterministic author → apply → re-render loop with Claude Code, with the remarks file living beside your capture so the commentary and the diagram cannot drift apart.
• Annotate a single session over MCP — walk through just the one session you care about, end to end, with its packet-detail pane intact.
• Rich markdown in annotations, including small comparison tables, rendered in every output format.
• --remark-width — widen the PDF annotation column for table-heavy remarks.
• New convert command — turn an extracted (and optionally hand-edited or AI-annotated) NDJSON flow straight into a diagram with no PCAP, FXT, or Wireshark at render time. Renders to PDF, standalone HTML, or the combined PDF + interactive viewer.
• convert --format html — a single self-contained HTML sequence diagram with per-message detail cards and remarks, nothing else to host or hand off.
• FXT field-name validation — field names are now checked against Wireshark's registry, so a template that names a nonexistent field is caught up front instead of silently producing an empty diagram. New visualether validate command checks a template without rendering.
• Sharper parameter attribution — each arrow carries only the fields its own PDU contributed, so a frame holding several PDUs reads correctly.
• Order the diagram axes from hosts.txt — --axis-order hosts lays the lanes out left to right in the order the hosts file lists them, instead of by first appearance, so the hub sits where you put it.
• Limit the diagram to the hosts you name — --filter-hosts drops any message touching an address the hosts file does not list, so a busy capture renders as just the conversation you care about. Both flags work on generate and convert.
• Six new protocol samples — f1ap (CU/DU F1 Setup, UE context, RRC/NAS relay), e1ap (CU-CP ↔ CU-UP bearer context), f1u (NR-U flow control over GTP-U), xnap (Xn Setup and UE context retrieval for RRC Resume), x2ap (X2 Setup and SgNB Addition), and oran-fronthaul (O-RAN 7.2 eCPRI CUS-plane with PTP and SyncE).
• EN-DC sample rebuilt — 4g-5g-nsa-connection renders the LTE attach and the X2AP SgNB Addition with both radio stacks. The earlier version was not correct.
• Windows Start Menu and desktop shortcuts — the installer now leaves something to click after installing.
• Community Edition page limit raised from 10 to 30 pages, so the richer protocol flows render whole.

9.0 — The launch release

Turn Wireshark captures into readable sequence diagrams — for humans and AI. The line matured across nine point releases and launched publicly with 9.0.9.

• PCAP → sequence diagram — convert Wireshark pcap/pcapng captures into publication-quality PDF sequence diagrams driven by FXT templates.
• 88 built-in protocol templates spanning routing, VoIP (SIP/RTP), 4G/5G radio, web (HTTP/1–3, gRPC), TLS/security, databases, and IoT.
• AI-assisted analysis via the built-in MCP server — detect protocols, generate templates, render diagrams, extract sessions, fields, and endpoints, and answer "why did this session fail?" in plain English.
• Works with any MCP client — Claude Code, Cursor, VS Code, Windsurf, Codex, Zed, and more; visualether mcp config prints the setup for each.
• Session tracking — follow each session's lifecycle and render per-session diagrams, with cross-file continuity for ring-buffer captures.
• Interactive Capture Atlas — browse a capture's flows in your browser with visualether serve: clickable diagrams, session navigation, and expandable Wireshark packet detail, with serve management (list/stop).
• Four output formats — PDF, HTML, NDJSON, and Markdown, plus a size-aware auto mode and a lazy renderer for very large captures.
• Multi-capture handling — render each --input independently, or merge ring-buffer PCAPs into one logical capture with --merge-inputs.
• 4G/5G radio-stack decoding — auto-decode UDP-framed and DLT_USER MAC/RLC/PDCP captures, with guided per-UE decryption setup (materialize_config).
• Three editions — free Community, plus Professional and Server for watermark-free unlimited output, all formats, and AI analysis.