- Timestamps now show "Jun 28 10:58" for older messages (dropped
"Yesterday" which overflowed the fixed-width column)
- Timestamp column widened from 52px to 72px to fit date+time
- state_snapshot now includes known_peers (historically seen, not
currently connected) so the UI can resolve aliases in history
- MessagePane aliasFor falls back to knownPeers map
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
history_loaded events were fired when peers exchanged history gossip,
but the browser UI often connects to the daemon after that handshake
has already happened. Now the daemon pushes recent messages for all
known rooms immediately after the state_snapshot on each new IPC
connection, so the browser always gets history regardless of timing.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
After hello verification, the connecting peer sends history_request to
the first peer it meets (one per room, no fan-out). The responder queries
SQLite and replies with a history_chunk. Received history is stored via
INSERT OR IGNORE (mid dedup) and emitted as history_loaded IPC events.
- proto: MsgHistoryRequest/Chunk types, HistoryEntry, EvtHistoryLoaded,
ComputeMsgID (sha256 content-addressed ID), MsgID field on ChatMessage
- store: ALTER TABLE ADD COLUMN msg_id + unique index migration (idempotent);
RecentMessagesSince query (msg_id IS NOT NULL filter); msg_id persisted on save
- mesh: RequestHistoryFrom, HandleHistoryRequest, HandleHistoryChunk methods;
historyRequested/historyFirstPeer state to ensure single-peer requests
- peer: dispatch history_request/history_chunk; RequestHistoryFrom after hello;
stamp MsgID on incoming chat messages
- ipc: stamp MsgID on outgoing group chat messages
- EXTENSIONS.md: EXT-007 documented
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Daemon:
- Per-network download directories (-download-dir flag, set_download_dir IPC)
- File transfer resume after disconnection: .tmp.meta sidecars survive
interruption; resume_offset in file-accept lets sender seek and continue
- set_download_dir IPC command; download_dir reported in state_snapshot
Protocol:
- PeerMessage.ResumeOffset (EXT-006) for file transfer resume
- IpcMessage.ShareNetworks json tag changed from "networks" to "network_ids"
to fix duplicate json tag collision with Networks []NetworkInfo
Desktop app (cmd/app):
- Wails v2 shell embedding daemon logic directly (no subprocess)
- System tray on Linux/Windows via getlantern/systray; macOS hides to Dock
- OS notifications for message_received and file_complete via Wails events
- notray build tag for headless/CI builds without GTK tray headers
- build-app.sh: builds web frontend, copies dist, runs wails build
Web / PWA:
- manifest.json + Apple touch icon meta tags for mobile "Add to Home Screen"
- PNG icons (192px, 512px, 180px) generated from SVG
- Wails EventsOn("notify") hook in App.tsx for native OS notifications
CI:
- .gitea/workflows/build.yml: server binaries cross-compiled for 5 platforms,
desktop app for Linux amd64, release artifacts published on v* tags
Docs:
- README: download dir, file transfer resume, desktop app, PWA, CI sections
- EXTENSIONS.md: EXT-004 daemon mode marked shipped; EXT-006 resume added
- FUTURE.md: roadmap updated
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
/room <name> in the TUI sends create_room to the daemon, which persists
it in the rooms SQLite table and echoes room_created back. state_snapshot
now includes persisted rooms so they survive reconnects. Tab navigation
and room rendering pick them up automatically.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Signed invites: waste: URI gains inviter+sig fields (Ed25519); hello
carries the invite so receiving peers can verify against known keys
- RequireInvite per-network flag: rejects peers without valid signed invite
- Hash-based hang links: #waste:base64 fragment pre-fills join form without
server-side leakage of network name
- Multi-share: shares.json (daemon) + waste_shares localStorage (browser);
IPC add_share/remove_share/list_shares commands
- EXTENSIONS.md: addendum documenting all waste-go protocol deviations from
YAW/2; all extensions are additive and backward compatible
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Daemon side:
- internal/shares: new package reads/writes shares.json next to identity
- proto: add ShareEntry type, add_share/remove_share/list_shares IPC commands,
shares_list event, path field on FileEntry
- netmgr: load shares.json on startup; ScanAllShares combines legacy ShareDir
with shares.json entries (recursive walk with relative paths)
- mesh: ScanFiles callback lets manager inject multi-share scanning without
mesh knowing about shares.json
- ipc: handle add_share, remove_share, list_shares commands
Browser side:
- ShareManager component replaces FolderPicker: shows list of named shares
with remove/re-pick buttons; persists share records to waste_shares in
localStorage (name, global flag, networkId)
- FolderPicker retained for internal use; Sidebar now uses ShareManager
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Protocol:
- invite.go: include full 64-char `net` hash in waste: invite blob
(matches yaw2's `net` field — any client parsing the base64 JSON can
join without knowing the plaintext name). Expose NetHash() helper.
- netmgr: add JoinByHash() — join via full 64-char hex hash alone,
storing the short ID as display name. Enables joining yaw2 networks
from a URL that only carries the hash.
- anchor: expose RunByHash() so netmgr can pass a pre-computed hash
directly without a name→hash roundtrip.
- ipc/proto: add network_hash field to join_network — routes to
JoinByHash when present and network_name is absent.
Web UI:
- Parse ?net=<64hex> (yaw2 URL param) and ?a=<anchor> in addition to
existing ?n= / ?invite= params. Hash-only joins send network_hash.
- Sidebar shows yaw: contact card (yaw:<masterID>?n=<alias>) using the
master identity — compatible with yaw2 contact card format. Click to
copy to clipboard.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Messaging:
- Append \n to WS IPC sends so bufio.Scanner fires (messages from web
were silently dropped — the line scanner never saw a newline)
- Handle session_ready in web store to update peer aliases after hello
Alias resolution:
- Seed AddPeer alias from SQLite cache so returning peers resolve
immediately, before hello exchange completes
- Remove stale SavePeer call from AddPeer (was writing placeholder IDs)
- Add PeerAlias() point-lookup to store
Onboarding redesign:
- Three-state UI: disconnected (daemon instructions), connecting,
connected-no-network (join form)
- Read ?n= / ?network= / ?anchor= / ?invite=waste:... URL params to
pre-fill the join form for invite links
- Show daemon alias + peer ID before any network is joined
(master_alias + master_id now included in state_snapshot)
- Identity backup: export (passphrase-protected yaw-key-backup-1 JSON,
download button) and restore (paste + passphrase verify)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add --ws-port flag to waste-daemon; when set, starts a WebSocket IPC
server alongside the existing TCP one. The web UI connects to this from
the browser. Uses nhooyr.io/websocket with localhost-only origin policy.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Each network now carries its own share dir, set at join_network time via
optional share_dir field or updated live with set_share_dir. The global
-share-dir daemon flag becomes a fallback default.
- proto: add ShareDir/DownloadDir to NetworkInfo and IpcMessage
- netmgr: Join accepts shareDir override; SetShareDir updates live
- ipc: wire join_network share_dir and set_share_dir command
- daemon: remove -share-dir from auto-join path (pass "" for default)
- test-network.sh: per-network join with share_dir; isolation verification
section confirms alice/friends and alice/work share dirs are independent
- test-tui.sh: join_network with share_dir; peer IDs resolved after join
All tests pass: YAW/2.1 FS, share isolation, file transfer, persistence.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
File chunks travel over a per-transfer "f:<xid>" WebRTC DataChannel —
direct DTLS-encrypted P2P, the anchor never sees file data. Once the
initial handshake is done the anchor can disappear and transfers continue.
Key design choices:
- Receiver sends "ok" on the file DC before sender streams, eliminating
a race where tiny files could be fully sent/closed before the receiver's
OnMessage handler is registered (open-race §6 analogue for data DCs).
- Auto-accept: receiver accepts every incoming offer immediately.
- Download dir: per-network at <data-dir>/downloads-<netid>/.
- Backpressure: bufferedAmountLowThreshold to avoid overwhelming the DC.
- SHA-256 verified on receive; mismatches emit EvtError and discard temp file.
- IPC: send_file {peer_id, path} → offers the named file from share dir.
- EvtFileComplete {transfer_id, path} emitted on success.
IPC command: {"type":"send_file","peer_id":"<hex>","path":"<filename>"}
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Five interop issues fixed against PROTOCOL.md:
- Join signature now covers nonce_raw || net_ascii (64-char UTF-8 hex
string) as specified in §5.1, not net_raw_bytes. Both anchor server
and client updated to match.
- Chat wire fields renamed to spec names: text/ts (Unix ms int64)
replacing body/sent_at (ISO timestamp). Flat layout on PeerMessage
matches §8 exactly; store and TUI updated accordingly.
- Direct messages now use the spec "pm" type (flat {type,mid,text,ts})
instead of chat+to. Receiver reconstructs a ChatMessage with
dm:<short-id> room for IPC/storage. §8 compliant.
- File transfer message types changed to spec hyphenated names:
file-offer, file-accept, file-cancel, file-done with spec field
names (name/size not filename/size_bytes). §9 compliant.
- DataChannel open-race (§14 gotcha #3) fixed with sync.Once: doOpen
fires on OnOpen callback or immediately if the channel is already
open when WireDataChannel is called (answerer race).
Also fixes two bugs found during testing:
- mid was missing from outgoing wire messages, causing all received
messages to arrive with mid="" and collide on the UNIQUE DB
constraint. mid is now included on all sent chat/pm messages; a
random mid is generated for any received message that omits it.
- Test scripts hardened: kill -9 + active lsof polling replaces blind
sleep for port cleanup; join_network sent before peer_field queries
(local_peer is now network-scoped and nil until joined).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
YAW/2 peer wire protocol is unchanged. Changes are local only.
internal/crypto:
- DeriveForNetwork(master, networkHash) — HKDF-SHA256 from master seed + network hash;
same master + same network always produces the same Ed25519 keypair (stable peer ID)
internal/proto:
- NetworkInfo type (network_id, network_name, local_peer)
- NetworkID field on IpcMessage (optional; commands default to first network when absent)
- Networks []NetworkInfo on state_snapshot (additive alongside existing local_peer)
- EvtNetworkJoined / EvtNetworkLeft events
internal/netmgr (new):
- Manager holds N independent Network contexts (derived identity, mesh, store, anchor)
- Join(name) creates context, derives identity, opens per-network DB, starts anchor client
- Leave(id) / LeaveAll() cancel contexts and close stores
- Resolve(netID) returns named network, or Default() when netID is empty (backward compat)
- Fan-out: Manager.Subscribe() receives tagged events from all networks
- Network IDs are the first 8 hex chars of SHA-256("yaw2-net:"+name) — stable and short
internal/ipc:
- Run(mgr, port) replaces Run(m *mesh.Mesh, port, anchorURL, joinFn)
- Commands without network_id route to mgr.Default() (backward compat)
- state_snapshot includes Networks array; local_peer/connected_peers still populated from first network
- generate_invite, get_file_list, send_message all respect network_id routing
cmd/daemon:
- Creates netmgr.Manager instead of mesh.Mesh directly
- --join and --share-dir pass through Config
- Auto-join via mgr.Join() before IPC starts
test-network.sh:
- Fix peer_name bash bug: $() with && inside triggers set -e; use if/elif/else instead
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- proto: FileEntry, FileListResp types; MsgFileListReq/Resp msg types;
CmdGetFileList + EvtFileList IPC types; Files field on IpcMessage
- mesh: ShareDir field + ScanShareDir(); on DataChannel open, auto-send
MsgFileListReq to new peer; handle MsgFileListReq (scan + reply) and
MsgFileListResp (emit EvtFileList to IPC subscribers)
- ipc: get_file_list command — own list returned immediately; remote peer
list requested via DataChannel (response arrives as EvtFileList event)
- daemon: -share-dir flag wired to mesh.ShareDir
- test scripts: pass -share-dir /home/frejoh/Downloads/{alice,bob,charlie};
test-network.sh verifies each peer's own file list via get_file_list
- FUTURE.md: document per-network share directories and multi-network design
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- internal/invite: Encode/Decode waste:<base64json{anchor,network}>
- proto: CmdGenerateInvite + EvtInviteGenerated + InviteString field
- ipc: track active network name; handle generate_invite (returns invite
string when joined to a network, errors if not joined or no anchor)
- daemon: --join <invite> flag — decodes anchor URL + network name,
sets anchor and auto-joins on startup
- tui: --join <invite> flag — extracts network name, skips -network
requirement; ctrl+i generates invite and shows it full-screen;
Esc dismisses the invite overlay
- FUTURE.md: document multi-network derived-identity design
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
ipc: CmdSendMessage now routes to the named recipient only (m.SendTo) when
the "to" field is set, rather than broadcasting to all peers. Group messages
continue to use Broadcast. This is the correct privacy behaviour for DMs.
test-network.sh:
- Resolve each peer's hex id from get_state before joining the network,
using a retry loop with timeout to handle slow daemon startup.
- Added a DM section: alice→bob, bob→alice, charlie→alice, alice→charlie.
DMs use the "to" + "room":"dm:<peer_id>" convention.
- Fixed jq -cn (compact) instead of jq -n (pretty) so the IPC newline-
delimited protocol receives a single JSON line per command.
- pretty() now renders DMs as 📨 DM <from→to> rather than 💬 #room.
- Persistence check now breaks down totals by group vs DM message count.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Each daemon writes to <data-dir>/messages.db on startup. Messages received
or sent are stored immediately; duplicate mids (INSERT OR IGNORE) are safe
to call multiple times. Peer aliases are upserted on peer_connected and
again after hello verification when the real nick is known.
Schema
- messages(mid UNIQUE, room, from_peer, body, sent_at) — mid is the YAW/2
dedup key added in the proto migration; index on (room, sent_at) for
efficient per-room queries.
- peers(peer_id PK, alias, last_seen) — cache of every peer ever seen,
used to resolve hex ids to names when peers are offline.
Wiring
- store.Open called in cmd/daemon/main.go, passed to mesh.New.
- mesh.Mesh holds *store.Store (nil-safe; persistence is optional).
- mesh.SaveMessage called in dispatchPeerMessage (incoming) and ipc
CmdSendMessage (outgoing) so the local node's own messages are stored.
- mesh.UpdatePeerAlias called after hello verification updates the alias
with the verified nick rather than the placeholder short-id.
Messages only accumulate from join time forward — no history replay to
late-joining peers; each node's view starts from when it connected.
test-network.sh: added SQLite verification block that queries each node's
DB after the test and prints message + peer counts.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
test-network.sh: shell script that starts anchor + alice + bob + charlie,
joins them to a named network, exchanges five messages, then has charlie
leave — verifying the full join/chat/leave cycle with coloured per-peer output.
ipc: two fixes exposed by the test script:
- Network join context was per-IPC-client, so the join was immediately
cancelled when the nc connection closed. Lifted join/leave state to the
Run() level (shared across all clients, protected by a mutex) so the
network stays connected independent of which client issued the command.
- Event-pusher goroutine could panic with "send on closed channel" when
the command loop closed writeCh while the pusher was mid-select. Added
defer recover() to both the pusher goroutine and the send helper, and
removed the default arm so the select blocks cleanly on done.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
