Merge branch 'master' of https://git.d-popov.com/popov/neuron-tai
This commit is contained in:
@@ -127,9 +127,9 @@
|
||||
display:flex; align-items:center; gap:8px; color:var(--dim); margin:0;
|
||||
}
|
||||
.chat-toolbar select {
|
||||
min-width:220px; max-width:min(420px, 50vw);
|
||||
min-width:280px; max-width:min(520px, 60vw);
|
||||
color:var(--fg); background:var(--chat-input-bg); border:1px solid var(--border);
|
||||
border-radius:6px; padding:6px 8px;
|
||||
border-radius:6px; padding:6px 8px; font-size:13px;
|
||||
}
|
||||
.chat-status { color:var(--dim); font-size:12px; margin-left:auto; }
|
||||
.chat-messages {
|
||||
@@ -277,6 +277,7 @@ const esc = s => String(s).replace(/[&<>"]/g,
|
||||
const usdt = v => (Math.round(v * 1e6) / 1e6).toFixed(6);
|
||||
const tps = v => (v === null || v === undefined) ? "?" : (Math.round(v * 10) / 10).toFixed(1);
|
||||
const copies = v => (v === null || v === undefined) ? "?" : Number(v).toFixed(2);
|
||||
const hp = v => (v === null || v === undefined) ? "?HP" : `${Number(v).toFixed(1)}HP`;
|
||||
const short = (s, n=14) => { s = String(s); return s.length > n ? s.slice(0, 6) + "…" + s.slice(-5) : s; };
|
||||
|
||||
function modelAliasKey(value) {
|
||||
@@ -831,6 +832,7 @@ let isLoggedIn = false;
|
||||
let accountApiKeys = [];
|
||||
let accountUsageRecords = [];
|
||||
let lastStats = null;
|
||||
let lastRouting = null;
|
||||
let availableModels = [];
|
||||
let chatHistory = [];
|
||||
let chatBusy = false;
|
||||
@@ -956,11 +958,7 @@ function selectChatSession(sessionId) {
|
||||
renderChatAuthHint();
|
||||
}
|
||||
|
||||
function deleteChatSession(sessionId, event) {
|
||||
if (event) {
|
||||
event.preventDefault();
|
||||
event.stopPropagation();
|
||||
}
|
||||
function deleteChatSession(sessionId) {
|
||||
if (chatBusy) return;
|
||||
const index = chatSessions.findIndex(item => item.id === sessionId);
|
||||
if (index < 0) return;
|
||||
@@ -1020,18 +1018,43 @@ function renderChatSessionList() {
|
||||
const active = session.id === activeChatSessionId ? " active" : "";
|
||||
const title = esc(chatSessionTitle(session));
|
||||
const when = esc(formatSessionTime(session.updatedAt || session.createdAt));
|
||||
const id = JSON.stringify(session.id);
|
||||
const sessionId = esc(session.id);
|
||||
return `<div class="chat-session-item${active}" role="button" tabindex="0"` +
|
||||
` onclick="selectChatSession(${id})"` +
|
||||
` onkeydown="if(event.key==='Enter'||event.key===' '){event.preventDefault();selectChatSession(${id});}">` +
|
||||
` data-session-id="${sessionId}">` +
|
||||
`<div class="chat-session-title">${title}</div>` +
|
||||
(when ? `<div class="chat-session-meta">${when}</div>` : "") +
|
||||
`<button type="button" class="chat-session-delete" title="Delete chat"` +
|
||||
` onclick="deleteChatSession(${id}, event)">×</button>` +
|
||||
` data-delete-session="${sessionId}">×</button>` +
|
||||
`</div>`;
|
||||
}).join("");
|
||||
}
|
||||
|
||||
function bindChatSessionList() {
|
||||
const list = $("chat-session-list");
|
||||
if (!list || list.dataset.bound) return;
|
||||
list.dataset.bound = "1";
|
||||
list.addEventListener("click", event => {
|
||||
const deleteBtn = event.target.closest("[data-delete-session]");
|
||||
if (deleteBtn) {
|
||||
event.preventDefault();
|
||||
event.stopPropagation();
|
||||
deleteChatSession(deleteBtn.getAttribute("data-delete-session"));
|
||||
return;
|
||||
}
|
||||
const item = event.target.closest("[data-session-id]");
|
||||
if (item) selectChatSession(item.getAttribute("data-session-id"));
|
||||
});
|
||||
list.addEventListener("keydown", event => {
|
||||
if (event.key !== "Enter" && event.key !== " ") return;
|
||||
const deleteBtn = event.target.closest("[data-delete-session]");
|
||||
if (deleteBtn) return;
|
||||
const item = event.target.closest("[data-session-id]");
|
||||
if (!item) return;
|
||||
event.preventDefault();
|
||||
selectChatSession(item.getAttribute("data-session-id"));
|
||||
});
|
||||
}
|
||||
|
||||
function switchDashboardTab(name) {
|
||||
if (name === "admin" && !isAdmin) name = "overview";
|
||||
if (name === "billing" && !isLoggedIn) name = "overview";
|
||||
@@ -1079,6 +1102,44 @@ function renderChatHistory() {
|
||||
history.scrollTop = history.scrollHeight;
|
||||
}
|
||||
|
||||
function findRoutingForModel(model, routing) {
|
||||
const models = (routing && routing.models) || {};
|
||||
const keys = [model.id, model.name, model.hf_repo, ...(model.aliases || [])].filter(Boolean);
|
||||
for (const key of keys) {
|
||||
if (models[key]) return models[key];
|
||||
const aliasKey = modelAliasKey(key);
|
||||
for (const [routeKey, info] of Object.entries(models)) {
|
||||
if (modelAliasKey(routeKey) === aliasKey) return info;
|
||||
}
|
||||
}
|
||||
return null;
|
||||
}
|
||||
|
||||
function chatModelHealthHp(servedCopies) {
|
||||
return hp(servedCopies);
|
||||
}
|
||||
|
||||
function chatModelTypicalTps(model, routing) {
|
||||
const info = findRoutingForModel(model, routing);
|
||||
const routes = (info && info.routes) || [];
|
||||
if (!routes.length) return null;
|
||||
const proven = routes.filter(r => r.status === "proven" && Number(r.tps) > 0);
|
||||
if (proven.length) return Math.max(...proven.map(r => Number(r.tps)));
|
||||
const priors = routes.map(r => Number(r.prior_tps ?? r.tps ?? 0)).filter(v => v > 0);
|
||||
return priors.length ? Math.max(...priors) : null;
|
||||
}
|
||||
|
||||
function chatModelOptionLabel(model, routing) {
|
||||
const base = model.name && model.name !== model.id
|
||||
? `${model.name} (${model.id})`
|
||||
: model.id;
|
||||
const health = chatModelHealthHp(model.servedCopies);
|
||||
const speed = chatModelTypicalTps(model, routing);
|
||||
const speedText = speed === null ? "?" : `${tps(speed)} tok/s`;
|
||||
const suffix = model.recommended ? " · recommended" : "";
|
||||
return `${base} · ${health} · ${speedText}${suffix}`;
|
||||
}
|
||||
|
||||
function renderChatModels() {
|
||||
const select = $("chat-model");
|
||||
if (!select) return;
|
||||
@@ -1094,11 +1155,8 @@ function renderChatModels() {
|
||||
selectedChatModel = preferred.id;
|
||||
localStorage.setItem("meshnet_chat_model", selectedChatModel);
|
||||
select.innerHTML = models.map(model => {
|
||||
const label = model.name && model.name !== model.id
|
||||
? `${model.name} (${model.id})`
|
||||
: model.id;
|
||||
const suffix = model.recommended ? " [recommended]" : "";
|
||||
return `<option value="${esc(model.id)}"${model.id === selectedChatModel ? " selected" : ""}>${esc(label + suffix)}</option>`;
|
||||
const label = chatModelOptionLabel(model, lastRouting);
|
||||
return `<option value="${esc(model.id)}"${model.id === selectedChatModel ? " selected" : ""}>${esc(label)}</option>`;
|
||||
}).join("");
|
||||
select.value = selectedChatModel;
|
||||
}
|
||||
@@ -1493,6 +1551,7 @@ async function sendChat() {
|
||||
renderChatStatus("stopped");
|
||||
}
|
||||
} else {
|
||||
console.error("chat stream failed", err);
|
||||
removeAssistantMessage(assistantMessage);
|
||||
const error = (err && err.message) || "request failed";
|
||||
chatHistory.push({ role: "error", content: error, model: selectedChatModel });
|
||||
@@ -1574,12 +1633,15 @@ async function refresh() {
|
||||
]);
|
||||
const [summary, settlements, wallets] = adminData;
|
||||
lastStats = stats;
|
||||
lastRouting = routing;
|
||||
availableModels = ((models && models.data) || []).map(model => ({
|
||||
id: model.id,
|
||||
name: model.name || model.id,
|
||||
hf_repo: model.hf_repo,
|
||||
recommended: Boolean(model.recommended),
|
||||
aliases: model.aliases || [],
|
||||
coverage: model.shard_coverage_percentage,
|
||||
servedCopies: model.served_model_copies,
|
||||
})).filter(model => model.id);
|
||||
renderHive(raft);
|
||||
renderNodes(map);
|
||||
@@ -1609,6 +1671,7 @@ async function refreshIfIdle() {
|
||||
}
|
||||
|
||||
refresh();
|
||||
bindChatSessionList();
|
||||
initChatSessions();
|
||||
bindChatPromptShortcuts();
|
||||
renderAccountPanel();
|
||||
|
||||
@@ -2729,6 +2729,11 @@ class _TrackerHandler(http.server.BaseHTTPRequestHandler):
|
||||
required_start,
|
||||
required_end,
|
||||
)
|
||||
served_copies = _served_model_copies(
|
||||
model_nodes,
|
||||
required_start,
|
||||
required_end,
|
||||
)
|
||||
aliases = [name]
|
||||
hf_repo = preset.get("hf_repo")
|
||||
if hf_repo and hf_repo not in aliases:
|
||||
@@ -2748,6 +2753,7 @@ class _TrackerHandler(http.server.BaseHTTPRequestHandler):
|
||||
"recommended": bool(preset.get("recommended", False)),
|
||||
"deployment": _deployment_summary(alive, preset),
|
||||
"shard_coverage_percentage": coverage,
|
||||
"served_model_copies": served_copies,
|
||||
})
|
||||
seen_ids.add(name)
|
||||
if hf_repo:
|
||||
@@ -2778,6 +2784,11 @@ class _TrackerHandler(http.server.BaseHTTPRequestHandler):
|
||||
aliases = [model_id, *[name for name in short_names if name != model_id]]
|
||||
required_start = 0
|
||||
required_end = max(node.num_layers for node in model_nodes) - 1
|
||||
coverage = _coverage_percentage(
|
||||
model_nodes,
|
||||
required_start,
|
||||
required_end,
|
||||
)
|
||||
data.append({
|
||||
"id": model_id,
|
||||
"object": "model",
|
||||
@@ -2787,7 +2798,8 @@ class _TrackerHandler(http.server.BaseHTTPRequestHandler):
|
||||
"hf_repo": model_id if any(node.hf_repo == model_id for node in model_nodes) else None,
|
||||
"aliases": aliases,
|
||||
"metadata": _model_metadata_from_nodes(model_nodes),
|
||||
"shard_coverage_percentage": _coverage_percentage(
|
||||
"shard_coverage_percentage": coverage,
|
||||
"served_model_copies": _served_model_copies(
|
||||
model_nodes,
|
||||
required_start,
|
||||
required_end,
|
||||
@@ -3457,6 +3469,7 @@ class _TrackerHandler(http.server.BaseHTTPRequestHandler):
|
||||
self.send_response(200)
|
||||
self.send_header("Content-Type", "text/event-stream; charset=utf-8")
|
||||
self.send_header("Cache-Control", "no-cache")
|
||||
self.send_header("X-Accel-Buffering", "no")
|
||||
self.end_headers()
|
||||
stream_usage: dict | None = None
|
||||
observed_stream_tokens = 0
|
||||
@@ -3499,7 +3512,7 @@ class _TrackerHandler(http.server.BaseHTTPRequestHandler):
|
||||
if usage is not None:
|
||||
stream_usage = usage
|
||||
except (BrokenPipeError, ConnectionResetError):
|
||||
pass
|
||||
client_gone = True
|
||||
if self._finalize_proxy_cancel(
|
||||
proxy_ctx=proxy_ctx,
|
||||
server=server,
|
||||
@@ -3537,6 +3550,7 @@ class _TrackerHandler(http.server.BaseHTTPRequestHandler):
|
||||
route_model=route_model,
|
||||
status=200,
|
||||
stream=True,
|
||||
client_disconnected=client_gone,
|
||||
tokens=tokens,
|
||||
elapsed_seconds=round(elapsed, 4),
|
||||
tokens_per_sec=round(tokens / elapsed, 4) if elapsed > 0 else 0.0,
|
||||
@@ -3738,6 +3752,7 @@ class _TrackerHandler(http.server.BaseHTTPRequestHandler):
|
||||
self.send_response(int(first.get("status", 200)))
|
||||
self.send_header("Content-Type", headers.get("Content-Type", "text/event-stream; charset=utf-8"))
|
||||
self.send_header("Cache-Control", "no-cache")
|
||||
self.send_header("X-Accel-Buffering", "no")
|
||||
self.end_headers()
|
||||
server: _TrackerHTTPServer = self.server # type: ignore[assignment]
|
||||
stream_usage: dict | None = None
|
||||
@@ -3811,6 +3826,7 @@ class _TrackerHandler(http.server.BaseHTTPRequestHandler):
|
||||
route_model=route_model,
|
||||
status=200,
|
||||
stream=True,
|
||||
client_disconnected=client_gone,
|
||||
relay=True,
|
||||
tokens=tokens,
|
||||
elapsed_seconds=round(elapsed, 4),
|
||||
|
||||
@@ -49,6 +49,42 @@ def test_dashboard_chat_uses_streaming_fetch():
|
||||
assert "stream: true" in html
|
||||
assert ".body.getReader()" in html
|
||||
assert '=== "[DONE]"' in html
|
||||
assert 'console.error("chat stream failed", err)' in html
|
||||
|
||||
|
||||
def test_dashboard_chat_model_selector_shows_health_and_speed():
|
||||
tracker = TrackerServer()
|
||||
port = tracker.start()
|
||||
try:
|
||||
html = urllib.request.urlopen(
|
||||
f"http://127.0.0.1:{port}/dashboard"
|
||||
).read().decode()
|
||||
finally:
|
||||
tracker.stop()
|
||||
|
||||
assert "chatModelHealthHp" in html
|
||||
assert "chatModelTypicalTps" in html
|
||||
assert "chatModelOptionLabel" in html
|
||||
assert "findRoutingForModel" in html
|
||||
assert "tok/s" in html
|
||||
assert "0.0HP" in html or "toFixed(1)}HP" in html
|
||||
|
||||
|
||||
def test_dashboard_chat_sessions_use_delegated_handlers():
|
||||
tracker = TrackerServer()
|
||||
port = tracker.start()
|
||||
try:
|
||||
html = urllib.request.urlopen(
|
||||
f"http://127.0.0.1:{port}/dashboard"
|
||||
).read().decode()
|
||||
finally:
|
||||
tracker.stop()
|
||||
|
||||
assert "bindChatSessionList" in html
|
||||
assert "data-session-id=" in html
|
||||
assert "data-delete-session=" in html
|
||||
assert 'onclick="selectChatSession(' not in html
|
||||
assert 'onclick="deleteChatSession(' not in html
|
||||
|
||||
|
||||
def test_dashboard_served_by_follower():
|
||||
|
||||
@@ -467,6 +467,7 @@ def test_tracker_logs_stream_progress_before_request_completes():
|
||||
method="POST",
|
||||
)
|
||||
response = urllib.request.urlopen(req, timeout=3.0)
|
||||
assert response.headers["X-Accel-Buffering"] == "no"
|
||||
first_line = response.readline()
|
||||
assert first_line.startswith(b"data:")
|
||||
assert chunk_sent.wait(timeout=1.0)
|
||||
|
||||
Reference in New Issue
Block a user