Compare commits

...
Author SHA1 Message Date
müde
fcd407da11 todo: mark terminal coherence pass as landed 2026-05-18 18:13:25 +02:00
müde
f8f2ccff52 agent terminal: coherence pass
layout
- unified prefix-column for every row kind: padding-left + negative
  text-indent so the glyph (→ ← · ◆ ✓ ✗ ⌁ !) sits in the same column
  whether the row is flat or a <details>. wraps hang under the body,
  not under the glyph.
- expandable rows drop the directional glyph from their summary text;
  the ▸/▾ disclosure marker from CSS sits in the prefix column instead,
  and the row's colour still carries cyan = outbound, muted = inbound.
- turn-start / turn-end de-weighted: bold/margin/tint dropped, the
  coloured left rule alone marks the boundary.

note classification
- stderr lines render orange with a `!` glyph (was muted `·`)
- operator-initiated notes (cancel/compact/model/new-session) render
  mauve italic (was muted `·` indistinguishable from harness chatter)
- catch-all .sys row escalates to orange `!` so unrecognised stream-json
  shapes surface for follow-up instead of hiding in muted noise

message-bearing rows
- send / ask / answer tool_use rich renderers default-open with the
  body inline; new ask + answer renderers (previously fell through to
  the generic JSON dump). recv tool_result also default-open, keyed by
  tracking tool_use_id → name across the stream so we know which
  result came from which tool.
- assistant text rows render markdown.
- bodies use vendored marked v4.0.2 (hive-fr0nt::MARKED_JS); falls
  back to plain text when the asset doesn't load.

extra-mcp tool pretty-print
- generic args formatter replaces the raw JSON dump for unknown tools
  (single-string field → `name k: "v"`; single dict / multi-field →
  trimmed `k: v · k: v …` summary)

dashboard .live .msgrow gets a text-indent: 0 reset so the new
hanging-indent metrics from TERMINAL_CSS don't leak into the flex-grid
broker rows.
2026-05-18 18:13:14 +02:00
müde
f827187341 agent ctx-badge: seed Bus::last_usage from latest turn_stats row on startup 2026-05-18 18:00:48 +02:00
müde
8a3e8bfb7f todo + terminal-rendering: ctx-badge cold-load, auto session-reset, more coherence-pass gripes 2026-05-18 18:00:46 +02:00
13 changed files with 402 additions and 53 deletions

View file

@ -18,7 +18,7 @@
## Dashboard
- **Per-agent terminal coherence pass**: row taxonomy + colour scheme has drifted — glyph reuse, ad-hoc `<details>` rules, undifferentiated stderr, `.sys` catch-all silently hiding unrecognised events. Full audit + proposed unified scheme in [`docs/terminal-rendering.md`](docs/terminal-rendering.md).
- ~~**Per-agent terminal coherence pass**~~ ✓ landed — unified prefix column (padding-left + negative text-indent so every row kind aligns), `<details>` summaries drop directional glyph (CSS `▸/▾` sits in prefix column instead), turn boundaries de-weighted (border-left only, no bold/margin/tint), stderr lines render orange `!`, operator-initiated notes mauve italic, catch-all `.sys` escalated to orange so unrecognised stream-json surfaces. Message-bearing tool calls (`send`/`ask`/`answer`/`recv`) render default-open with markdown bodies via vendored `marked` v4.0.2; assistant text also markdown-rendered. Extra-MCP tools get a generic args pretty-printer instead of raw JSON. See [`docs/terminal-rendering.md`](docs/terminal-rendering.md) for the as-built taxonomy.
- **Delivered-reminder rollups**: per-agent delivered-count chip (last 24h) + histogram of attempts-vs-successes on the container row. Needs `Broker::count_delivered_reminders_since(agent, ts)` (cheap COUNT against the `reminders` table, `WHERE agent = ?1 AND sent_at >= ?2`).
## Security
@ -61,6 +61,12 @@ how often the friction bites in normal use.
- **Surface per-turn stats on the agent web UI**: "N turns today" chip + rolling tool-call histogram tooltip on the model chip. (`open_threads` and `open_reminders` chips already landed via other paths — open-threads section on the page + reminder count chip on the container row.) Reads the per-agent `turn_stats.sqlite`.
- **Stats UI on the main dashboard**: per-agent rollups (avg turn duration, tokens-since-boot, top 5 tools) on the container row. Same data source, host-side aggregation query.
## Harness Behaviour
- **Persist + cold-load current context size on the per-agent page**: the `ctx-badge` (Claude Code's bottom-right "N tokens" indicator) currently only populates after the first `TokenUsageChanged` SSE event arrives, which is the *next* turn — until then the badge is empty. Operator can't see "this agent is at 78% context" before deciding to manually compact / reset / message it. Last known token usage should be persisted (likely a small `/state/hyperhive-token-usage` blob, or a row in turn_stats already has it — pull last row's totals on cold load) and returned by `/api/state` so the badge paints with real numbers on first render.
- **Auto session-reset when context is large and cache is cold**: today every turn uses `--continue`, so a long-lived agent carries its entire transcript forward indefinitely. When the next turn's context is above some threshold (rough starting point: ~50% of the session limit — hive startup alone burned ~15%, so the headroom disappears fast) *and* the prompt cache is no longer warm (last turn ended past the cache TTL), it's cheaper to start fresh than to re-send the whole history uncached. Open question: drop `--continue` vs. trigger `--compact` first — needs measurement of what each actually costs (uncached re-read of N tokens vs. a compact turn's own token spend + the post-compact uncached re-read). Decision should be data-driven, not guessed. Needs: a context-size estimate per turn (turn_stats already tracks token usage), a cache-warmth heuristic (time since last turn vs. cache TTL), and a one-shot fresh-session path in `turn.rs` mirroring the existing `↻ new session` button.
## Bugs
- **Token-budget exhaustion crashes the harness**: when claude's account hits its rate/token cap, the in-flight `claude --print` invocation returns an error the harness doesn't recognise as recoverable, the serve loop exits, and the container stays up with a dead daemon. Operator only notices when an unrelated wake fails to drive a turn. Want: detect the budget-exceeded class of failure (likely a specific stderr line or stream-json `rate_limit_event` shape), fire a `LiveEvent::StatusChanged("rate_limited")` or new status, surface as a red badge + banner on the dashboard + per-agent UI, and have the serve loop park (sleep N minutes, retry) instead of returning Err. Operator can also see "this agent is rate-limited until ~HH:MM" if claude tells us when. Inspect `crate::turn::run_claude`'s `bail!` paths + claude's stderr conventions for the budget error string.

View file

@ -71,6 +71,56 @@ shared `.live .<class>` styling).
autonomous harness chatter; nothing flags them as "operator
initiated."
7. **Turn-start / turn-end are visually overweight**: the
triggering event row + the closing row both get bold text,
top margins, and a full coloured background tint. The
coloured left rule alone already says "this is a turn
boundary" — the heavy chrome adds noise without information.
Drop the bold/margin/tint, keep the left rule.
8. **Left-alignment is incoherent across row types**: the `▸`
disclosure marker on expandable rows doesn't sit in the
same column as flat-row prefix glyphs (`→`, `←`, `·`), so
the prefix column wanders by row kind. Expandable tool_use
(`→ Name …`) and expandable tool_result (`▸ ← Nl …`) use
different layouts from each other too. Pick one prefix
column and align every row kind into it; the disclosure
marker should be visually inside that column, not pushed
to the side.
9. **Continuation lines aren't inset**: when a row's body
wraps (e.g. `→ foo bar baz quux …`), the second visual
line starts at column 0 rather than under the `f` of
`foo`, so wrapped content blurs into the next row. Want a
`text-indent` / hanging-indent rule so continuation lines
align with the start of the body, not the prefix glyph.
10. **Most message-bearing rows are collapsed by default**
even when the body is the whole point: `send`, `recv`,
`ask`, and friends hide their text behind a `<details>`
summary. The summary headline is rarely enough context.
Want these expanded by default, with collapse reserved
for genuinely heavy payloads (multi-page diffs, long
tool_result blocks).
11. **Tools from extra MCP servers aren't pretty-printed**:
`renderRichToolUse` only special-cases the built-in
hyperhive tools (Write/Edit/send). Anything coming from
`extraMcpServers` (matrix, bitburner, …) falls through
to the generic `→ Name args…` JSON dump. Either a
plugin-style hook in the renderer map, or at minimum a
nicer generic args-pretty-printer that handles common
shapes (single string arg, single dict arg, etc.).
12. **No markdown rendering on message bodies**: `send` /
`recv` / `ask` / `answer` / agent text content all
arrive as markdown-flavoured prose (bullets, fenced
code, bold, links) but render as raw text. Want a
minimal markdown pass — at least lists, fenced code,
inline code, bold/italic, links — applied to message
bodies and probably to the `assistant.content[].text`
`.text` row.
## Suggested coherence pass
Pick one scheme and audit all renderers to match. A concrete
@ -96,6 +146,48 @@ The `.sys` catch-all should escalate visually — a louder
shapes for future fix-up rather than hiding them in the muted
note stream.
### Layout rules to apply uniformly
- **One prefix column for every row kind** — flat rows and
expandable rows alike. The disclosure marker (`▸`) lives
inside that column, not as a separate gutter, so glyph
alignment doesn't shift between row types.
- **Hanging indent on wrapped bodies** so continuation lines
start under the first character of the body, not under the
prefix glyph. Probably `display: grid` with
`grid-template-columns: <prefix-col> 1fr` per row, or
`text-indent`/`padding-left` with negative `text-indent`.
- **Turn boundaries are rule-only** — drop the bold + margin
+ tint on `.turn-start` / `.turn-body` / `.turn-end-*`.
The coloured left rule alone carries the boundary signal.
- **Default-expanded message rows**: `send`, `recv`, `ask`,
`answer`, and short-ish `text` rows render their body
inline (no `<details>`). Reserve collapse for genuinely
heavy bodies — multi-page diffs, long tool_results,
thinking blocks past N lines.
### Renderer surface for extra MCP tools
`renderRichToolUse` currently switches on hard-coded tool
names. Replace with a registry keyed by
`mcp__<server>__<tool>` (and the built-in claude tool names)
that maps to a `(toolUse) -> Node` renderer. Per-agent extra
MCPs can register their own renderers via a small JS hook
(loaded the same way `extra-mcp.json` is loaded server-side);
the fallback is a generic args-pretty-printer that handles
single-string and single-dict shapes nicely instead of dumping
raw JSON.
### Markdown rendering
Apply a minimal markdown pass to message bodies (`send` /
`recv` / `ask` / `answer`) and to assistant `text` rows.
Scope: paragraphs, lists, fenced + inline code, bold/italic,
links. No HTML passthrough, no images, no tables —
explicitly bounded so we don't import a kitchen-sink parser
into the per-agent page. A small handwritten pass or a tiny
dep (e.g. `micromark` / `marked`) both work.
## Dashboard side (not covered here)
The main dashboard's message-flow pane is a different beast:

View file

@ -656,6 +656,79 @@
log.innerHTML = '';
function trim(s, n) { return s.length > n ? s.slice(0, n) + '…' : s; }
// Render a message body as markdown into a new <div class="md">.
// Wraps `marked.parse` so the per-row body element carries the
// `.md` class (CSS in TERMINAL_CSS scopes paragraph/code/list
// styles to it). Falls back to a plain text node if marked isn't
// loaded (network glitch, asset 404) so the body still renders.
function mdNode(text) {
const div = document.createElement('div');
div.className = 'md';
const src = String(text || '');
if (window.marked && typeof window.marked.parse === 'function') {
try {
marked.setOptions({ breaks: true, gfm: true });
div.innerHTML = marked.parse(src);
} catch (err) {
console.warn('marked failed', err);
div.textContent = src;
}
} else {
div.textContent = src;
}
return div;
}
// Build a default-open details row whose body is markdown-rendered.
// Used by send / ask / answer tool_use renderers and by `recv`
// tool_result so message-bearing rows show their content inline
// without an extra click.
function detailsOpenMd(api, cls, summary, body) {
const d = api.details(cls, summary, '');
d.open = true;
const pre = d.querySelector('pre.tool-body');
if (pre) {
pre.replaceWith(mdNode(body));
} else {
d.appendChild(mdNode(body));
}
return d;
}
// Generic args-pretty-printer for unknown / extra-MCP tools. The
// built-in switch handles the common claude/hyperhive tools; this
// is the fallback so an `mcp__matrix__send_message` or similar
// doesn't dump raw JSON. Heuristics: single string-valued field →
// `Name field: "value"`; single dict-valued field → `Name field
// {…}`; otherwise compact JSON. Always trimmed to fit a row.
function fmtArgsGeneric(name, input) {
const keys = Object.keys(input || {});
if (keys.length === 0) return name + '()';
if (keys.length === 1) {
const k = keys[0];
const v = input[k];
if (typeof v === 'string') {
const oneline = v.replace(/\s+/g, ' ').trim();
return name + ' ' + k + ': ' + JSON.stringify(trim(oneline, 100));
}
if (typeof v === 'number' || typeof v === 'boolean') {
return name + ' ' + k + ': ' + JSON.stringify(v);
}
}
// Multi-field: render `k: v` pairs with strings/numbers inlined and
// anything else summarised by type so the row stays readable.
const pretty = keys.slice(0, 4).map((k) => {
const v = input[k];
if (v == null) return k + ': null';
if (typeof v === 'string') {
const oneline = v.replace(/\s+/g, ' ').trim();
return k + ': ' + JSON.stringify(trim(oneline, 40));
}
if (typeof v === 'number' || typeof v === 'boolean') return k + ': ' + v;
if (Array.isArray(v)) return k + `: [${v.length}]`;
return k + ': {…}';
});
const tail = keys.length > 4 ? ' …+' + (keys.length - 4) : '';
return name + ' ' + pretty.join(' · ') + tail;
}
// Pretty-print a tool call: per-known-tool format, fallback to JSON
// for unknown tools.
function fmtToolUse(c) {
@ -678,7 +751,7 @@
case 'mcp__hyperhive__kill': return short + ' ' + (input.name || '');
case 'mcp__hyperhive__request_apply_commit':
return short + ' ' + (input.agent || '') + ' @ ' + (input.commit_ref || '').slice(0, 12);
default: return name + ' ' + trim(JSON.stringify(input), 200);
default: return fmtArgsGeneric(short, input);
}
}
// Build a "rich" tool_use row for tools whose input has a body we
@ -709,37 +782,73 @@
+ '\n'
+ newLines.map(l => '+ ' + l).join('\n');
}
const summary = '→ ' + name + ' ' + path + ' · '
// Summaries on expandable rows omit the row's directional glyph
// (`→`) — the disclosure marker (`▸/▾`) from CSS sits in the
// prefix column for every row kind, and the row's cyan colour
// already signals "outbound tool".
const summary = name + ' ' + path + ' · '
+ (minus ? '-' + minus + ' ' : '') + '+' + plus;
return api.detailsDiff('tool-use', summary, body);
}
// Message-bearing tools render default-open with a markdown body so
// the operator sees the content without an extra click. send / ask
// address a target; answer attaches to an existing question id.
if (name === 'mcp__hyperhive__send') {
const to = input.to || '?';
const body = String(input.body || '');
const headline = body.replace(/\s+/g, ' ').trim().slice(0, 80);
const lines = body.split('\n').length;
const summary = '→ send → ' + to + (lines > 1 ? ` · ${lines}L` : '')
+ (headline ? ' · ' + headline + (body.length > 80 ? '…' : '') : '');
return api.details('tool-use', summary, body);
return detailsOpenMd(api, 'tool-use',
'send → ' + to + (lines > 1 ? ` · ${lines}L` : ''),
body);
}
if (name === 'mcp__hyperhive__ask') {
const to = input.to || 'operator';
const q = String(input.question || '');
const lines = q.split('\n').length;
return detailsOpenMd(api, 'tool-use',
'ask → ' + to + (lines > 1 ? ` · ${lines}L` : ''),
q);
}
if (name === 'mcp__hyperhive__answer') {
const id = input.id != null ? String(input.id) : '?';
const a = String(input.answer || '');
const lines = a.split('\n').length;
return detailsOpenMd(api, 'tool-use',
'answer #' + id + (lines > 1 ? ` · ${lines}L` : ''),
a);
}
return null;
}
// Track tool_use_id → tool name so we can decide on rendering when the
// matching tool_result lands later. Lets us default-open the body for
// message-bearing tools (`recv`) while keeping shell/file tool output
// collapsed unless the operator clicks. Cleared on /clear; otherwise
// grows with the session — entries are tiny strings.
const toolNameById = new Map();
function renderToolResult(c, api) {
const txt = Array.isArray(c.content)
? c.content.map(p => p.text || '').join('')
: (c.content || '');
const summary = '← ' + (() => {
const trimmed = txt.replace(/\s+/g, ' ').trim();
const sourceName = c.tool_use_id ? toolNameById.get(c.tool_use_id) : null;
const isMessageBearing = sourceName === 'mcp__hyperhive__recv';
const trimmed = txt.replace(/\s+/g, ' ').trim();
const summaryBody = (() => {
if (!trimmed) return '(empty)';
if (trimmed.length <= 120) return trimmed;
const lines = txt.split('\n').filter(l => l.length).length;
const headline = trimmed.slice(0, 90) + '…';
return `${lines}L · ${headline}`;
})();
// Flat row: keep the `←` glyph in the prefix column. Details rows
// drop it — the `▸/▾` disclosure marker sits in that column via CSS.
if (isMessageBearing && txt.trim()) {
return detailsOpenMd(api, 'tool-result-block',
'recv ← ' + summaryBody, txt);
}
if (!txt.trim() || txt.length <= 120) {
api.row('tool-result', summary);
api.row('tool-result', '← ' + summaryBody);
} else {
api.details('tool-result-block', summary, txt);
api.details('tool-result-block', summaryBody, txt);
}
}
// Pretty-render claude's background-task subagent events
@ -783,12 +892,19 @@
}
if (v.type === 'assistant' && v.message && v.message.content) {
for (const c of v.message.content) {
if (c.type === 'text' && c.text && c.text.trim()) api.row('text', c.text);
if (c.type === 'text' && c.text && c.text.trim()) {
// Assistant prose renders with markdown — claude often
// emits bullets / fenced code / inline code; raw text
// loses the structure.
const row = api.row('text', '');
row.appendChild(mdNode(c.text));
}
else if (c.type === 'thinking') {
const txt = (c.thinking || c.text || '').trim();
api.row('thinking', txt ? '· ' + txt : '· thinking …');
}
else if (c.type === 'tool_use') {
if (c.id && c.name) toolNameById.set(c.id, c.name);
if (!renderRichToolUse(c, api)) {
api.row('tool-use', '→ ' + fmtToolUse(c));
}
@ -802,7 +918,10 @@
}
return;
}
api.row('sys', '· ' + trim(JSON.stringify(v), 200));
// Catch-all for unrecognised stream-json shapes. Loud (orange) so
// silently-dropped event types surface in the scrollback for
// follow-up classification.
api.row('sys', '! ' + trim(JSON.stringify(v), 200));
}
// Count open turns across the backfill replay so the live banner +
@ -844,7 +963,22 @@
(ev.ok ? '✓' : '✗') + ' turn ' + (ev.ok ? 'ok' : 'fail')
+ (ev.note ? ' — ' + ev.note : ''));
},
note(ev, api) { api.row('note', '· ' + ev.text); },
note(ev, api) {
const t = String(ev.text || '');
// stderr lines coming off the claude pump get an orange `!`
// glyph so they're not visually fused with ambient harness
// chatter. Operator-initiated notes (/cancel, /compact,
// /model, new-session) get a mauve italic affordance so the
// scrollback distinguishes "the operator did this" from
// "the harness did this on its own."
if (t.startsWith('stderr:')) {
api.row('note stderr', '! ' + t);
} else if (t.startsWith('operator:')) {
api.row('note op', '· ' + t);
} else {
api.row('note', '· ' + t);
}
},
stream(ev, api) {
const v = Object.assign({}, ev); delete v.kind;
renderStream(v, api);

View file

@ -39,6 +39,7 @@
<div id="term-input" class="term-input"></div>
</div>
<script src="/static/marked.js" defer></script>
<script src="/static/hive-fr0nt.js" defer></script>
<script src="/static/app.js" defer></script>
</body>

View file

@ -74,6 +74,11 @@ async fn main() -> Result<()> {
let login_state = Arc::new(Mutex::new(initial));
let bus = Bus::new();
let stats = TurnStats::open_default();
if let Some(s) = &stats
&& let Some(u) = s.last_usage()
{
bus.seed_usage(u);
}
let files = turn::TurnFiles::prepare(&cli.socket, &label, mcp::Flavor::Agent).await?;
let turn_lock: TurnLock = Arc::new(tokio::sync::Mutex::new(()));
plugins::install_configured(&cli.socket, Some("manager")).await;

View file

@ -64,6 +64,11 @@ async fn main() -> Result<()> {
let login_state = Arc::new(Mutex::new(initial));
let bus = Bus::new();
let stats = TurnStats::open_default();
if let Some(s) = &stats
&& let Some(u) = s.last_usage()
{
bus.seed_usage(u);
}
let files = turn::TurnFiles::prepare(&cli.socket, &label, mcp::Flavor::Manager).await?;
let turn_lock: TurnLock = Arc::new(tokio::sync::Mutex::new(()));
plugins::install_configured(&cli.socket, None).await;

View file

@ -378,6 +378,15 @@ impl Bus {
self.emit(LiveEvent::ModelChanged { model: value });
}
/// Seed `last_usage` at startup without emitting a SSE event.
/// Used by the bin entrypoints to backfill from the most recent
/// `turn_stats` row so the per-agent web UI's `ctx-badge` paints
/// real numbers on cold load instead of staying empty until the
/// next turn finishes.
pub fn seed_usage(&self, usage: TokenUsage) {
*self.last_usage.lock().unwrap() = Some(usage);
}
/// Record the latest token usage from a completed turn.
pub fn record_usage(&self, usage: TokenUsage) {
*self.last_usage.lock().unwrap() = Some(usage);

View file

@ -156,6 +156,34 @@ impl TurnStats {
tracing::warn!(error = ?e, "turn_stats: insert failed");
}
}
/// Token counts from the most recently inserted row, if any. Lets
/// the harness seed `Bus::last_usage` on startup so the per-agent
/// web UI's `ctx-badge` paints with real numbers on cold load
/// instead of waiting for the next `TokenUsageChanged` SSE event.
/// Best-effort: any sqlite error returns `None` and the caller
/// falls back to the empty state.
#[must_use]
pub fn last_usage(&self) -> Option<crate::events::TokenUsage> {
let conn = self.inner.lock().unwrap();
conn.query_row(
"SELECT input_tokens, output_tokens,
cache_read_input_tokens, cache_creation_input_tokens
FROM turn_stats
ORDER BY started_at DESC
LIMIT 1",
[],
|row| {
Ok(crate::events::TokenUsage {
input_tokens: u64::try_from(row.get::<_, i64>(0)?).unwrap_or(0),
output_tokens: u64::try_from(row.get::<_, i64>(1)?).unwrap_or(0),
cache_read_input_tokens: u64::try_from(row.get::<_, i64>(2)?).unwrap_or(0),
cache_creation_input_tokens: u64::try_from(row.get::<_, i64>(3)?).unwrap_or(0),
})
},
)
.ok()
}
}
fn default_path() -> PathBuf {

View file

@ -94,6 +94,7 @@ pub async fn serve(
.route("/static/agent.css", get(serve_css))
.route("/static/app.js", get(serve_app_js))
.route("/static/hive-fr0nt.js", get(serve_shared_js))
.route("/static/marked.js", get(serve_marked_js))
.route("/api/state", get(api_state))
.route("/events/stream", get(events_stream))
.route("/events/history", get(events_history))
@ -190,6 +191,13 @@ async fn serve_shared_js() -> impl IntoResponse {
)
}
async fn serve_marked_js() -> impl IntoResponse {
(
[("content-type", "application/javascript")],
hive_fr0nt::MARKED_JS,
)
}
#[derive(Serialize)]
struct StateSnapshot {
/// Bus seq at the moment this snapshot was assembled. Clients dedupe

View file

@ -668,6 +668,10 @@ summary:hover { color: var(--purple); }
align-items: baseline;
gap: 0.5em;
padding: 0.1em 0;
/* Override the per-agent-terminal's hanging-indent metrics from
TERMINAL_CSS the dashboard's broker rows are flex grids, not
glyph-prefixed text, and don't want the prefix column. */
text-indent: 0;
}
.live .msgrow .msg-body {
flex: 1 1 0;

6
hive-fr0nt/assets/marked.min.js vendored Normal file

File diff suppressed because one or more lines are too long

View file

@ -60,41 +60,54 @@
from { opacity: 0; transform: translateY(4px); }
to { opacity: 1; transform: translateY(0); }
}
/* Unified prefix column for every row kind. The glyph (` · !`)
is the first character of the row's text content; `padding-left` reserves
the column and `text-indent: -1.4em` pulls the glyph back into it. Wrapped
continuation lines then start under the body, not under the glyph, so
wraps don't blur into the next row. `details.row` summaries reuse the
same metrics below. */
.live .row {
white-space: pre-wrap;
word-break: break-word;
padding: 0.05em 0;
line-height: 1.45;
border-left: 2px solid transparent;
padding-left: 0.5em;
padding-left: 1.9em;
text-indent: -1.4em;
margin: 0.1em 0;
}
.live .row + .row { border-top: 0; }
/* Row-kind colours. Pages register renderers that emit these classes;
any class no page emits is just dead CSS, which is fine. */
.live .turn-start {
color: var(--amber);
font-weight: bold;
margin-top: 1em;
border-left-color: var(--amber);
padding-top: 0.3em;
}
.live .turn-start:first-child { margin-top: 0; }
.live .turn-body {
color: var(--fg);
font-weight: normal;
margin-top: 0.15em;
padding-left: 1.2em;
opacity: 0.85;
}
.live .turn-end-ok { color: var(--green); border-left-color: var(--green); margin-bottom: 0.4em; }
.live .turn-end-fail { color: var(--red); border-left-color: var(--red); margin-bottom: 0.4em; }
.live .text { color: var(--fg); padding-left: 1.2em; }
.live .thinking { color: var(--muted); font-style: italic; padding-left: 1.2em; }
.live .tool-use { color: var(--cyan); padding-left: 1.2em; }
.live .tool-result { color: var(--muted); padding-left: 1.2em; }
.live .result { color: var(--green); padding-left: 0.5em; }
.live .sys, .live .note { color: var(--muted); }
any class no page emits is just dead CSS, which is fine. Turn-framing
classes carry their signal entirely on the coloured border-left rule
no bold, no top/bottom margins, no background tint. The chrome was
overweight for what's just a "this is a boundary" marker. */
.live .turn-start { color: var(--amber); border-left-color: var(--amber); }
/* turn-body is a child block under turn-start carrying the wake-prompt
body; reset text-indent so wrapped content stays under its own column
instead of pulling back into the parent's prefix. */
.live .turn-body { color: var(--fg); text-indent: 0; margin-top: 0.15em; }
/* Any child block (markdown body, nested details) resets the parent
row's hanging indent so the content lays out from column 0 of the
body area. */
.live .row .md, .live .row > details { text-indent: 0; }
.live .turn-end-ok { color: var(--green); border-left-color: var(--green); }
.live .turn-end-fail { color: var(--red); border-left-color: var(--red); }
.live .text { color: var(--fg); }
.live .thinking { color: var(--muted); font-style: italic; }
.live .tool-use { color: var(--cyan); }
.live .tool-result { color: var(--muted); }
.live .result { color: var(--green); }
.live .note { color: var(--muted); }
/* Distinguish stderr lines (orange) and operator-initiated notes
(mauve, lightly emphasised) from ambient harness chatter so the
eye picks out anomalies + operator actions in the scrollback. */
.live .note.stderr { color: var(--amber); }
.live .note.op { color: var(--purple); font-style: italic; }
/* The .sys catch-all fires when renderStream landed an event shape it
couldn't classify. Make it visually loud so silently-dropped event
types surface for follow-up. */
.live .sys { color: var(--amber); }
.live .unread-badge {
color: var(--amber);
font-weight: normal;
@ -137,28 +150,31 @@
pointer-events: auto;
}
.tail-pill:hover { filter: brightness(1.1); }
/* Expandable rows reuse the flat-row prefix metrics (padding-left +
negative text-indent) so the disclosure glyph (` / `) lands in
exactly the same column as flat-row prefix glyphs (` · `).
Summary text omits the per-row directional glyph (the row colour
already carries cyan = outbound tool, muted = inbound result) so
the prefix column doesn't have to fit two glyphs side-by-side. */
details.row {
white-space: normal;
padding-left: 0.5em;
}
details.row > summary {
cursor: pointer;
color: var(--muted);
list-style: none;
white-space: pre-wrap;
word-break: break-word;
}
details.row > summary::before {
content: '▸ ';
color: var(--muted);
display: inline-block;
width: 1em;
color: inherit;
}
details.row[open] > summary::before { content: '▾ '; }
details.row.tool-result-block > summary { color: var(--muted); }
details.row > pre.diff-body {
margin: 0.3em 0 0.4em 1.2em;
details.row > pre.diff-body,
details.row > pre.tool-body {
margin: 0.3em 0 0.4em 0;
padding: 0.4em 0.6em;
text-indent: 0;
background: rgba(255, 255, 255, 0.02);
border-left: 2px solid var(--purple-dim);
white-space: pre-wrap;
@ -166,17 +182,44 @@ details.row > pre.diff-body {
max-height: 22em;
overflow-y: auto;
}
details.row > pre.tool-body { color: var(--fg); }
details.row > pre.diff-body .diff-add { color: var(--green); }
details.row > pre.diff-body .diff-del { color: var(--red); }
details.row > pre.diff-body .diff-ctx { color: var(--fg); }
details.row > pre.tool-body {
margin: 0.3em 0 0.4em 1.2em;
/* Markdown body inside a row (assistant text, send/recv/ask/answer
message bodies). Inline elements get muted accents; block elements
reset the parent row's hanging indent so content lays out cleanly. */
.live .row .md p { margin: 0.2em 0; }
.live .row .md p:first-child { margin-top: 0; }
.live .row .md p:last-child { margin-bottom: 0; }
.live .row .md code {
background: rgba(255, 255, 255, 0.06);
padding: 0.05em 0.3em;
border-radius: 3px;
font-size: 0.95em;
}
.live .row .md pre {
margin: 0.3em 0;
padding: 0.4em 0.6em;
background: rgba(255, 255, 255, 0.03);
background: rgba(255, 255, 255, 0.04);
border-left: 2px solid var(--purple-dim);
color: var(--fg);
text-indent: 0;
white-space: pre-wrap;
word-break: break-word;
max-height: 22em;
overflow-y: auto;
}
.live .row .md pre code {
background: transparent;
padding: 0;
border-radius: 0;
}
.live .row .md a { color: var(--cyan); text-decoration: underline; }
.live .row .md strong { color: inherit; font-weight: bold; }
.live .row .md em { color: inherit; font-style: italic; }
.live .row .md ul, .live .row .md ol { margin: 0.2em 0 0.2em 1.4em; padding: 0; }
.live .row .md li { margin: 0.05em 0; }
.live .row .md blockquote {
margin: 0.2em 0;
padding-left: 0.6em;
border-left: 2px solid var(--purple-dim);
color: var(--muted);
}

View file

@ -32,3 +32,11 @@
pub const BASE_CSS: &str = include_str!("../assets/base.css");
pub const TERMINAL_CSS: &str = include_str!("../assets/terminal.css");
pub const TERMINAL_JS: &str = include_str!("../assets/terminal.js");
/// Vendored [marked](https://github.com/markedjs/marked) v4.0.2 UMD bundle.
/// Exposes a global `marked` with `.parse(src, opts)` returning HTML. Used
/// by the per-agent terminal to render message bodies (`send` / `recv` /
/// `ask` / `answer` / assistant text) with basic markdown — paragraphs,
/// lists, fenced + inline code, bold/italic, links. Vendored rather than
/// CDN-loaded so the per-agent page works on operator machines without
/// internet egress (the container itself never fetches it).
pub const MARKED_JS: &str = include_str!("../assets/marked.min.js");