vinzenz/hyperhive
vinzenz/hyperhive
Watch
0
0
Fork
You've already forked hyperhive
 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

fix(dashboard): ctx·Nk card badge showed 0k — use context_tokens()

Browse source 
/api/dashboard-state reported ctx_tokens from raw ctx_usage.input_tokens,
which is only the *uncached* input. With prompt caching the bulk of the
prompt is cache-read, so input_tokens is ~0 and every SW4RM card showed
'ctx·0k'. Use ctx_usage.context_tokens() (input + cache-read + cache-creation
= the real window size) to match the agent page (ships the full ctx_usage)
and turn.rs's cache-TTL check. Doc comment updated.
This commit is contained in:
iris 2026-06-21 21:27:28 +02:00 • committed by mara
commit 5eacc99146
1 changed files with 9 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

12
hive-ag3nt/src/web_ui.rs
View file

@ -664,8 +664,9 @@ struct DashboardState {
/// `None` when no status is set. /// `None` when no status is set.
#[serde(skip_serializing_if = "Option::is_none")] #[serde(skip_serializing_if = "Option::is_none")]
status_set_at: Option<i64>, status_set_at: Option<i64>,
/// Input token count from the most recent completed turn (`ctx_usage.input_tokens`). /// Full context-window size from the most recent completed turn
/// `None` until the first turn finishes. /// (`ctx_usage.context_tokens()` = input + cache-read + cache-creation).
/// `None` until the first turn finishes. Drives the `ctx·Nk` card badge.
#[serde(skip_serializing_if = "Option::is_none")] #[serde(skip_serializing_if = "Option::is_none")]
ctx_tokens: Option<u64>, ctx_tokens: Option<u64>,
/// Effective context-window budget for the current model. Same /// Effective context-window budget for the current model. Same
@ -714,7 +715,12 @@ async fn api_dashboard_state(State(state): State<AppState>) -> axum::Json<Dashbo
.bus .bus
.api_context_window() .api_context_window()
.unwrap_or_else(|| crate::events::context_window_tokens(&model)); .unwrap_or_else(|| crate::events::context_window_tokens(&model));
let ctx_tokens = state.bus.last_ctx_usage().map(|u| u.input_tokens); // Full context-window size = input + cache-read + cache-creation. Using
// raw `input_tokens` here reported only the *uncached* sliver, which is
// ~0 once prompt caching kicks in — so every card showed `ctx·0k`. Match
// the agent page (which ships the whole `ctx_usage` and sums it) and the
// cache-TTL logic in turn.rs, both of which use `context_tokens()`.
let ctx_tokens = state.bus.last_ctx_usage().map(|u| u.context_tokens());
axum::Json(DashboardState { axum::Json(DashboardState {
status_text, status_text,
status_set_at, status_set_at,