mirror of
https://github.com/wu736139669/hapi.git
synced 2026-08-05 06:24:37 +00:00
Junmo KimandGitHub
e45fde51e9
* refactor(claude): thread session's selected model into SDKToLogConverter
Adds an optional selectedModel field to the converter's context, wired
from session.getModel() in the launcher, so a later commit can seed the
turn-1 contextWindow estimate for presets whose system/init model
arrives without the "[1m]" suffix. No behavior change yet.
* fix(claude): key contextWindow cache by model to stop 1M/200k flicker
The remote launcher re-emits system/init on every turn for the same
converter instance. Its init-time estimate only checked whether the
model string ended in "[1m]", but current claude CLI versions strip
that suffix from system/init for some 1M presets (fable[1m] arrives as
"claude-fable-5"), so the estimate guessed 200k for them. The one
authoritative value is result.modelUsage[<model>].contextWindow, which
arrives after the heuristic has already injected 200k into that turn's
assistant message and then gets clobbered back to 200k by the very
next turn's init - producing the observed 200k<->1M oscillation in the
web status bar.
Cache the authoritative contextWindow per model id instead of a single
session-wide number, and only let system/init seed a heuristic guess
for a model that has no cached value yet, so a same-model re-init no
longer downgrades an already-learned value.
Two observed facts about the CLI's model ids drive the design:
system/init.model and the result.modelUsage keys always agree with
each other within a session (both bare for plain/fable[1m], both
suffixed for opus[1m]/sonnet[1m]), while each per-turn assistant
message reports its model bare and thus can't distinguish a 200k plain
preset from its 1M "[1m]" variant on tiers where they share a base id.
So the cache is keyed on the raw id (init/result agree, no
normalization) and assistant injection looks the value up via
resolvedModel (the last init id) rather than the lossy message.model.
Keying raw keeps a plain preset and its [1m] variant on distinct
entries; looking up via resolvedModel also means sidechain (Task
subagent) messages carry the main session window rather than the
subagent's own, since the web status bar picks the most recent usage
message without filtering sidechains and would otherwise flicker to
the subagent's smaller window while it runs.
For presets whose init model arrives bare even though they are 1M
(fable[1m]), the originally-selected preset - which preserves the
"[1m]" suffix - seeds the turn-1 estimate, kept live across mid-session
model switches via updateSelectedModel() (called from the launcher on
every turn) so it never goes stale.
* fix(web): recognize [1m] suffix on full Claude model ids in budget fallback
getContextBudgetTokens already special-cased "[1m]" for short preset
values (e.g. "opus[1m]") but fell through to the default 200k budget
for full model ids (e.g. "claude-opus-4-8[1m]"), which is what the CLI
now reports once context_window isn't available and this fallback is
consulted. Check the suffix on that branch too so it stays a correct
last-resort even without a session-provided context_window.
* refactor(web): merge duplicate Claude context-budget branches
isClaudeModelPreset(trimmedModel) and the startsWith('claude-') branch
below it had become byte-for-byte identical bodies after the [1m]
suffix check was added to both. Merge them into one condition; no
behavior change.
* fix(claude): distinguish fable from fable[1m] when the CLI reports both bare
The per-model contextWindow cache keyed on the raw system/init model id,
on the assumption that a 1M preset and its plain form always land on
distinct ids. That holds for opus[1m]/sonnet[1m] (the CLI reports the
"[1m]" suffix on their init and result ids) but not for fable: the CLI
reports both "fable" and "fable[1m]" with the same bare id
"claude-fable-5". So the "seed only if not already cached" guard would
skip re-seeding when switching fable[1m] -> fable, leaving the stale 1M
in place until fable's own result arrived - the same switch flicker this
change set out to remove, just for fable specifically.
Fold the selected preset's "[1m]" back into the cache key
(computeContextWindowKey): when the init model arrives bare but the
session selected an "[1m]" preset, key the entry as "<id>[1m]" so the 1M
and plain variants stay distinct; ids the CLI already suffixed are left
as-is. Seeding, lookups, and the current-model result entry all use this
resolved key. Subagent result entries (e.g. haiku) keep their own raw id
so the session's "[1m]" is never folded onto a model that isn't the
selected one.
hapi-web
React Mini App / PWA for monitoring and controlling hapi sessions.
What it does
- Session list with status, pending approvals, todos, and summaries.
- Chat view with streaming updates and message sending.
- Permission approval and denial workflows.
- Permission mode and model selection.
- Machine list and remote session spawn.
- File browser and git status/diff views.
- PWA install prompt and offline banner.
Runtime behavior
- When opened inside Telegram, auth uses Telegram WebApp init data.
- When opened in a normal browser, you can log in with
CLI_API_TOKEN:<namespace>(orCLI_API_TOKENfor the default namespace). - The login screen includes a top-right hub picker; if unset, the app uses the same origin it was loaded from.
- Live updates come from the hub via SSE.
Routes
See src/router.tsx for route definitions.
/- Redirect to /sessions./sessions- Session list./sessions/$sessionId- Chat interface./sessions/new- Create new session./sessions/$sessionId/files- File browser with git status./sessions/$sessionId/file- File viewer with diff support./sessions/$sessionId/terminal- Terminal interface./settings- Application settings.
Features
Session list (src/components/SessionList.tsx)
- Active/inactive status indicator.
- Session title from name, summary, or path.
- Todo progress display.
- Pending permission request count.
- Agent flavor label (claude/codex/gemini).
- Model mode display.
Chat interface (src/components/SessionChat.tsx)
- Message thread with infinite scroll.
- Composer for sending messages.
- Permission mode toggle (default/acceptEdits/auto/bypassPermissions/plan).
- Model selection (default/sonnet/sonnet[1m]/opus/opus[1m]).
- Session abort and mode switch controls.
- Context size display.
- Per-session scratchlist (
src/components/AssistantChat/ScratchlistPanel.tsx)- Workbench panel for held notes/drafts; distinct from the queue.
- Add/delete/reorder entries; promote to composer (copy) or queue (send).
- Persists across reloads via
localStoragekeyed per session. - Keyboard shortcut: Ctrl/Cmd+Shift+S to focus the add-input.
File browser (src/routes/sessions/files.tsx)
- Git status view (staged/unstaged files).
- File search with ripgrep.
- Navigate to file viewer.
File viewer (src/routes/sessions/file.tsx)
- File content display with syntax highlighting.
- Staged/unstaged diff view.
Terminal (src/routes/sessions/terminal.tsx)
- Remote terminal via xterm.js
- Real-time via Socket.IO
- Resize handling
Voice assistant
- ElevenLabs integration (@elevenlabs/react)
- Real-time voice control
New session (src/components/NewSession/)
Modular session creation:
- Machine selector
- Directory input with recent paths
- Agent type selector
- Model selector
- Permission mode toggle (YOLO mode)
Authentication
See src/hooks/useAuth.ts and src/hooks/useAuthSource.ts.
- Telegram Mini App: Uses initData from WebApp SDK.
- Browser: Uses CLI_API_TOKEN from login prompt.
- JWT tokens with auto-refresh.
Data fetching
See src/hooks/queries/ for query hooks and src/hooks/mutations/ for mutations.
- Sessions, messages, machines via TanStack Query.
- Git status and file operations.
- Optimistic updates for message sending.
Real-time updates
See src/hooks/useSSE.ts.
- SSE connection to
/api/events. - Session/message/machine update events.
- Automatic cache invalidation on events.
Stack
React 19 + Vite + TanStack Router/Query + Tailwind + @assistant-ui/react + xterm.js + @elevenlabs/react + socket.io-client + workbox + shiki.
Source structure
src/router.tsx- Route definitions.src/components/- UI components.src/hooks/- Data fetching and state hooks.src/api/client.ts- API client.src/types/api.ts- Type definitions.
Development
From the repo root:
bun install
bun run dev:web
If testing in Telegram, set:
HAPI_PUBLIC_URLto the public HTTPS URL of the dev server.CORS_ORIGINSto include the dev server origin.
Tests
Unit tests run under vitest + jsdom:
bun run test:web
End-to-end browser tests for the scratchlist component (real Chromium, real
inert focus blocking, real localStorage round-trips) live at the repo root
under e2e/:
bun run test:e2e # headless
bun run test:e2e:ui # Playwright UI mode (debug)
The spec drives a Vite-served fixture page (web/e2e-fixtures/scratchlist-fixture.html)
that mounts the production ScratchlistPanel in isolation, so no hub /
auth / socket setup is required.
Build
bun run build:web
The built assets land in web/dist and are served by hapi-hub. The single executable can embed these assets.
Standalone hosting
You can host web/dist on a static host (GitHub Pages, Cloudflare Pages) and point it at any hapi hub:
- Build the web app. If your static host uses a subpath, set the Vite base:
bun run build:web -- --base /<repo>/
- Deploy
web/distto your static host. - Set hub CORS to allow the static origin (
HAPI_PUBLIC_URLorCORS_ORIGINS). - Open the static site, click the top-right Hub button on the login screen, and enter the hapi hub origin.
Clear the hub override in the same dialog to return to same-origin behavior.