Populars: Dades actualitzades el 21 de novembre del 2026

Categories

Internet1.421
Comunicació329
Àudio i vídeo938
Gràfics i disseny566
Ofimàtica i productivitat1.184
Eines per a desenvolupadors3.288
Eines d’IA610
Seguretat i privadesa408
Eines del sistema3.194
Jocs545
Educació i ciència242
Xarxa i TI294

Baixar cc-sessions-viewer 0.1.3 per a Windows

Support cc/codex/gemini sessions viewer, token usage statistics, global search, resume, and export to html

Aquesta és una versió anterior. La versió actual és la 0.6.3. cc-sessions-viewer 0.6.3 · Data de publicació:

Fitxers d’instal·lació 0.1.3

ArquitecturaTipusÀmbitMida del fitxerBaixaSHA256
64 bits EXE (NSIS) Usuari actual — github.com VirusTotal
64 bits MSI Tots els usuaris — github.com VirusTotal

L’enllaç i l’SHA256 provenen del manifest comunitari del Gestor de paquets del Windows. Convé comparar el hash del fitxer baixat abans d’executar-lo.

Notes de la versió

Added
- Embedded terminal — "Open in window" — resume or start a session inside the app window via an xterm.js terminal backed by portable-pty, instead of shelling out to Terminal.app. Chat header gains List / View / per-session terminal tabs.
- macOS menu-bar tray + close-to-tray — closing the window hides to a tray icon (Show / Statistics / Settings / Quit) instead of quitting; ⌘Q still exits.
- Lossless JSON export + Export history view — export a session to a re-importable JSON envelope; a sidebar view lists past exports (capped at 50, dedup by original path) that reopen the original transcript.
- "This month" stats range — separate from "Last 30 days"; matches calendar-month accounting tools and stays honest in the first week of the month.
- Auto display-name for new models — claude-… / gpt-… / gemini-… IDs render as "Opus 4.9" / "GPT-5.6 Codex" / "Gemini 3 Pro" by pattern, so brand-new versions display correctly with no table edit.
- Live model pricing view + sidebar "More" menu — kebab dropdown next to the trash icon holds Export history and a new Live model pricing page. Lists Claude / Codex / Gemini rates from LiteLLM upstream (Context / Input / Output / Cache read / Cache write columns), sticky search box (Enter to filter) + brand-icon anchor chips for jumping between families, Refresh button with a Stats-style full-area loader. Sorted newest version first (version-tuple based, tier names ignored); filters noise (@default aliases, gpt-oss-*, image / audio / realtime / transcribe / search-preview variants, gpt-35-* Azure dupes).
- Markdown rendering for chat text — GFM tables + Mermaid diagrams — renderText now parses GFM tables (header + separator + body, with :-- / --: / :--: alignment, inline formatting in cells, horizontal scroll wrapper for wide tables) and mermaid fenced blocks (lazy-loaded mermaid.js, light/dark theme-reactive, error fallback shows source + error). Applies to all three agents through the single ChatView text path; HTML export bakes the SVG in at export time so the offline file renders without a runtime mermaid dep.
- JSON syntax highlighting in tool calls and JSON tool results — tool_use args (always JSON) are pretty-printed and colorized (key / string / number / bool / null tokens). Tool result text that looks like JSON (strips cat -n line numbers first to detect) gets the same treatment in place — original layout preserved, just tokens colored. Color tokens go light/dark via separate hues. Applies to both the chat view (via ChatView.vue + ToolResult.vue) and the HTML export. Detection is relaxed enough to still color the truncated JSON that Read with limit:N returns (the file is cut mid-object, JSON.parse can't succeed, but the "key": pattern survives).
- Unified-diff syntax highlighting in tool results — text-form git diff / patch output in tool_result (e.g. Bash running git diff) now renders with row-level coloring: file headers, index / --- a/x / +++ b/x metadata, @@ -m,n +p,q @@ hunk headers, and + / - / context rows each get their own class. Existing structured-patch results (Claude structuredPatch) keep going through the interactive DiffBlock component — this only affects the text path that used to render as a plain <pre>. Detection order is diff-before-JSON so a patch on a JSON file colors as diff, not as malformed JSON. Mirrored to the HTML export.
- Codex session filtering — internal & archived — reads Codex's SQLite threads table to tag internal / archived sessions, queries the Codex app-server via JSON-RPC for ranking metadata, and scans ~/.codex/archived_sessions/ when archived visibility is enabled. Session list shows rank, "review session", and "archived" pill badges. Two new toggles in settings: "Show internal sessions" and "Show archived sessions" (archived defaults to on).
- Codex (blue-toned light) & Dracula (classic dark) themes — two new theme presets alongside the existing light / dark / system. CSS variables now cover font sizes and diff colors for full theme adaptability. Codex / Dracula items also appear in the native macOS menu bar.
- "Use external terminal" setting — toggle in Settings → Advanced to resume / start sessions in the system terminal (macOS Terminal.app, Windows PowerShell/cmd, Linux gnome-terminal/konsole/xterm) instead of the built-in xterm.js. Off by default (embedded terminal is the default).
- Settings redesign — General / Advanced tabs — settings split into two tabs: General (language, theme, data, about) and Advanced (terminal, Codex). Language and theme pickers replaced with custom Geist-style dropdown menus; Codex / terminal toggles use compact iOS-style switches instead of oversized card buttons.
Changed
- "All time" stats range → "Last 6 months" — unbounded scans were slow and rarely useful; bounded to 6 calendar months back. Stale 'all' in localStorage silently migrates to 'months6'.
- Model pricing now live from LiteLLM — replaced the hand-curated table with a runtime fetch from LiteLLM (24h disk cache, retry on failure). New models price automatically on next launch.
- Sidebar refresh button moved next to {agent} · N projects — was on the topbar far from the agent switcher; now sits on the agent/count row so "refresh this agent" reads more naturally. Only the active agent reloads; other agents stay untouched.
- Batch select / delete / export controls moved into list body — previously lived in the topbar, causing visual overlap with other icon rows. Now inline in the session-list and trash-list headers, closer to the content they act on.
- Single-session stats now fold Claude sub-agent JSONLs into the parent — opening "Session stats" from a parent now feeds <parent>/subagents/*.jsonl into the same Aggregator, so cost / calls match the global by-session leaderboard row. Codex / Gemini unchanged (no sub-agent concept).
- HTML export now renders GFM tables, inline markdown, and Mermaid diagrams — previously was escapeHtml + <br> with no markdown. Now runs through renderText; mermaid SVGs are baked in at export time (one-shot, current-theme color), so the exported .html stays self-contained and offline-readable.
- Release notes built from conventional commits — changelogithub groups feat/fix/perf commits and adds a contributors footer.
- Conventional commit types lowercased — Feat → feat, etc., so changelogithub groups correctly.
- Dev-only MCP Bridge in debug builds — lets an AI assistant screenshot, snapshot the DOM, run JS, and watch IPC against the running app.
Fixed
Stats accuracy — all numbers now reconcile with codeburn across every range:
- Gemini cost ~2× too high and calls ~70% too high — tokens.input is the total prompt size including cached tokens; we billed it at input rate and billed cached at cache-read rate, double-charging the cached portion. Fix: input_tokens = totalInput − cached. Also skip streaming sub-events (no tokens / no model / all-zero) that aren't independently billed.
- Claude session count inflated by ~50% — subagent JSONLs (<project>/<parent>/subagents/agent-*.jsonl) counted as standalone sessions in Stats but not in the sidebar. Fix: fold them onto the parent session id. Cost / tokens unchanged.
- Anthropic 1h cache_creation underbilled ~6–8% — 1h cache writes cost 1.6× the 5-minute rate, but we read only the lump-sum field and billed everything at 5min. Fix: read the ephemeral_1h_input_tokens split, add the 0.6× premium.
- OpenAI reasoning tokens billed at $0 — reasoning_output_tokens (the hidden chain-of-thought) was dropped from cost_usd. Fix: bill at output rate. Anthropic unaffected (already folded into output_tokens).
- Codex calls ~3× inflated — every function_call / agent_message became a separate call. Fix: emit one CallRecord per event_msg.token_count event (one real API call), folding tool metadata into the next call.
- "Today" / "7 days" / "30 days" rolled by 24h, not calendar days — KPIs inflated 5–10× when a session crossed local midnight. Fix: switch to local calendar-day boundaries.
- Range filter only at session level — entire sessions counted toward "Today" if the file's mtime fell in the window, regardless of when individual turns happened. Fix: apply the window per-turn.
- Cost / token formatting — $38.55 no longer rounds to $39, 240.5K no longer rounds to 241K. Always 2 decimals for USD, always 1 decimal for tokens.
- Opus 4.8 mislabeled "Opus 4" and 3× overpriced — missing pricing-table entry fell through to the base Opus 4 tier.
- OpenAI / Gemini rate cards re-verified — fixed several wrong or placeholder rows (gpt-5.1-codex-mini, the gpt-5.2 / 5.4 / 5.5 family, gemini-3.1-pro-preview, gemini-3-flash-preview, gemini-2.5-flash-lite cache read).
- refresh_pricing Tauri command was a sync blocking call — ureq::get(...).call() ran on the main thread with a 20s timeout, freezing the webview (CSS animations, mouse cursor, everything). Now async fn + spawn_blocking — measured 0 stalls > 100ms over a 6-second refresh.
- GPT 5.x not sorted above 4.x in pricing view — gpt-oss-120b parsed "120" as a version, sorting it above gpt-5 (version [5]). Now filters non-chat variants (gpt-oss-*, image / audio / realtime / transcribe / search-preview, gpt-35-* Azure dupes) before sort.
- Horizontal scrollbar in light-mode markdown tables looked nearly black — global scrollbar styles didn't reach .md-table-wrap. Now scoped scrollbar styling (7px thin, transparent track, color-mix(--text, 22%, transparent) thumb), Firefox via scrollbar-color.

https://github.com/jerrywu001/cc-sessions-viewer/blob/HEAD/CHANGELOG.md#v013

El més popular a Assistents d’IA i LLM

Mostra-ho tot
  1. Claude CodeAssistents d’IA i LLM · Anthropic PBC
  2. Codex CLIAssistents d’IA i LLM · OpenAI, Inc.
  3. ChatGPT by lencxAssistents d’IA i LLM · lencx
  4. GPT4AllAssistents d’IA i LLM · Nomic
  5. JanAssistents d’IA i LLM · Jan
  6. oh-my-piAssistents d’IA i LLM · Can Bölük
  7. T3 CodeAssistents d’IA i LLM · T3 Tools
  8. WSL ManagerAssistents d’IA i LLM · Bostrot
  9. CC SwitchAssistents d’IA i LLM · farion1231
  10. LobeHubAssistents d’IA i LLM · LobeHub

Etiquetes

Comparteix