EILMELDUNGEN LIVE
🔧 AI Nachrichten Generate playable HTML5 games with Mistral Medium and GLM(28.08.2026 um 16:00 Uhr)
🔧 AI Nachrichten Claude Fable 5.1 Just Set a New AI Performance Record(02.09.2026 um 11:00 Uhr)
🔧 AI Nachrichten Sam Altman Says This AI Agent Could Run Forever(02.09.2026 um 14:00 Uhr)
🔧 AI Nachrichten New Qwen 3.8 Max Update Is SCARY GOOD!(02.09.2026 um 18:00 Uhr)
🔧 AI Nachrichten Generate playable HTML5 games with Mistral Medium and GLM(28.08.2026 um 16:00 Uhr)
🔧 AI Nachrichten Claude Fable 5.1 Just Set a New AI Performance Record(02.09.2026 um 11:00 Uhr)
🔧 AI Nachrichten Sam Altman Says This AI Agent Could Run Forever(02.09.2026 um 14:00 Uhr)
🔧 AI Nachrichten New Qwen 3.8 Max Update Is SCARY GOOD!(02.09.2026 um 18:00 Uhr)

17 🕛 kürzlich 5 Min Lesezeit CVE-RADAR
0

GitHub Release: cline/cline vsdk/sdk/v0.0.82 (02.09.2026)

↗ Quelle (GitHub · cline/cline)
🗣️ Stimme:
GitHub Release: cline/cline vsdk/sdk/v0.0.82 (02.09.2026)
Avatar
$ git clone https://github.com/cline/cline.git

  • Fixed tool calling being silently disabled for gateway models whose catalog entry declares no capabilities. Three separate producers built gateway model definitions with their own hand-written capability translations and had drifted; the builtin-provider path emitted ["text"] where the others emitted undefined, and that list read as an authoritative denial that stripped every tool definition from requests to Dify, SAP AI Core, opencode, and the Codex CLI. One shared translator now serves every producer

  • Fixed an empty capability list stripping image input. Two readers bypassed modelHasCapability and tested capabilities directly, where an empty list is not nullish but matches nothing: the file-read tool silently dropped every image from the request, and model pickers reported models as definitively lacking vision, attachments, and reasoning when nothing had been declared

  • Langfuse tracing now works in minified release builds. initializeLangfuseTelemetry identified the OpenTelemetry provider by constructor name, which minification renames, so every production build silently initialized with readiness=false while unminified dev runs worked. Detection is now structural, and a foreign provider already holding the global slot gets the Langfuse span processor attached rather than being silently bailed on. Langfuse also moved to AI SDK 7 telemetry

  • New SessionImportService discovers and imports conversation history from Claude Code (~/.claude/projects), Codex (~/.codex/sessions), and opencode (opencode.db), translating each into Cline's native message format and persisting it as a completed, listable, resumable session. Imports are transactional, idempotent per source, and sanitized so the transcript is valid to resume against any provider

  • The hub is never retired out from under running work. upgradeManagedHub and retireIncompatibleHub now establish the drain barrier before any activity check, and refuse to replace a hub that does not accept the drain — previously an idle snapshot could authorize a shutdown that killed a session admitted moments later

  • Restoring a checkpoint now refuses to run when HEAD has moved past the commit the checkpoint was taken on. The restore's git reset --hard silently knocked any later commits off the branch, leaving them reachable only through the reflog; the branch move is now a compare-and-swap so a commit landing mid-restore cannot be lost either

  • Remote (SSE/streamable HTTP) MCP servers now get a 10s connect budget. An unreachable but enabled server previously spent the full 60s request timeout inside session.create, blowing past the hub's 30s cap and tearing down the whole session

  • apply_patch now preserves a file's own CRLF line endings across updates

  • Global rules are now also discovered at ~/Cline/Rules. The VS Code Rules tab resolves the Documents folder with xdg-user-dir DOCUMENTS, which prints bare $HOME on WSL and headless installs, so rules written there never reached the system prompt

  • OpenAI Codex (ChatGPT subscription) sign-in now fails fast with an actionable error when callback port 1455 is already in use, instead of opening a browser to a flow whose callback can never arrive. OAuth redirect errors such as access_denied are surfaced rather than collapsed into "Missing authorization code"

  • A transient token-refresh failure no longer logs out Codex and OpenAI-compatible-account users. Only a rejected refresh token now forces re-authentication; network errors, timeouts, and server 5xx rethrow instead

  • Aborting a session now propagates to the work it spawned — delegated subagents and teammates are cancelled with it, and aborted teammate tasks persist as cancelled instead of staying open

  • Agent-created schedules are now anchored in ~/.cline/schedules rather than inheriting whichever chat workspace they happened to be created in, where they were invisible to workspace-scoped listings and tied to folders that may be cleaned up. Schedules created with an explicit workspace are unchanged

  • Fixed hub-managed schedules being wiped by cron reconciliation on hub restart

  • Scheduled sessions now carry their schedule's identity — id, name, cron run id, and a stable 1-based run number that a later cancellation never shifts — so clients can group runs under the schedule that produced them

  • Token-authenticated hub connections can now request cross-workspace schedule access, so a host can list schedules that live outside the workspace it registered. Workspace-bound clients stay scoped

  • Session history is now searchable through a server-side full-text index, exposed as a hub command, instead of clients loading every session to filter locally

  • Cline provider models now load from the live catalog rather than only the bundled one, so newly published models appear without a package update. Explicit runtime and models.json overrides still take precedence

  • The message the model receives when a user rejects a tool call now names the rejected tool and reads as a user decision rather than an error

  • Provider 401/403 responses are now classified as an auth error class on the existing errorClass plumbing, so hosts can render credential rejections as actionable guidance instead of a raw provider body

  • The provider catalog now exposes a computed configured flag, so a settings entry seeded with only a default model and no credentials no longer reads as a connected provider

  • Tool results can now carry media: images returned by a tool are extracted and exposed as attachments instead of raw base64 text

  • The hub no longer creates filesystem watchers on agenda spec directories while the todo tool is disabled, and updateTask reconciles external spec edits itself, so a spec edited on disk no longer fails every subsequent update with a signature mismatch

  • Refreshed the model catalog. Adds ten providers (Bothub, OpenReason, SenseNova (China), TokenGo, TokenRouter, Vancine, Volcengine Ark, Volcengine Ark Coding Plan, above.dev, and klokintegration.se) and updates model lists and pricing across providers. This is an unusually wide refresh: the resolved default model changes for 57 providers. Most consequentially, Anthropic now resolves to Claude Fable 5.1 instead of Claude Opus 5, and Amazon Bedrock, Vertex, OpenRouter, Vercel AI Gateway, Kilo Gateway, LLM Gateway, DevPass, DigitalOcean, CrossModel, Eden AI, and NanoGPT follow it to Fable 5.1. If you use any provider without pinning a model, expect a different default


Full Changelog: sdk/sdk/v0.0.81...sdk/sdk/v0.0.82

Vollständiger Original-Bericht
Ausführliche Details, Code-Beispiele & Hersteller-Stellungnahme auf github.com.
↗ Original-Artikel auf github.com lesen
Wie bewertest du diesen Beitrag?
1 Klick Feedback
Teilen mit Netzwerk & Team:

Community-Analysen & Experten-Meinungen 0

Verfasse deine eigene Analyse, teile Workarounds oder diskutiere diesen Vorfall im Blog.
Noch keine Community-Analyse verfasst. Markiere einen Textabschnitt oder klicke oben auf Eigene Analyse verfassen“!
Community Pulse: Relevanz-Einschätzung
1 Klick Experten-Votum
🔴 Akute Relevanz 58%
🟡 In Evaluierung 23%
🟢 Keine Auswirkung 10%
Spannende Innovation 9%
Verwandte Story-Cluster & Quellen (Vektor-KI)
Port 8095 Engine
1 Quelle
Generate playable HTML5 games with Mistral Medium and GLM
1 Quelle
New AI waifus, new Deepseek, realtime worlds, Happy Shrimp, tiny TTS: AI NEWS
1 Quelle
Ox Alpha reveal, realtime Minimax, Qwen Next, Hy4, robot olympics: AI NEWS