Repository navigation
2.1.0-GA merge back to main - #16
Merged
Merged
Conversation
knownSDKReleases in Components and the 15 SDK examples is now 2.0.0-GA (previous) + 2.1.0-GA; defaultSDKVersion = "2.1.0-GA"; 1.0.0-GA dropped (jevdk and mcp-chat gain 2.1.0-GA next to 2.0.0-GA). Each module's checksum is the v2.1.0-GA release's .sha256, verified against the downloaded zips (Core 8ca2eda6, Inference 21ce035b, Remote 99e6c240, Claude 5dc5bc59). The examples READMEs that name the default version now say 2.1.0-GA; examples/README.md's knownSDKReleases / .sha256 / LOCALLM_SDK_VERSION samples too. CHANGELOG: 2.1.0 GA, fixes only (Inference bundle ID lab.locallm.sdk.inference, not mlx-swift_Cmlx, which App Store Connect rejects). docs/api-surface.md from the published 2.1.0-GA binaries: surface identical to 2.0.0's. Built: Components, MCPAppsHost and every example with swift build (fresh scratch path), VistaNova with xcodebuild. check-public-hygiene passes. Co-Authored-By: Claude <noreply@anthropic.com>
The DMG doesn't carry Apple's managed PCC entitlement, so 2.1.0's DMG has no PCC; the App Store version, not yet reviewed, does. Co-Authored-By: Claude <noreply@anthropic.com>
Folds the runtime bump into 2.1.0 GA (not announced yet): Gemma 4 array arguments, gpt-oss Harmony and Muse-Glimmer tool calls, and a malformed tool call now fails the turn. The Inference checksum listed is still the first 2.1.0-GA build's; it changes when the Inference xcframework is rebuilt on the new pin. tested-models.md points to the maintained model list.
The 2.1.0-GA xcframeworks were re-published on 2026-10-15 from locallmlab-sdk 16e3f39: mlx-swift-lm 30d755c (upstream 8ec6a2c plus our fix for Qwen3 calls in an unclosed <think>) and the bridge draining generation before failing a turn on a rejected tool call. Core 48c45168, Inference db452ed4, Remote b07318d7, Claude 95859239, each verified against the downloaded zip; Inference's bundle ID is lab.locallm.sdk.inference. Replaces the earlier 2.1.0-GA checksums in Components, the 15 SDK examples, examples/README.md and the CHANGELOG; the CHANGELOG's MLX entry now names 30d755c and both fixes. docs/api-surface.md regenerated from the final binaries: surface unchanged. Built: Components, MCPAppsHost and every example with swift build (fresh scratch path, SwiftPM artifact cache flushed), VistaNova with xcodebuild. code-buddy's doc-comment task on Qwen3-8B: 6 of 6 runs complete on the published binaries. Co-Authored-By: Claude <noreply@anthropic.com>
…G and SDK xcframeworks.
…tool-call errors §10e (new): what uploading LocalLM Lab and JevDK to App Store Connect turned up. Embed LocalLMLabSDKInference from 2.1 as is (2.0's mlx-swift_Cmlx bundle ID fails with 90277; drop a copied mlx-swift_Cmlx.bundle, a duplicate bundle ID can fail with 90685), with a check for invalid bundle IDs. Private Cloud Compute: a managed capability on the App ID, signed only when the profile grants it, run in the main executable, and gated on the process's own entitlement, since FoundationModels traps without it (availability and probe included). §10's intro and §10d's App ID step point there, and the PCCModelProvider row no longer suggests probe() can detect a missing entitlement. MLXModelProvider: since 2.1 a tool call the runtime can't accept fails the turn with LocalLMLabError.generation instead of returning the markup as the answer; calls inside reasoning stay inert except a <think> that opens on the call and never closes (Qwen3). Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: Claude <noreply@anthropic.com>
…bility probe GLM-4-0414 tool results, Llama 3.1 multi-call turns, Mistral-Small-3.2's missing tool template, template choice when tools are requested, and a capped, time-limited capabilityProbe. SDK commits ebc088c and f6eac83 on ancientcomputing/locallmlab-sdk 2.1.0-GA. Co-Authored-By: Claude <noreply@anthropic.com>
The template-fitting fixes (GLM-4-0414 observation role, Llama 3.1 one call per turn, the built-in Mistral tool template, tool-capable template preference) and the bounded capability probe ship in the republished 2.1.0-GA, so they move under its Fixed section.
The 2.1.0-GA xcframeworks were re-published on 2026-10-17 with the MLX tool-use fixes (locallmlab-sdk f6eac83: chat-template fitting, the capability-probe cap, subfolder downloads) and Inference rebuilt from 8e5d287 without the test target's fixtures bundle. Core 8d541e06, Claude f54472c2, Remote 34725717, Inference 77d3d1a1, each verified against the downloaded zip. Replaces the 2.1.0-GA checksums in Components, the 15 SDK examples, examples/README.md and the CHANGELOG's checksum block. docs/api-surface.md from the final binaries: surface unchanged. Co-Authored-By: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
2.1.0-GA merge back to main