RepoJournal
Local LLMs Local LLMs
59 wires and counting

$ follow Local LLMs

Keep up with Local LLMs in about 3 minutes: what actually shipped — the commits, pull requests, releases, and security advisories that matter.

or

fair warning: these emails are deeply technical. diffs, version numbers, CVEs, benchmark deltas. if that's not your idea of a good read, this isn't your newsletter.

Folds into your digest — weekly by default, monthly if you prefer. Unsubscribe in one click.

$ status

wire 2026-07-23
stories 212
RSS Archive Reviews Topics

© 2026 RepoJournal Home Showcase Explore How it works Privacy

$ the-wire · showcase

OLLAMA AGENTS NOW IMPORT SKILLS FROM CODEX AND CLAUDE

By RepoJournal · Filed · About Local LLMs · Composed from the cited sources · methodology

dhiltgen ch-wan ParthSareen m1el Pascal sagearc connorcarpenter15
14 people shipped this
dhiltgen @dhiltgen 2 cited ch-wan @ch-wan 2 cited ParthSareen @ParthSareen 1 cited m1el @m1el 1 cited Pascal @ServeurpersoCom 1 cited sagearc @sagearc 1 cited connorcarpenter15 @connorcarpenter15 1 cited ColinZ22 @ColinZ22 1 cited rankaiyx @rankaiyx Niklas Wenzel @nikwen Oliver Simons @ORippler noooop @noooop njhill @njhill ispobock @ispobock

Ollama shipped a skills import system that lets coding agents pull validated skill directories from external sources, while simultaneous fixes to GLM tool-call parsing and Laguna model alignment prevent agents from silently failing.

The agent-TUI now supports `/skills import codex|claude|pi` with nested slash-command completion, validating and reloading skill catalogs without overwriting existing entries [1]. This lands as GLM's tool-call parser gets a critical fix: the parser was buffering incomplete calls and returning empty responses on stream end, leaving agents unable to continue when models partially emitted closing tags [2]. Ollama also aligned Laguna with upstream llama.cpp, removing local implementation and translating older GGUF metadata names to keep existing models loading [3]. Across the llama.cpp ecosystem, WebGPU gained depthwise convolution support ported from Vulkan [4], while the UI now lets users defer reasoning decisions to the server instead of forcing a setting that breaks unconditional-reasoning models like Gemma 4 [5]. vLLM's Rust frontend extracted request preparation into engine-free processors [6] and added a gRPC abort control RPC for canceling requests by ID [7], while ROCm support now handles AMD Quark DeepSeek-V4-MXFP4 checkpoints with fused shared expert support [8]. SGLang's config refactor introduced a structured RuntimeContext API with per-domain namespace bags [9], though a partial revert removed migrated config-namespace reads after correctness defects surfaced [10].

Action items

References

  1. [1] agent: import skills from coding agents ↗ ollama/ollama
  2. [2] model/parsers: finalize incomplete GLM tool calls ↗ ollama/ollama
  3. [3] model: align Laguna with upstream llama.cpp ↗ ollama/ollama
  4. [4] webgpu : add CONV_2D_DW (depthwise conv2d) kernel (#25847) ↗ ggml-org/llama.cpp
  5. [5] ui: Add a "Default" option for the reasoning selector (#25846) ↗ ggml-org/llama.cpp
  6. [6] [Rust Frontend] Extract request preparation from the inference path ↗ vllm-project/vllm
  7. [7] [Rust Frontend][gRPC] Add abort control RPC ↗ vllm-project/vllm
  8. [8] [ROCm] Fused Shared Expert Support for AMD Quark DeepSeek-V4 Model Checkpoints ↗ vllm-project/vllm
  9. [9] runtime_context: add resolved-config namespace bags and accessors ↗ sgl-project/sglang
  10. [10] Revert RuntimeContext config-namespace reads/roles (#31813–#31817) ↗ sgl-project/sglang

$ ls local-llm/reviews/ # the step-back reads

the week in review Sparse attention lands across vLLM, SGLang, llama.cpp the month in review Speculative decoding consolidates across the stack as DeepSeek V4 lands
Topics: Local LLMs AI / ML Full archive → Weekly reviews →

Quick answers

What shipped in Local LLMs on July 23, 2026?
Ollama shipped a skills import system that lets coding agents pull validated skill directories from external sources, while simultaneous fixes to GLM tool-call parsing and Laguna model alignment prevent agents from silently failing. In total, 101 commits, 101 pull requests, and 10 releases landed.
Who contributed to Local LLMs on July 23, 2026?
14 developers shipped this update, including ParthSareen, dhiltgen, m1el, rankaiyx, Niklas Wenzel, Oliver Simons, Pascal, and sagearc, and 6 more.
What were the notable Local LLMs updates?
agent: import skills from coding agents, model/parsers: finalize incomplete GLM tool calls, and model: align Laguna with upstream llama.cpp.
♥

$ log Local LLMs --recent

Metal MoE/SSM fusion lands, MiMo-V2.5 fp8 sharding fixed 158 shipped Ollama exposes thinking levels, llama.cpp widens Hexagon and architecture coverage 250 shipped GGUF offset corruption fixed, vLLM restores Dynamo KV metadata 244 shipped MLX engine graduates out of x/, ROCm fixes silent attention corruption 268 shipped

$ log --all # elsewhere on the wire

@openclaw Cold-start Gateway diagnostics stop crying wolf, doctor repairs plugin drift Agentic Coding Cline desktop bundles shrink, goose unblocks its own clicks

$ ls showcases/

Rails Django Laravel Node.js Go Vercel Vue.js Kubernetes HashiCorp Arch Linux Linux Anthropic

$ init your-desk

RepoJournal can write one of these for any repo you care about. Sign in with GitHub, pick your repos and sources, and your desk gets a wire like this one every morning.