The AI Toolchain — issue -240, December 20, 2025

KoboldCpp v1.104 adds --smartcache dynamic KV snapshotting, --autofit GPU layer fitting, and pipeline-parallel opt-in via --pipelineparallel.

$ git clone --branch v1.104 https://github.com/LostRuins/koboldcpp.git
# already have the repo? check out this version:
$ git checkout v1.104

$ koboldcpp --model mymodel.gguf --smartcache --host 0.0.0.0 --port 5001

$ koboldcpp --model mymodel.gguf --autofit

• ›Adds --smartcache flag: a dynamic KV state snapshot system that automatically saves and reuses KV states, accelerating context-swap scenarios (e.g. AI Horde, shared instances) and supporting RNN/Hybrid models such as Qwen3Next and RWKV; max KV states increased to 6.
• ›Adds --autofit flag: enables upstream llama.cpp automatic GPU layer fitting (-fit), overwriting manual layer configs and tensor overrides to find a better GPU fit automatically.
• ›Adds --pipelineparallel flag to opt into pipeline parallelism on multi-GPU setups (previously the default); trades higher memory usage for faster speed.
• ›Adds remove_limits payload/--sdgendefaults override to bypass the new cfg_scale cap of 3.0 for Z-Image.
• ›Adds support for GLM4.6V and GLM4.6V Flash multimodal models.

llama.cpp server gains autoload-on-startup and preset-only option support

$ git clone --branch b7487 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b7487

• ›Adds load-on-startup support to llama-server, allowing a model to be loaded automatically when the server starts rather than requiring a separate API call.
• ›Adds preset-only options to llama-server, enabling configuration presets to be applied without requiring a full model specification.

The open source coding agent.

OpenCode v1.0.180 adds API endpoints to delete and update message parts.

$ git clone --branch v1.0.180 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v1.0.180

• ›Adds endpoints to delete and update individual message parts via the API.

AI agent toolkit: unified LLM API, agent loop, TUI, coding agent CLI

Pi v0.25.0 adds interruptible tool execution and two free Google OAuth providers for Gemini and Claude models.

$ git clone --branch v0.25.0 https://github.com/earendil-works/pi.git
# already have the repo? check out this version:
$ git checkout v0.25.0

$ # Inside pi, run:
/login
# Then select: Antigravity

• ›Enables mid-task agent interruption: queuing a message while tools are running skips remaining tools and processes your new message immediately.
• ›New Google Gemini CLI OAuth provider: access Gemini 2.0/2.5 models for free via Google Cloud Code Assist using /login → 'Google Gemini CLI'.
• ›New Google Antigravity OAuth provider: access Gemini 3, Claude sonnet/opus thinking models, and GPT-OSS models for free via /login → 'Antigravity'.
• ›The /model command now filters to only the models specified by --models, preventing accidental selection of out-of-scope providers.

PydanticAI v1.37.0 adds runtime model switching and DynamicToolset for TemporalAgent, plus Vertex AI image output controls.

$ git clone --branch v1.37.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v1.37.0

ImageGenerationTool(output_format='jpeg', output_compression=80)

• ›Adds output_compression and output_format parameters to ImageGenerationTool for Vertex AI Gemini image models.
• ›Enables TemporalAgent to switch model at agent.run-time, allowing per-run model selection.
• ›Adds DynamicToolset support in Temporal, enabling runtime-defined tool sets for Temporal workflows.
• ›Adds a model profile flag for APIs that support native output but still require JSON schema in instructions.
• ›Updates known Groq model names to add production/preview variants and remove deprecated entries.