Repository navigation

feat(local): replace Ollama with a bundled llama.cpp runtime, a curated GGUF catalog, and a hardware-aware fit estimator - #1819

Merged

AnishSarkar22 merged 61 commits into

MODSetter:devfrom

AnishSarkar22:feat/llama-cpp

Sep 22, 2026

Merged

feat(local): replace Ollama with a bundled llama.cpp runtime, a curated GGUF catalog, and a hardware-aware fit estimator#1819
AnishSarkar22 merged 61 commits into
MODSetter:devfrom
AnishSarkar22:feat/llama-cpp

Commits on Sep 19, 2026
Commits on Sep 20, 2026
- build(electron): stage a pinned llama.cpp build in place of Ollama and llmfit
  
  AnishSarkar22
  and
  claude
  committed
  750ba30
- ci(release-local): pin ubuntu-22.04 and smoke llama-server instead of Ollama
  
  AnishSarkar22
  and
  claude
  committed
  4fafae3
- feat(electron): run llama-server in router mode as a sidecar
  
  AnishSarkar22
  and
  claude
  committed
  ca26d10
- feat(config): point the backend at llama-server and drop the llmfit settings
  
  AnishSarkar22
  and
  claude
  committed
  64dafcb
- feat(llm/gguf): read GGUF headers and derive model shape from local or remote files
  
  AnishSarkar22
  and
  claude
  committed
  efb89c8
- feat(llm/hardware): probe devices, libraries, and system memory through ggml
  
  AnishSarkar22
  and
  claude
  committed
  8ca33e8
- feat(llm/fit): estimate KV cache, compute buffers, and load plans per model
  
  AnishSarkar22
  and
  claude
  committed
  a1170b2
- feat(llm/providers): accept a transport and json_schema in the OpenAI-compatible provider
  
  AnishSarkar22
  and
  claude
  committed
  d25b3aa
- feat(llm/providers): add the llamacpp provider with presets, downloads, and thinking control
  
  AnishSarkar22
  and
  claude
  committed
  0a86d4d
- feat(llm/catalog): replace recommendations with a curated manifest and Hugging Face search
  
  AnishSarkar22
  and
  claude
  committed
  cdab93b
- refactor(llm): remove the Ollama provider and the llmfit recommendation module
  
  AnishSarkar22
  and
  claude
  committed
  9ffe32d
- feat(llm): resolve and select generation models through the llamacpp provider
  
  AnishSarkar22
  and
  claude
  committed
  dea61e4
- feat(egress): split the ollama_pull grant into model_download and model_search
  
  AnishSarkar22
  and
  claude
  committed
  c0f9cd2
- feat(db): migrate selected models and egress grants off Ollama in revision 0012
  
  AnishSarkar22
  and
  claude
  committed
  2004f02
- fix(secrets): raise UnreadableSecretError instead of leaking InvalidToken
  
  AnishSarkar22
  and
  claude
  committed
  f439630
- feat(chat): classify an unreadable stored key as a provider auth error
  
  AnishSarkar22
  and
  claude
  committed
  a5da65c
- feat(frontend): show fit badges and model search in the model catalog
  
  AnishSarkar22
  and
  claude
  committed
  12abfb0
- refactor(frontend): rename Ollama-era copy and tests for the llamacpp runtime
  
  AnishSarkar22
  and
  claude
  committed
  768f4f1
- chore(scripts): add the curated model refresh script backed by Hugging Face
  
  AnishSarkar22
  and
  claude
  committed
  156437d
- docs(local): update the README and phase 7 spec for the llama.cpp runtime
  
  AnishSarkar22
  and
  claude
  committed
  2d91219
- refactor(llm): remove the Ollama provider and the llmfit recommendation module
  AnishSarkar22
  committed
  accad5b
- Merge remote-tracking branch 'upstream/dev' into feat/llama-cpp
  AnishSarkar22
  committed
  d833cc3
Commits on Sep 21, 2026
Commits on Sep 22, 2026

Commits

Commits on Sep 19, 2026

Commits on Sep 20, 2026

Commits on Sep 21, 2026

Commits on Sep 22, 2026