9 releases 9 tags

RSS feed

• v0.2.3 60c19fb34d

  Compare

  v0.2.3 Stable

  olemd released this 2026-10-19 15:56:38 +02:00 | 9 commits to main since this release

  • Own LLM: claudia run NAME --llm URL --llm-model MODEL runs Claude Code against any server with an Anthropic Messages API (Ollama, llama-server, vLLM, LiteLLM) instead of Anthropic. Remembered per instance; --no-llm switches back.
  • The token comes from $CLAUDIA_LLM_TOKEN (or --llm-token-env VAR) on the host and never lands in the config, the command line or --dry-run. No claude.ai login is needed, also for -p.
  • localhost URLs reach the host. In full network mode the container then gets access to the host's loopback (every service on 127.0.0.1, not only the LLM).
  • --restricted allows the LLM host automatically, including ports other than 443.
  • --llm-context TOKENS tells Claude Code the model's context window. The server must give the model that window too: Ollama's default of 4k silently cuts Claude Code's ~15k-token prompt.
  • --llm-only sends nothing to Anthropic (no telemetry, update checks or auto-updates). With --restricted only the LLM and --allow-domain hosts are allowed.
  • claudia list shows an LLM column.

  Downloads

  • Source code (ZIP)
    0 downloads
  • Source code (TAR.GZ)
    0 downloads

  ---

  • claudia-0.2.3-py3-none-any.whl
    1 download · 2026-10-19 15:59:04 +02:00 · 26 KiB
  • claudia-0.2.3.tar.gz
    0 downloads · 2026-10-19 15:59:05 +02:00 · 21 KiB
  • claudia_0.2.3_all.deb
    0 downloads · 2026-10-19 15:59:05 +02:00 · 23 KiB