r/o

don't stream responses until we're done 3d5dd6cb parent ec40d27c

The LLM generates half-formed opinions with tool calls, then a final response overwriting the earlier one. Let\'s not.

authored by ~lotte

1
export PIPA_API_KEY=pipapeep
2
export GITLAB_TOKEN=glpat-peeeeeep
3
export GITHUB_TOKEN=ghp_dooooooWawawaaooo
4
export PIPA_LLM_MODEL=qwen3-coder:30b
5
export PIPA_LLM_BASE_URL=http://localhost:11434/v1
6
export PIPA_LLM_ENABLED=true
7