Submission method

Produce a new result cell

Run one model at a time. Keep the result file and the evidence directory written beside it.

Ollama

MODEL=qwen2.5:7b-instruct
OUT=results/ollama-qwen2.5-7b-instruct.json
cargo run -p willitcall -- run \
  --model "$MODEL" \
  --server ollama \
  --out "$OUT"
cargo run -p willitcall -- validate "$OUT"

llama.cpp

MODEL_PATH=/absolute/path/to/model.Q4_K_M.gguf
OUT=results/llamacpp-model-q4_k_m.json
cargo run -p willitcall -- run \
  --model "$MODEL_PATH" \
  --server llamacpp \
  --out "$OUT"
cargo run -p willitcall -- validate "$OUT"

Pull request checklist

preflight clean (no contention override), or the override is explained
result file schema-valid
evidence transcripts included
empty responses cross-checked on a second server per the seeding protocol

Read CONTRIBUTING.md for the complete contribution rules.