Discovered through heavy dogfooding. The report + budget surface token counts and a per-model breakdown, but not a money cost. For heterogeneous runs (deepseek/qwen/kimi/claude all at different prices) the token total doesn't convey relative spend — e.g. 42k claude tokens cost far more than 42k deepseek tokens.
Proposed: an optional cost estimate driven by a price table (pandacode already ships PANDACODE_BAMBOO_PRICE_FILE); show ~$X next to total tokens and in the per-model breakdown when a price table is available. Opt-in, no behavior change when absent.
Severity: usability / nice-to-have.
Discovered through heavy dogfooding. The report + budget surface token counts and a per-model breakdown, but not a money cost. For heterogeneous runs (deepseek/qwen/kimi/claude all at different prices) the token total doesn't convey relative spend — e.g. 42k claude tokens cost far more than 42k deepseek tokens.
Proposed: an optional cost estimate driven by a price table (pandacode already ships
PANDACODE_BAMBOO_PRICE_FILE); show~$Xnext to total tokens and in the per-model breakdown when a price table is available. Opt-in, no behavior change when absent.Severity: usability / nice-to-have.