@javenlu233/dsh-client-ui-turn-cost

@javenlu233/dsh-client-ui-turn-cost

English | 中文

Cost readout for the Web chat view: the session total (with its cache-hit / cache-miss / output split and cache-hit rate) in the composer stats strip, plus a compact per-turn increment in each assistant message's action row.

• Data: the session total rides the durable sessionCost projection (whole log, paging- and compaction-proof, priced per request by its model, including captured web_search model usage); the per-turn increment folds that turn's assistant nodes client-side.
• Pricing: the DeepSeek price table (CNY per 1M tokens), date-aware: the default route is now deepseek-flash; V4.1-Flash rates from 2026-09-10 12:00 Beijing time are hit / miss / output 0.04 / 2 / 8 at peak and 0.02 / 1 / 4 off-peak, while historical usage keeps its historical rate.
• Slots: two list slots — conversation.composer.dock (session total) and conversation.chat.assistant-actions (per-turn); composing this plugin out removes both.
• Dependency: requires the host @javenlu233/dsh-session-cost projection plugin; without it the session total is absent and only the per-turn increment renders.

Model Experience

None; nothing here reaches a model request.

KV Cache effect

None; this package neither assembles nor sends a provider request.

Known Limitations and Deferred Work

• The per-turn increment is precise only for the loaded window; compacted turns contribute to the session total but have no per-turn line.
• The price table is configurable through the host session-cost plugin's cordis.yml config, not the settings UI.
• The stats dialog lists web_search model spend in its own cut when any search usage was captured; conversation and search tokens for the same model still share the "By model" row.