Preface¶
When using DSH Web, a single conversation may involve multiple models, cache hits, non-cache hits for inputs, output tokens, and different peak/off-peak periods. If fees are not displayed directly in the interface, developers usually need to refer to the table prices themselves and then split them into input, output, cache hits, and peak/off-peak tiers.
Addressing this scenario, dsh-conversation-cost places DeepSeek usage fees into the statistics row at the bottom of the DSH Web conversation, displaying RMB and USD amounts in real-time. DSH’s plugin architecture also allows such display capabilities to be installed, upgraded, and rolled back independently.
Positioning¶
dsh-conversation-cost is a DSH Web plugin, with the repository at Ayaka157/dsh-conversation-cost and an MIT license.
It solves the problem of viewing the current DeepSeek usage fees within the DSH conversation interface: providing RMB + USD dual-currency fees in the bottom statistics row, and handling cache hits, peak/off-peak pricing, and mixed model pricing.
Core Features¶
Display Fees in the Bottom Statistics Row¶
The plugin displays the current conversation’s DeepSeek usage fees in real-time in the statistics row at the bottom of the DSH Web conversation.
The amounts are displayed in both RMB and USD:
- RMB is taken from the official Chinese page table prices.
- USD is taken from the official English page table prices.
- Both are independent of each other; exchange rates are not converted.
The fees are the last group in the statistics row, using the same separator, alignment, and font size as “Rounds/Steps, Time Taken, TTFT, Cache Hits, Input/Output”, and automatically adapting to Chinese and English.
Per-Request Billing¶
The plugin registers a costLedger session projection on the Host side, folding request events one by one into “Tier × Model” token buckets. It does not perform aggregated estimation or replay logs.
Model pricing rules are as follows:
- Each request is priced using the model source included in the message.
- Failed requests fall back to request headers.
- Mixed model sessions are priced according to their respective models.
- Unknown models are not priced and are labeled.
Cache and Peak/Off-peak¶
Cache hits and non-cache hits for inputs are priced separately, while cache writes are priced at ¥0. The plugin does not double-charge inputTokens.
Peak/off-peak pricing automatically categorizes each request into the following tiers based on its occurrence time:
- Standard (Flat rate)
- Peak
- Off-peak (Idle)
Before August 17, 2026, 00:00 (Beijing Time), it is all standard pricing. Peak/off-peak is determined by the request emission time step/start, which is a convention of this plugin; if the official stance is clarified later, it can be adjusted at a specific point.
Tier Details¶
When hovering over the statistics row, you can view the subtotals for the Standard, Peak, and Off-peak tiers, as well as the unknown model prompt.
Installation and Enablement¶
Install from GitHub¶
The following command installs the plugin from GitHub, followed by starting DSH Web:
npx -p @deepseek-ai/dsh dsh plugin --profile web add github:Ayaka157/dsh-conversation-cost
dsh web
The first command installs the plugin, and the second command starts the Web interface.
Install from Local Directory¶
You can also clone the repository first and then install it from the directory:
git clone git@github.com:Ayaka157/dsh-conversation-cost.git
cd dsh-conversation-cost
dsh plugin --profile web add .
This method is suitable for scenarios where you need to view local source code first and then install the plugin from the current directory.
Upgrade¶
To upgrade, re-run the GitHub installation command for a fresh install, then restart DSH Web for the changes to take effect:
npx -p @deepseek-ai/dsh dsh plugin --profile web add github:Ayaka157/dsh-conversation-cost
After restarting, DSH Web will display fees according to the status of the currently installed plugins.
Typical Usage¶
-
Install the plugin and run
dsh web. -
Start a conversation and view the current conversation fees in the bottom statistics row.
-
Hover over the statistics row to view the subtotals for Standard, Peak, and Off-peak, as well as the unknown model prompt.
-
When upgrading the plugin, re-run the installation command for a fresh install, then restart DSH Web.
Suitable Scenarios and Notes¶
Suitable for scenarios in DSH Web where you need to quickly verify conversation fees, especially for conversations involving mixed models, cache hits, and peak/off-peak periods.
Please note before use:
- Prices will change according to official table prices. Verified materials mention that the prices in the table are official table prices scraped on August 14, 2026, and may change at any time; peak/off-peak pricing takes effect on August 17, 2026, 00:00 (Beijing Time).
- When
costLedgeris not ready, for example, just installed or historical backfilling is incomplete, the plugin falls back to an approximate amount of “Aggregated Usage × Current Tier” and adds a “≈” label. - Unknown models are not priced and are reflected in relevant prompts in the statistics row.
- The plugin runs with the permissions of the current
dshprocess. You should check the source code and license before installing. - The repository license is MIT.
The DSH plugin ecosystem emphasizes “everything is a plugin”; the community directory is an independent site with no official affiliation to DeepSeek/Huansuan, and should not be treated as an official app store.
Links¶
GitHub: Ayaka157/dsh-conversation-cost
The verified materials do not provide a directory page URL, so this article does not list the directory page link.