> For the complete documentation index, see [llms.txt](https://docs.darcyiq.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.darcyiq.com/dispatch/use-dispatch/overview.md). # Overview What Overview shows about spend, savings and recent requests, and how it differs for Dispatch Users and Dispatch Admins Overview answers the question most people open Dispatch with: what is this costing right now? It shows the last 30 days of usage at a glance, with savings, spend, where spend is heading, and a live list of recent requests. You don't configure anything here. When you need a different time window or a different breakdown, Overview sends you to Cost Reporting with the question already asked. Overview answers for whoever is reading it. A **Dispatch User** sees the keys they own. A **Dispatch Admin** sees every key in the organization. {% hint style="info" %} **Watching your first request?** **Recent requests** at the bottom of the page refreshes every 15 seconds, so you can send a message from your tool and watch it arrive without reloading. {% endhint %} ## The Page Header The page greets you by first name and says what it's showing: usage over the last 30 days, across the keys you own or across every key in the organization. | Control | Who sees it | What it does | | --------------- | --------------- | ----------------------------------------------------------------------------------------------------------------------------------------------- | | **Quick start** | Everyone | Reopens Dispatch's guided setup, which can create a key, connect a tool and wait for your first request | | **Invite** | Dispatch Admins | Opens **Invite a colleague**, where you enter a first name, last name, work email and **Dispatch role**. See [Users](/dispatch/manage/users.md) | Dispatch Admins also see up to three setup suggestions under the header, such as **Set a spend ceiling** or **Invite your team**, until setup is complete. Each one reopens the guided setup at that step. You can dismiss them, and the **Setup** count in the top bar keeps the remaining steps one click away. Dispatch Users see their suggestions on the API keys page instead. ## Headline Figures Four cards across the top summarize the 30-day window. | Card | What it shows | | --------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | **Total savings** | What Dispatch saved, split into **Routing** (how much less your tiers cost than the model each tier stands in for) and **Prompt Optimization** (tokens removed before requests reached a model). Shows a dash when no comparison could be priced for the models in the window. | | **Total cost** | What the window cost, with the change against the 30 days before it | | **Organization budget** or **Run rate** | See below | | **Effective token rate** | What you paid per million tokens, shown for input and output separately, or blended into one figure when the split isn't available | The third card depends on your role and on whether a budget is set: * **Dispatch Admins with an organization budget** see **Organization budget**, with the budget's window in brackets. It shows the share used, a bar that turns amber and then red as the budget fills, and what you're on track to spend over the budget's window at the current rate. * **Dispatch Admins without a budget** see **Run rate**: spend per day, and roughly what that comes to over 30 days, noted as "no limit set". * **Dispatch Users** always see **Run rate** for their own keys. The organization's budget isn't shown to them. The card is hidden until there is spend in the window. ## Usage Counters A row of smaller tiles follows. Which ones appear depends on your traffic. * **Request and token counters**, such as **Requests**, **Input tokens** and **Output tokens** * **Cache** — the **Hit rate** (the share of input tokens served from cache) and the **Saving** that caching produced * **Speed** — median and tail (p50, p90, p99) figures for **Throughput** in tokens per second, **First token** on streamed answers, and total **Latency** * **Escalated** and **Auto adoption** — for organizations using Auto, how often Auto sent a turn to its stronger tier, and what share of the requests counted on the tile asked for Auto. When your organization has more than one Auto, a picker on the tile switches between them. ## Where Spend Is Heading **The next 30 days** projects spend at your current rate, as a sentence ("At this rate, the next 30 days come to about...") and as a chart. The chart shows recorded days as bars and the projection as a dashed line with a shaded range. A note under the sentence says how far to trust it, from "A rough guide only" to "Based on a steady few weeks". Use the **7d**, **14d** and **30d** buttons to choose how much history the projection is based on. The 30-day default is the steadiest. A shorter window follows a recent change faster, which helps when your team only started using Dispatch recently. This panel doesn't appear until there's spend to project from. ## What Dispatch Saved **What Dispatch saved** shows each day as a stacked bar: what you spent, then what routing saved, then what Prompt Optimization saved. The top of each bar is what the day would have cost without Dispatch. Switch between **7d** and **30d** to change the range. If Prompt Optimization isn't saving anything yet, Dispatch Admins see a note pointing to where it's turned on. See [Guardrails](/dispatch/manage/manage/guardrails.md). ## Break This Down Once there's spend in the window, **Break this down** offers shortcuts that open Cost Reporting with a report already set up. | Shortcut | Answers | Who sees it | | ----------------------- | ------------------------------------ | --------------- | | **Spend over time** | Daily, weekly or hourly spend | Everyone | | **Spend by person** | Who is spending, most first | Dispatch Admins | | **Spend by tier** | Which tiers carry the bill | Everyone | | **Tier mix over time** | Whether the mix of tiers is drifting | Everyone | | **Savings by tier** | Where the saving comes from | Everyone | | **Spend by tag** | Spend by business unit or project | Everyone | | **Who uses which tier** | People against tiers | Dispatch Admins | See [Cost Reporting](/dispatch/cost-and-insights/cost-reporting.md). ## Breakdowns Below the shortcuts, Overview splits the window's spend a few ways. Each section only appears when it has something to show. * **Input versus output** — how the bill divides between the two directions, and the rate paid for each * **Spend by tier** — spend and share per tier, shown when at least two tiers have spend * **Spend by tag** (Dispatch Users see **Spend by tag on your keys**) — spend per tag, shown once at least one tag has spend. A key can carry several tags, so its spend counts under each of them and the percentages can add up to more than 100%. Rows for keys with no tag, and for spend recorded before tags were, appear in italics, and hovering explains which is which. See [Tags](/dispatch/manage/manage/tags.md). * **Auto** — appears only when Auto is available but nothing in the window asked for it ## Habits **Your habits** (or **Habits across the organization** for Dispatch Admins) describes how Dispatch is being used rather than how much it cost. It shows a bar of spend by tier, plus four figures: | Figure | Meaning | | ------------------------ | --------------------------------------------------------- | | **Served from cache** | The share of the prompt billed at the cheaper cached rate | | **Sessions** | Conversations, however many turns each | | **Cost per session** | What one conversation costs, on average | | **Requests per session** | Turns an agent takes to finish | Nobody is compared with anybody here. A Dispatch User sees only their own keys. ## Recent Requests **Recent requests** lists individual requests, newest first, and updates every 15 seconds while the tab is visible. It appears even when there's no spend yet, because a key's first request is when people most want to watch it. Click **Show older requests** to page back. | Column | What it shows | | ------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **When** | How long ago the request finished, plus a status badge when it didn't simply complete | | **Key** (Dispatch Admins see **Key and person**) | The key's label and, where Dispatch recognizes it, the tool that sent the request. Dispatch Admins also see the key's owner. | | **Tier** | The tier that served the request. For a request that asked for Auto, an arrow shows which tier Auto chose. A second line shows the settings the request was sent with, such as effort or temperature, in italics where the tier filled in a setting the request left out. | | **Tokens in / out** | Input and output tokens, plus the share served from cache and any thinking tokens | | **Cost** | What the request cost | | **Took** | Total time, plus time to first token and tokens per second where measured | The status badges you may see are: | Badge | Meaning | | -------------------- | ---------------------------------------------------------------------------------- | | **Cancelled** | A streamed request was cancelled before it finished | | **Over budget** | A spend limit had been reached, so the request was refused | | **Blocked** | A guardrail's content policy refused the request | | **Unknown model** | The request named something that isn't one of your tiers | | **Invalid request** | Dispatch couldn't read the request | | **Not signed in** | The request wasn't authenticated | | **Hit output limit** | The request completed, but the answer stopped at the output limit, so it's partial | | **Failed** | Any other failure | Every response from Dispatch also carries this information in its `x-darcy-*` headers, for tools that don't show it themselves. ## The Budget Meter (Dispatch Admins) Dispatch Admins also see a meter at the foot of the sidebar on every page. It shows the organization's prepaid credit balance, labelled **Left for the organization**. When an organization budget is set, it also shows what's left in the budget's daily, weekly or monthly window, with a bar that turns amber at 75% used and red at 90%. **Add credits** opens Billing & Budget. See [Billing & Budget](/dispatch/manage/manage/billing-and-budget.md). ## How Overview Differs by Role | | Dispatch User | Dispatch Admin | | --------------------------------------------------------- | ----------------- | -------------------------------------------------------------- | | Whose usage | The keys you own | Every key in the organization | | Budget card | **Run rate** only | **Organization budget**, or **Run rate** when no budget is set | | **Invite** button and setup suggestions | — | ✓ | | **Spend by person** and **Who uses which tier** shortcuts | — | ✓ | | Key owner shown in **Recent requests** | — | ✓ | | Prompt Optimization note | — | ✓ | | Sidebar budget meter | — | ✓ | ## Next Steps | Goal | Documentation | | ----------------------------------------------- | ----------------------------------------------------------------- | | Build your own breakdown, over any window | [Cost Reporting](/dispatch/cost-and-insights/cost-reporting.md) | | Find unusual traffic, idle keys and savings | [Insights](/dispatch/cost-and-insights/insights.md) | | Set the organization's budget or add credits | [Billing & Budget](/dispatch/manage/manage/billing-and-budget.md) | | Understand the tiers your spend is split across | [Models](/dispatch/use-dispatch/models.md) | | See spend per key | [API Keys](/dispatch/use-dispatch/api-keys.md) | --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.darcyiq.com/dispatch/use-dispatch/overview.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.