AIsa Console

A tour of the AIsa Console: the dashboard for balance and 30-day trends, API keys, the model playground, and per-request usage logs - plus the usage API for reading the same totals from code.

The AIsa Console is where you manage keys and balance, try models in the browser, and audit every request.

Page	Use it to
Dashboard	See balance, 30-day spend and requests, error rate, and model split
API Keys	Create, limit, disable, and revoke keys - see Authentication
Billing	Top up your balance - see Billing & Credits
Playground	Send a real request to any model and copy it as cURL
Usage & Logs	Inspect each request: endpoint, key, model, latency, tokens, cost, status

Dashboard

The Dashboard is the Console’s landing page. It summarizes spend, traffic, and account health for the last 30 days.

New API Key and Top Up sit next to the page title. If a first top-up promotion is active, the Top Up button shows it.

• Total Balance - credit available for API spend, drawn down as requests are processed.
• Spend · 30D - what your traffic cost over the last 30 days, with the daily average.
• Requests · 30D - calls in the same window, with the error rate. A rising error rate is usually the first sign of a bad key, an unavailable model, or a malformed request.
• Active Keys - enabled keys out of total, and when any key was last used.

The Spend and Requests charts break the same totals down by day; hover or use the arrow keys to read a single day. Requests flat while spend rises usually means traffic moved to more expensive models or longer contexts. Model Split shows which models your calls went to, and the Account panel shows your plan, verification status, and last sign-in.

Usage figures cover a fixed 30-day window; balance and keys are current state. For other ranges or per-request detail, use Usage & Logs.

Playground

The Playground sends real requests through the same endpoints your code uses and shows the exact request body and raw response. You need at least one enabled key.

1. Key - choose which of your enabled keys authenticates the request. The full key never reaches the browser, and usage is attributed to that key, so a dedicated key keeps Playground traffic separate.
2. Route - Chat completions (POST /v1/chat/completions, OpenAI-compatible) or Messages (POST /v1/messages, Anthropic-compatible). The model list reloads, because the routes serve different models.
3. Model - every model the key can reach on that route. If you are on trial credit and a model needs a paid balance, Run Test is disabled for it.
4. Parameters - Stream, Temperature (opt-in; off means the model default), Max tokens, and the System and User prompts.
5. Run Test - becomes Stop while the request is in flight.

The debug panel shows the request as JSON and as a cURL command using $AISA_API_KEY - the fastest way to move a working configuration into a script. If a prompt contains a single quote ('), escape it before running the cURL in a shell.

The output panel shows the extracted Result and the raw Response, with latency and the x-request-id. Use that ID to find the call in Usage & Logs or when contacting support. Playground requests are billed exactly like API requests.

Usage and Logs

Usage & Logs is the per-request record of the last 30 days of traffic through your keys. Every figure on the page follows the filters you apply.

The summary cards show Requests, Tokens (input plus output), Cache Tokens (read plus write), and Usage Value - the traffic’s worth at list rates - with the amount actually charged below it. The two differ when a discount, promotion, or refund applied.

Search matches request ID, key, endpoint, and model, so pasting a request ID from an error or the Playground jumps to that call. Filter by Stream (streaming or not) and Status (success, failed, refunded).

Column	What it shows
Time	When the request arrived, with its request ID
Endpoint	The route that served it, such as /v1/chat/completions
API Key	The key’s name, or its prefix if unnamed
Model	The model that processed the request
Stream	stream or sync
Latency	Total duration, with time to first token
Tokens	Total, with the input / output split
Cache	Total cache tokens, with the read / write split
Usage Value	Value at list rates; the charged amount appears below if different
Status	success, failed, or refunded

Typical uses: filter by Failed to debug a call, find which model and key drove a cost spike, confirm prompt caching works by comparing cache reads with total tokens, and audit everything one key has done.

Query usage with an API key

GET https://api.aisa.one/v1/usage returns aggregate usage for the authenticated key by default. Use the same key that made the calls. It is not a per-request log; use the Console above to inspect individual calls.

# Positive Unix seconds in UTC; this example selects the last 24 hours.
start_time=$(python3 -c 'import time; print(int(time.time()) - 86400)')
curl --fail-with-body --silent --show-error --get \
  "https://api.aisa.one/v1/usage" \
  --header "Authorization: Bearer ${AISA_API_KEY:?Set AISA_API_KEY}" \
  --data-urlencode "start_time=$start_time" \
  --data-urlencode "bucket_width=1d" \
  --data-urlencode "scope=key"

Parameter	Meaning
start_time	Required positive Unix timestamp in seconds, not milliseconds or a date string.
end_time	Optional Unix seconds; defaults to the current time. Must be later than start_time. The interval includes start and excludes end: [start_time, end_time).
Window	At most 31 days (31 × 24 hours); split longer periods into adjacent windows.
bucket_width	Only 1d is supported, with UTC daily buckets.
scope	Defaults to key. account requires OAuth authentication; an ordinary API key cannot request account scope.

totals.charged_micros_usd is the actual charged amount; divide integer micro-USD by 1,000,000 for USD. totals.usage_value_micros_usd is usage value and can differ from the charge. Compare the same key, time window and currency when reconciling. This aggregate does not prove the outcome of a specific request or that a historical top-up succeeded.

When data fails to load

If balance, key, or usage data cannot be fetched, an inline warning appears at the top of the Dashboard with a Refresh button that retries all of them.