Guide updated
Implemented using the official billing API and public Billing console logic; live-account verification is pending. Key creation, billing access, and post-payment reconciliation have not been tested in a signed-in account. Automated tests are not live verification.
- Required credentials
- DeepInfra API Key
- Monitoring currency
- USD · US dollars
- What is monitored
- Account prepaid Credits
- Shared DeepInfra account Prepaid Credits, following the official console formula −(stripe_balance + recent). stripe_balance is the signed account ledger; recent is usage since the last invoice. Both are dollars, not cents. This is not a token-price estimate or a separate API-key or Quota Monitor project wallet.
How to check your DeepInfra balance manually
- Sign in to the DeepInfra account that pays for your API calls and open Billing.
- Find Prepaid Credits and read the USD amount; a negative amount represents money owed.
- Keep model-specific credits, monthly spending, and key limits separate. Use Add funds if you need to refill the general wallet.
Connect DeepInfra to Quota Monitor
1. Confirm the DeepInfra account and wallet
Sign in to DeepInfra and open Billing directly. Locate Prepaid Credits. Confirm this is the account your application actually uses, not another provider's credits or your Vercel wallet.
Check: you can identify the target account's USD Prepaid Credits. No payment is required to connect monitoring. If you only want a manual balance check, you can stop here.
2. Open API Keys, not SSH Keys
Select Keys in the sidebar, open the API Keys tab, and choose New API key. This opens Generate API key. The direct link below also opens Keys; no cloud IAM policy is needed.
Check: this is the same DeepInfra account's API-key list, not SSH Keys, another platform's keys, or a model name. Menu names come from official public page text; the signed-in flow still awaits live verification.
3. Generate and save a dedicated key
Enter Quota Monitor in API Key name and use the dialog's generate button. Copy the complete generated secret to secure storage, not its Key ID, display suffix, or a sample request. You do not need Also export this key to Vercel.
Check: you have saved the full secret. The official dialog asks you to copy it immediately. If lost, create another dedicated key; do not delete keys used by running applications. Never put keys in chat, tickets, or screenshots.
4. Understand permissions and existing restrictions
An ordinary API key is not a balance-only read token and can make chargeable model calls. A separate name helps auditing and revocation but does not narrow permissions. If your team uses key limits or IP restrictions, ask an administrator to confirm billing access for the monitoring server; your laptop's IP is not its server egress IP.
Check: you accept the key's actual privileges without weakening account security for this test. A key limit is not account funds. Billing reads with a zero key limit have not been tested, so do not treat zero as a verified read-only solution.
5. Save the connection inside a Quota Monitor project
Open the target project → add provider connection → add new connection. Choose DeepInfra (USD Credits), enter a recognizable connection name, and paste the full secret from step 3 into the DeepInfra API Key password field. Save and connect to the project.
Check: no Bearer prefix, API URL, Account ID, provider project ID, or Key ID is needed. Monitoring uses USD. The key is encrypted and only its last four characters are displayed. For MCP setup, use the secure entry link instead of giving the key to an agent.
6. Sync and reconcile Prepaid Credits
Wait for first sync or choose Query latest balance. Confirm a new successful query time, then compare with Prepaid Credits in the same account's Billing page. Do not compare monthly spend or model-scoped promotional credits.
Check: account, amount, sign, currency, and time match. Live verification remains pending. Report redacted amounts, error codes, and query times if they differ, never your key. A retained snapshot after a failed sync is not a fresh balance.
7. Inspect the breakdown when needed
Open the information button beside the amount to see the signed ledger and already-deducted uninvoiced usage. For example, stripe_balance = -20 and recent = 3.25 give 16.75 USD in Prepaid Credits, not 20 USD. Do not divide these fields by 100.
Check: recent is not subtracted twice or mislabeled as this month's costs. A negative result is money owed under this billing calculation. Model-restricted Credits have separate eligibility and are not added to this general wallet.
8. Top up directly with DeepInfra
Choose Open provider on the card. In the same account's Billing → Prepaid Credits, select Add funds, review DeepInfra's current payment terms, and pay yourself. Managed billing accounts without this action should follow the provider's billing instructions.
Check: payment has posted before syncing again. Quota Monitor never calls the funding API, charges your payment method, or changes auto-top-up, limits, or payment settings.
9. Replace or revoke the dedicated key safely
Create a new dedicated key, add a replacement Quota Monitor connection, and verify sync. Remove every project reference to the old connection, delete the unused connection, then delete its dedicated key in DeepInfra Keys. Plan a replacement window first if your plan has no free connection slot.
Check: no production key used by another application was revoked. Removing a connection from one project does not revoke upstream credentials. Revoke an unused dedicated key on the provider when monitoring ends.
What Quota Monitor reads from DeepInfra
- The public GET /payment/checklist documents the signed ledger and uninvoiced usage. The combined formula and USD units follow the official Billing page implementation, pending live reconciliation. Missing either amount fails sync; we never substitute zero, use stripe_balance alone, or deduct monthly usage again.
- Model-restricted scoped_credits are excluded, as are account spend caps, per-key limits, and other platforms' wallets. Projects using one account share funds; do not sum duplicate balances.
- Negative results mean money owed and are preserved, not made absolute or clamped to zero. An explicit suspended=true marks the account unavailable. Other responses do not guarantee successful model calls or override other limits.
- This is the latest successful billing snapshot. Usage and payments may take time to post; refreshing cannot accelerate settlement. This integration does not report monthly costs: recent is not a calendar-month total. We do not change auto-top-up or payment methods.
- The dashboard shows the latest successful snapshot. Manual refresh queries the provider again but cannot eliminate upstream billing delays. An old balance is not a new result after a failed query.
DeepInfra FAQ and troubleshooting
- Why do billing reads return 401 or 403 when model calls work?
- Check the account, complete non-revoked key, lack of a Bearer prefix, and that you did not paste a Key ID. Review IP restrictions and billing access. Working model calls do not prove billing access; contact DeepInfra if authorization is unavailable. Do not switch to Cookies, more privileged tokens, or disabled security policies.
- Why isn't the balance the absolute value of stripe_balance?
- Official Billing includes uninvoiced usage recent and displays −(stripe_balance + recent) in USD. Negating only the ledger omits that usage. These two fields are dollars; do not apply the cents conversion used by the funding API or scoped_credits.
- Why does it differ from promotional credits, limits, or another project?
- Only general account Prepaid Credits are monitored, not model-restricted scoped_credits, per-key limits, or monthly budgets. Multiple projects using one account share the same money; never add duplicate wallet balances.
- What if a top-up is missing, credits are negative, or sync fails?
- Check the account and payment status and wait for posting before refreshing. Negative amounts remain visible as money owed. Retry later after 429; missing fields and service errors do not become zero. Inspect the last successful time and error code without disclosing your key.
- Can I use a balance-only key or automatic top-up?
- No compatible balance-only read scope has been confirmed; ordinary keys can spend funds. Quota Monitor does not use funding or automatic top-up APIs. Use official Billing manually if you cannot accept key custody.
Official DeepInfra references
- Official reference · docs.deepinfra.com/api-reference/billing/get-checklist(opens in a new tab)
- Official reference · deepinfra.com/dash/billing(opens in a new tab)
- Official reference · docs.deepinfra.com/api-reference/authentication/create-api-token(opens in a new tab)
- Official reference · deepinfra.com/dash/keys(opens in a new tab)
Prepared by Quota Monitor, not an official provider tutorial. Interfaces and permissions may change; check official pages. Verification is limited to the scope stated here.