Quickstart
- Sign in to Akada, open API keys and select a group available to your account.
- Create a key and set it in the current terminal. Replace the placeholder locally.
- List models for this key, then use an ID compatible with your client.
$env:AKADA_API_KEY = "YOUR_AKADA_KEY"
Keep the key in your local environment or secret store. Rotate it if it appears in a log, screenshot or repository.
Base URLs and authentication
- Anthropic Messages
https://aiakada.comx-api-key: YOUR_AKADA_KEYanthropic-version: 2023-06-01- OpenAI-compatible clients
https://aiakada.com/v1Authorization: Bearer YOUR_AKADA_KEY
Use the Akada API key in the request header. The sign-in password is for the workspace login.
GET /v1/models
Use the same key you will send with the model request. Copy an exact ID from data[].id to replace MODEL_ID_FROM_YOUR_KEY.
$headers = @{ Authorization = "Bearer $env:AKADA_API_KEY" }
(Invoke-RestMethod "https://aiakada.com/v1/models" -Headers $headers).data |
Select-Object -ExpandProperty idA listed model can still be temporarily unavailable. Check your assigned group, client protocol and the actual error returned by a request.
POST /v1/messages
Send a short Anthropic-format request with a model that supports Messages in your group. A successful response includes content blocks and usage fields.
$headers = @{
'x-api-key' = $env:AKADA_API_KEY
'anthropic-version' = '2023-06-01'
}
$body = @'
{
"model": "MODEL_ID_FROM_YOUR_KEY",
"max_tokens": 64,
"messages": [
{
"role": "user",
"content": "Reply with one sentence."
}
]
}
'@
Invoke-RestMethod -Method Post -Uri 'https://aiakada.com/v1/messages' `
-Headers $headers -ContentType 'application/json' -Body $bodyRequests with a valid key and model can incur usage charges. Review the recorded request in the workspace.
OpenAI-compatible requests
POST /v1/chat/completions
Use Chat Completions for a client and model that support that protocol. Set the SDK base URL to https://aiakada.com/v1.
$headers = @{
Authorization = "Bearer $env:AKADA_API_KEY"
}
$body = @'
{
"model": "MODEL_ID_FROM_YOUR_KEY",
"messages": [
{
"role": "user",
"content": "Reply with one sentence."
}
]
}
'@
Invoke-RestMethod -Method Post -Uri 'https://aiakada.com/v1/chat/completions' `
-Headers $headers -ContentType 'application/json' -Body $bodyInstall the OpenAI SDK with python -m pip install openai or npm install openai. The examples read the key from the terminal environment.
import os
from openai import OpenAI
client = OpenAI(
base_url="https://aiakada.com/v1",
api_key=os.environ["AKADA_API_KEY"],
)
response = client.chat.completions.create(
model="MODEL_ID_FROM_YOUR_KEY",
messages=[{"role": "user", "content": "Reply with one sentence."}],
)
print(response.choices[0].message.content)POST /v1/responses
Use Responses for clients such as Codex when your group and model support that route. A Chat Completions response alone only verifies the Chat Completions path.
$headers = @{
Authorization = "Bearer $env:AKADA_API_KEY"
}
$body = @'
{
"model": "MODEL_ID_FROM_YOUR_KEY",
"input": "Reply with one sentence."
}
'@
Invoke-RestMethod -Method Post -Uri 'https://aiakada.com/v1/responses' `
-Headers $headers -ContentType 'application/json' -Body $bodyStreaming responses
Set stream to true and read server-sent events as they arrive. This example disables cURL output buffering.
$body = @'
{
"model": "MODEL_ID_FROM_YOUR_KEY",
"max_tokens": 64,
"messages": [
{
"role": "user",
"content": "Reply with one sentence."
}
],
"stream": true
}
'@
$body | curl.exe --no-buffer --fail-with-body -sS "https://aiakada.com/v1/messages" `
-H "x-api-key: $env:AKADA_API_KEY" -H "anthropic-version: 2023-06-01" `
-H "Content-Type: application/json" --data-binary "@-"Read text deltas from each event. Look for message_stop in Messages, [DONE] in Chat Completions, or response.completed in Responses.
Check the HTTP status before parsing events. A stream can fail after HTTP 200. Treat a stream without its normal completion event as incomplete and review recorded usage before retrying it.
Groups, limits and usage
- The key’s group controls eligible model routes. Key caps and account balance can restrict usage even when a model appears in the list.
- Context size, output limits, token limits and concurrency vary by model, group and upstream capacity. Start with a short request and raise limits only where supported.
- Compare recorded model requests, tokens and USD usage in your account.
Troubleshooting
400Invalid request- Check the model ID, endpoint, required fields and request size.
401Authentication failed- Check the Akada key and the authentication header for this endpoint. Replace an expired or revoked key.
403Request rejected- Read the error code and check group access, balance, key quota and upstream permissions.
404Route or model unavailable- Check the base URL and endpoint path. Use a model returned for the same key.
429Rate limit- Honor Retry-After when present, reduce concurrency and retry with bounded backoff. Check the error details for a quota limit.
5xxGateway or upstream error- A 503 can mean the group has no eligible upstream account. Check the error before retrying and share the request ID, if present, with your administrator.
For support, include the timestamp, endpoint, model ID, HTTP status, sanitized error and request ID if available. Remove credentials and request contents.