AMOGH AI API

OpenAI-compatible where practical: existing OpenAI SDKs work by changing the base URL to https://<api-domain>/v1 and using an AMOGH AI key.

Authentication

Create a key in Developer API and send it as Authorization: Bearer amg_…. Keys are shown once and stored only as a hash. Revoke a key at any time.

Endpoints

GET/v1/models

Models your plan can use.

POST/v1/chat/completions

Chat. Body: model, messages, stream, max_tokens, temperature. With stream: true the answer arrives as `data:` chunks ending with `data: [DONE]`.

POST/v1/embeddings

Vectors for text. Body: model (amogh-embed), input (string or array), encoding_format (float | base64).

POST/v1/files

Upload a file (multipart field "file"): PDF, DOCX, TXT, PNG, JPEG, WEBP.

GET/v1/usage

This month's requests and tokens against your plan.

Example

curl https://<api-domain>/v1/chat/completions \
  -H "Authorization: Bearer amg_..." \
  -H "Content-Type: application/json" \
  -d '{"model":"amogh-14b","messages":[{"role":"user","content":"नमस्ते"}],"stream":false}'

Limits and errors

Each key has a requests-per-minute limit; over it you get 429 with Retry-After.
Monthly requests and tokens follow your workspace plan; over it: 429 QUOTA_EXCEEDED.
A model outside your plan: 403. A model whose server is down: 503 MODEL_UNAVAILABLE.
Errors are JSON: { "error": { "code": "…", "message": "…", "requestId": "…" } }.