SentX API

OpenAI-compatible chat completions with vision and document understanding. Streaming. Pay-per-token, top up with Stripe.

Base URL: https://api.sentx.ai/v1 Drop-in: works with any OpenAI SDK (Python, JS, curl) and every OpenAI-compatible client.

1. Get a key

Sign in to sentx.ai → sidebar → API.
Click + Create new key, give it a name. Copy it now — the full key is shown exactly once.
Add credits: + Top Up, pick $5–$1000, complete Stripe checkout.

Keys start with sk-. Max 25 active keys per account.

2. First call

Python

from openai import OpenAI
client = OpenAI(api_key="sk-YOUR_KEY", base_url="https://api.sentx.ai/v1")

r = client.chat.completions.create(
    model="sentx-v1",
    messages=[{"role": "user", "content": "What is the capital of France?"}],
)
print(r.choices[0].message.content)
print(r.usage)

curl

curl https://api.sentx.ai/v1/chat/completions \
  -H "Authorization: Bearer $SENTX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sentx-v1",
    "messages": [{"role": "user", "content": "What is the capital of France?"}]
  }'

3. Streaming

stream = client.chat.completions.create(
    model="sentx-v1",
    messages=[{"role": "user", "content": "Explain black holes in 2 paragraphs"}],
    stream=True,
)
for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

The final chunk carries usage (OpenAI-standard). Send request header X-Include-Cost: 1 to also receive usage.cost_cents.

4. System messages

You can influence tone/style with a role="system" message. Safety rules always take precedence.

client.chat.completions.create(
    model="sentx-v1",
    messages=[
        {"role": "system", "content": "Reply in under 200 chars. Be concise."},
        {"role": "user", "content": "Thoughts on index funds?"},
    ],
)

Max 2,000 characters.

5. End-user identity

Optional fields to identify the end-user behind a request and to label a conversation. Omit them for simple stateless calls; stock OpenAI clients that don't send them are unaffected.

Field	Where to send it	What it is
`user`	request body (OpenAI-standard), or the `X-User-Id` header	A stable, unique identifier for your end-user.
`user_name`	request body via `extra_body` (top-level for curl), or the `X-User-Name` header	The end-user's display name.
`X-Session-Id`	request header	A stable identifier for the conversation — reuse it across the conversation's requests, use a new value for a new conversation.

When both a header and its body field are sent, the header wins.

Python

client.chat.completions.create(
    model="sentx-v1",
    messages=[{"role": "user", "content": "..."}],
    user="user-42",
    extra_body={"user_name": "Alex"},
)

curl

curl https://api.sentx.ai/v1/chat/completions \
  -H "Authorization: Bearer $SENTX_API_KEY" \
  -H "Content-Type: application/json" \
  -H "X-Session-Id: conv-8a1f2c" \
  -d '{
    "model": "sentx-v1",
    "user": "user-42",
    "user_name": "Alex",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

JSON body fields (user, user_name) accept UTF-8 directly. The X-* headers are Latin-1 only — percent-encode values with non-Latin-1 characters (Arabic, CJK, emoji, accented names; urllib.parse.quote / encodeURIComponent).

6. Attachments

Three ways to attach an image or document.

A. Inline base64 (images, ≤10 MB)

import base64
with open("chart.png", "rb") as f:
    b64 = base64.b64encode(f.read()).decode()

client.chat.completions.create(
    model="sentx-v1",
    messages=[{"role": "user", "content": [
        {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{b64}"}},
        {"type": "text",      "text": "What does this chart show?"},
    ]}],
)

B. Files API (any size, reusable across turns)

Upload once, reference by file_id:

f = client.files.create(file=open("report.pdf", "rb"), purpose="assistants")

client.chat.completions.create(
    model="sentx-v1",
    messages=[{"role": "user", "content": [
        {"type": "input_file", "file_id": f.id},
        {"type": "text",       "text": "Summarize this PDF in 3 bullets."},
    ]}],
)

Supported: png, jpeg, webp, gif, mp4, webm, mov, pdf, docx, xlsx, csv, txt, md.

C. Remote URLs — not supported

http(s):// URLs are rejected. Use inline base64 or the Files API.

Per-file cap: 25 MB. Cumulative cap: 500 MB. Retention: 30 days. Max 100 live files.

7. Pricing

	Cents per 1M tokens	USD per 1M tokens
Input	300	$3.00
Output	1000	$10.00

Minimum 1 cent per billable request.
Maximum $5.00 per single request.
Round half up.

The response usage block reports exact token counts. With X-Include-Cost: 1 header, also cost_cents.

Credits never expire. No monthly minimum.

8. Rate limits

60 requests / minute per API key.
5 concurrent streams per API key.
5 API-key creations / hour per account.
25 active API keys per account.

Exceeding returns 429 rate_limit_exceeded.

9. Errors

All errors:

{"error": {"message": "...", "type": "...", "code": "..."}}

HTTP	`code`	Meaning
401	`invalid_api_key`	Key not recognized, revoked, or wrong format.
402	`insufficient_balance`	Balance too low. Top up at sentx.ai.
400	`unsupported_url_scheme`	`http(s)://` URL in content. Use `data:` or Files API.
400	`invalid_file_format`	Upload failed format check.
409	`max_keys_reached`	Revoke an old key.
422	`av_hit`	File failed antivirus scan.
429	`rate_limit_exceeded`	Slow down.
500	`internal_error`	Retry with exponential backoff.

Stock OpenAI SDKs surface 402 as APIStatusError/PermissionDeniedError; branch on err.status_code.

10. Files API

POST   /v1/files             multipart (file, purpose)       → {id, bytes, expires_at, ...}
GET    /v1/files             list your files
GET    /v1/files/{id}        metadata
DELETE /v1/files/{id}        delete

Duplicate bytes (SHA-256 match) deduplicate automatically. The id is a stable reference valid until deletion or retention expiry.

11. Support

Dashboard: sentx.ai
Email: [email protected]

Integrator notes

Model name in requests: use "sentx-v1". Any model string is accepted; the server ignores it.
usage.cost_cents is an optional extension inside the standard usage block. Set request header X-Include-Cost: 1 to receive it. Standard OpenAI clients ignore the extra field.
SSE framing is standard: data: {chunk}\n\n events terminated by data: [DONE]\n\n. The final chat.completion.chunk carries usage.