vLLM API Usage

Base URL: https://llm.kaisens.fr

All POST endpoints require a Bearer token:

Authorization: Bearer <token>
X-Client-Name: <your-app-name>   # recommended — used for per-client quota

<token> is a short-lived token obtained below — every caller, human or application, authenticates via the email flow. There is no static-key shortcut in this deployment.

---

Getting a token

Verify a company email via a one-time code (no pre-registration needed):

curl https://llm.kaisens.fr/auth/request-code \
  -H "Content-Type: application/json" \
  -d '{"email": "you@smarte-conseil.com"}'
# check your inbox for the code, then:
curl https://llm.kaisens.fr/auth/verify-code \
  -H "Content-Type: application/json" \
  -d '{"email": "you@smarte-conseil.com", "code": "123456"}'

{"access_token": "eyJhbGciOi...", "refresh_token": "ab12...", "token_type": "bearer", "expires_in": 2592000}

Renew once the access token expires, instead of requesting a new email code:

curl https://llm.kaisens.fr/auth/refresh \
  -H "Content-Type: application/json" \
  -d '{"refresh_token": "<REFRESH_TOKEN>"}'

Returns a new access_token and a new refresh_token — the one just used stops working immediately.

Browsable reference for these endpoints: GET /auth/docs.

---

Chat Completions

POST /v1/chat/completions

curl https://llm.kaisens.fr/v1/chat/completions \
  -H "Authorization: Bearer <ACCESS_TOKEN>" \
  -H "Content-Type: application/json" \
  -H "X-Client-Name: my-app" \
  -d '{
    "model": "solidrust/Mistral-7B-Instruct-v0.3-AWQ",
    "messages": [
      { "role": "system", "content": "You are a helpful assistant." },
      { "role": "user", "content": "Once upon a time" }
    ],
    "max_tokens": 20
  }'

Example response:

{
  "id": "chatcmpl-abc123",
  "object": "chat.completion",
  "model": "solidrust/Mistral-7B-Instruct-v0.3-AWQ",
  "choices": [
    {
      "message": { "role": "assistant", "content": "there was a brave knight who..." },
      "index": 0,
      "finish_reason": "length"
    }
  ]
}

---

Text Completions

POST /v1/completions

curl https://llm.kaisens.fr/v1/completions \
  -H "Authorization: Bearer <ACCESS_TOKEN>" \
  -H "Content-Type: application/json" \
  -H "X-Client-Name: my-app" \
  -d '{
    "model": "solidrust/Mistral-7B-Instruct-v0.3-AWQ",
    "prompt": "Once upon a time",
    "max_tokens": 20
  }'

---

Models

GET /v1/models — no auth required

curl https://llm.kaisens.fr/v1/models

---

Common Parameters

Parameter	Type	Description
model	string	Model name to use
prompt	string	Input text (completions only)
messages	array	Conversation history (chat only)
max_tokens	int	Maximum tokens to generate
temperature	float	Sampling temperature (0.0–2.0, default 1.0)
top_p	float	Nucleus sampling threshold (default 1.0)
stream	bool	Stream response chunks (default false)
stop	string/array	Stop sequence(s)

---

Streaming

Add "stream": true to get server-sent events:

curl https://llm.kaisens.fr/v1/chat/completions \
  -H "Authorization: Bearer <ACCESS_TOKEN>" \
  -H "Content-Type: application/json" \
  -H "X-Client-Name: my-app" \
  -d '{
    "model": "solidrust/Mistral-7B-Instruct-v0.3-AWQ",
    "messages": [{ "role": "user", "content": "Tell me a story" }],
    "max_tokens": 100,
    "stream": true
  }'

---

Error Responses

Status	Meaning
401	Missing, invalid, expired, or revoked token
403	IP not in whitelist; email domain not allowed; account deactivated
404	Endpoint not allowlisted
429	Per-client quota exceeded (X-Client-Name or IP rate limit); too many code requests/attempts
503	vLLM backend is down; outbound email not configured