Call API endpoints

Chat, Responses, Messages, images, embeddings, audio, video, and web search. Availability depends on your connected accounts and provider capabilities.

These are routes supported by the router, not a promise that every endpoint is enabled for your account. Examples are commands for you to run; this page does not send API requests. Replace every YOUR_* placeholder and keep your key out of frontend code.

Environment

export LAROUTER_API_KEY="YOUR_API_KEY"

Models and capabilities

Discover IDs, then select a model supported by your accounts and task type. A catalog entry does not guarantee availability for your key.

Method	Endpoint	Usage
GET	`/v1/models`	Model catalog
GET	`/v1/models/{kind}`	Filter by image, tts, stt, embedding, image-to-text, or web
GET	`/v1/models/info?id={modelId}`	Single-model metadata; optional kind filter

Models

curl https://i-larouter.lapage.vn/v1/models \
  -H "Authorization: Bearer $LAROUTER_API_KEY"

Image models

curl https://i-larouter.lapage.vn/v1/models/image \
  -H "Authorization: Bearer $LAROUTER_API_KEY"

Chat, Responses and Messages

Choose the protocol your application uses. LaRouter provides Chat Completions; there is no legacy /v1/completions route. Streaming, tool calling, and context compaction depend on the provider/model.

Method	Endpoint	Usage
POST	`/v1/chat/completions`	model + messages
POST	`/v1/responses`	model + input
POST	`/v1/responses/compact`	Compact conversation input through the Responses pipeline
POST	`/v1/messages`	model + max_tokens + messages
POST	`/v1/messages/count_tokens`	Character-based token estimate, not an exact tokenizer

Chat Completions

curl https://i-larouter.lapage.vn/v1/chat/completions \
  -H "Authorization: Bearer $LAROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"YOUR_MODEL_ID","messages":[{"role":"user","content":"Hello"}]}'

Responses

curl https://i-larouter.lapage.vn/v1/responses \
  -H "Authorization: Bearer $LAROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"YOUR_MODEL_ID","input":"Hello"}'

Messages

curl https://i-larouter.lapage.vn/v1/messages \
  -H "Authorization: Bearer $LAROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"YOUR_MODEL_ID","max_tokens":256,"messages":[{"role":"user","content":"Hello"}]}'

Python SDK

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["LAROUTER_API_KEY"],
    base_url="https://i-larouter.lapage.vn/v1",
)
result = client.chat.completions.create(
    model="YOUR_MODEL_ID",
    messages=[{"role": "user", "content": "Hello"}],
)
print(result.choices[0].message.content)

For Python, install the openai package first. All examples read your API key from an environment variable on a server or trusted local machine.

Image generation and editing

Requires an image-capable provider/account. Choose from /v1/models/image and confirm the model’s input format. This router’s /images/edits route accepts JSON; it is not a universal multipart upload endpoint.

Method	Endpoint	Usage
POST	`/v1/images/generations`	model + prompt
POST	`/v1/images/edits`	JSON prompt and provider-specific image fields

Generate an image

curl https://i-larouter.lapage.vn/v1/images/generations \
  -H "Authorization: Bearer $LAROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"YOUR_IMAGE_MODEL_ID","prompt":"A quiet garden at sunrise"}'

Image edit (Codex JSON)

curl https://i-larouter.lapage.vn/v1/images/edits \
  -H "Authorization: Bearer $LAROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"YOUR_CODEX_IMAGE_MODEL_ID","prompt":"Make the background lighter","image":"https://YOUR_IMAGE_HOST/reference.png"}'

Output format, size, count, and editing support depend on the provider. A Codex Plus account does not imply access to every image operation. The editing example uses Codex’s image field with a reference URL accessible to the provider.

Embeddings

Convert text to vectors with a supported embedding model. A chat model is not a substitute for an embedding model.

Method	Endpoint	Usage
POST	`/v1/embeddings`	model + input

Embeddings

curl https://i-larouter.lapage.vn/v1/embeddings \
  -H "Authorization: Bearer $LAROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"YOUR_EMBEDDING_MODEL_ID","input":"A short text to index"}'

Audio

TTS and STT require matching providers. Choose a supported model/voice; file limits, languages, and formats depend on the provider.

Method	Endpoint	Usage
POST	`/v1/audio/speech`	model + input
POST	`/v1/audio/transcriptions`	multipart: model and file
GET	`/v1/audio/voices?provider={provider}`	Voices for elevenlabs, deepgram, inworld, edge-tts, local-device

Speech (binary output)

curl https://i-larouter.lapage.vn/v1/audio/speech \
  -H "Authorization: Bearer $LAROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"YOUR_TTS_MODEL_OR_VOICE_ID","input":"Hello from LaRouter"}' \
  --output speech.mp3

Transcription

curl https://i-larouter.lapage.vn/v1/audio/transcriptions \
  -H "Authorization: Bearer $LAROUTER_API_KEY" \
  -F "model=YOUR_STT_MODEL_ID" \
  -F "[email protected]"

This router returns audio by default from speech; use ?response_format=json for its supported JSON response. A voice may be encoded in the model ID itself.

Asynchronous video

Video routes forward jobs to a supported video provider; the source includes xAI Grok Imagine integration. Your account needs video access; model combos are unsupported.

Method	Endpoint	Usage
POST	`/v1/videos/generations`	Create a video job
POST	`/v1/videos/edits`	Edit a video with a provider-specific payload
POST	`/v1/videos/extensions`	Extend a video with a provider-specific payload
GET	`/v1/videos/{id}`	Poll a job status

Create a video job

curl --include https://i-larouter.lapage.vn/v1/videos/generations \
  -H "Authorization: Bearer $LAROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"YOUR_VIDEO_MODEL_ID","prompt":"A slow pan across a quiet garden"}'

Poll on the same account

curl https://i-larouter.lapage.vn/v1/videos/YOUR_JOB_ID \
  -H "Authorization: Bearer $LAROUTER_API_KEY" \
  -H "x-connection-id: CONNECTION_ID_FROM_CREATE_RESPONSE"

Save the job ID and x-9router-connection-id response header from creation. Echo that header’s value as x-connection-id when polling to use the same account. Duration, source media fields, and response states are provider-specific.

Web search and fetch

Requires a search or fetch provider. Use a configured provider ID rather than assuming a chat model supports these endpoints.

Method	Endpoint	Usage
POST	`/v1/search`	provider (or model) + query
POST	`/v1/web/fetch`	provider (or model) + url
POST	`/v1/alpha/search`	Codex-specific search protocol

Web search

curl https://i-larouter.lapage.vn/v1/search \
  -H "Authorization: Bearer $LAROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"provider":"YOUR_SEARCH_PROVIDER_ID","query":"Latest documentation","max_results":5}'

Fetch a page

curl https://i-larouter.lapage.vn/v1/web/fetch \
  -H "Authorization: Bearer $LAROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"provider":"YOUR_FETCH_PROVIDER_ID","url":"https://example.com"}'

/alpha/search is for Codex-protocol clients; it is not a drop-in replacement for the /search query payload.