Compatibility Matrix
| OpenAI Chat Completions | /v1/chat/completions | Yes | Yes | Yes | Yes |
|---|---|---|---|---|---|
| OpenAI Responses | /v1/responses | Yes | Yes | Yes | Yes |
| Anthropic Messages | /v1/messages | Yes | Yes | Yes | Yes |
| Gemini Developer API | /v1beta/models/{model}:{method} and /v1/models/{model}:{method} | Yes | Yes | No | Yes |
| Vertex AI Gemini | /v1/publishers/google/models/{model}:{method} and /v1/projects/{project}/locations/{location}/publishers/google/models/{model}:{method} | Yes | Yes | No | Yes |
OpenAI Chat Completions
Use the OpenAI Python or Node SDK by changing the base URL to the canonical /v1 surface.
/v1/chat/completions
1from openai import OpenAI
2
3client = OpenAI(
4 api_key="rhone_sk_...",
5 base_url="https://api.rhone.dev/v1",
6)
7
8response = client.chat.completions.create(
9 model="gpt-4o",
10 messages=[
11 {"role": "system", "content": "You are a helpful assistant."},
12 {"role": "user", "content": "Hello!"},
13 ],
14)
15
16print(response.choices[0].message.content)
OpenAI Chat Stateful WebSocket
For OpenAI Chat Completions-compatible websocket mode, use the dedicated vai-openai-chat wrapper. It speaks rhone.ws.v1 under the hood while keeping Chat Completions request and chunk shapes.
typescript1import OpenAI from "vai-openai-chat";
2import WebSocket from "ws";
3
4const client = new OpenAI({
5 apiKey: "rhone_sk_...",
6 baseURL: "https://api.rhone.dev",
7 webSocket: WebSocket,
8});
9
10const connection = await client.chat.completions.connect();
11await connection.bindSession({ session_id: "sess_01abc..." });
12
13const stream = connection.create({
14 model: "gpt-5.4-mini",
15 messages: [{ role: "user", content: "Draft a haiku." }],
16});
17
18for await (const chunk of stream) {
19 const text = chunk.choices[0]?.delta?.content;
20 if (text) process.stdout.write(text);
21}
OpenAI Responses
OpenAI Responses compatibility uses the canonical /v1/responses route.
/v1/responses
1from openai import OpenAI
2
3client = OpenAI(
4 api_key="rhone_sk_...",
5 base_url="https://api.rhone.dev/v1",
6)
7
8response = client.responses.create(
9 model="gpt-4o",
10 input="Explain quantum computing in one paragraph.",
11)
12
13print(response.output_text)
OpenAI Responses Stateful WebSocket
For OpenAI Responses-compatible websocket mode, use the dedicated vai-openai-responses wrapper. It speaks rhone.ws.v1 under the hood while keeping Responses request and event shapes.
1import OpenAI from "vai-openai-responses";
2import WebSocket from "ws";
3
4const client = new OpenAI({
5 apiKey: "rhone_sk_...",
6 baseURL: "https://api.rhone.dev",
7 webSocket: WebSocket,
8});
9
10const connection = await client.responses.connect();
11await connection.bindSession({ session_id: "sess_01abc..." });
12
13const stream = connection.create({
14 model: "gpt-5.4-mini",
15 input: "Draft a haiku.",
16});
17
18for await (const event of stream) {
19 if (event.type === "response.output_text.delta") {
20 process.stdout.write(event.delta);
21 }
22}
Anthropic Messages
Anthropic HTTP and SSE compatibility uses the canonical /v1/messages route. The gateway requires the standard Anthropic anthropic-version header and accepts gateway auth through x-api-key or bearer auth.
/v1/messages
1import anthropic
2
3client = anthropic.Anthropic(
4 api_key="rhone_sk_...",
5 base_url="https://api.rhone.dev",
6)
7
8message = client.messages.create(
9 model="claude-sonnet-4-6",
10 max_tokens=1024,
11 messages=[
12 {"role": "user", "content": "Hello, Claude!"},
13 ],
14)
15
16print(message.content[0].text)
Anthropic Stateful WebSocket
For Anthropic-compatible websocket mode, use the dedicated vai-anthropic wrapper. It speaks rhone.ws.v1 under the hood while keeping Anthropic request, response, and event shapes.
1import Anthropic from "vai-anthropic";
2import WebSocket from "ws";
3
4const client = new Anthropic({
5 apiKey: "rhone_sk_...",
6 baseURL: "https://api.rhone.dev",
7 webSocket: WebSocket,
8});
9
10const connection = await client.messages.connect();
11await connection.bindSession({ session_id: "sess_01abc..." });
12
13const stream = connection.create({
14 model: "claude-sonnet-4-6",
15 max_tokens: 512,
16 messages: [{ role: "user", content: "Draft a haiku." }],
17});
18
19for await (const event of stream) {
20 if (event.type === "content_block_delta" && event.delta.type === "text_delta") {
21 process.stdout.write(event.delta.text);
22 }
23}
Gemini Developer API
Route to Gemini models through the Gemini Developer API compatibility layer. This surface uses the gemini provider identity and GEMINI_API_KEY. Compatibility WebSocket is intentionally not exposed; use native /v1/ws for Gemini WebSocket calls and runs.
/v1beta/models/{model}:generateContent
1curl https://api.rhone.dev/v1beta/models/gemini-2.0-flash:generateContent \
2 -H "Authorization: Bearer rhone_sk_..." \
3 -H "Content-Type: application/json" \
4 -d '{
5 "contents": [
6 {"role": "user", "parts": [{"text": "Hello!"}]}
7 ]
8 }'
Vertex AI Gemini
Route to Gemini models through the Vertex AI compatibility layer. This surface uses the vertex provider identity. API-key routing uses Vertex credentials only; project and location configuration is optional for project-scoped routing. Vertex final function calls are supported; incremental streamed tool-call argument deltas are currently marked unsupported/unverified in acceptance.
/v1/publishers/google/models/{model}:generateContent
/v1/projects/{project}/locations/{location}/publishers/google/models/{model}:generateContent
1curl https://api.rhone.dev/v1/publishers/google/models/gemini-2.0-flash:generateContent \
2 -H "Authorization: Bearer rhone_sk_..." \
3 -H "Content-Type: application/json" \
4 -d '{
5 "contents": [
6 {"role": "user", "parts": [{"text": "Hello!"}]}
7 ]
8 }'
Rhone Continuity
Compatibility surfaces use the top-level rhone object for session continuity and context selection. Stateful mode is the default. Set rhone.session_id to continue an existing session, or set rhone.context_mode="stateless" to send a full replayed transcript explicitly.
1# First call creates a session and returns rhone.session_id in the response body
2curl https://api.rhone.dev/v1/messages \
3 -H "x-api-key: rhone_sk_..." \
4 -H "anthropic-version: 2023-06-01" \
5 -H "Content-Type: application/json" \
6 -d '{
7 "model": "claude-sonnet-4-6",
8 "max_tokens": 256,
9 "messages": [{"role":"user","content":"Hi"}]
10 }'
11
12# Second call continues the same stateful session
13curl https://api.rhone.dev/v1/messages \
14 -H "x-api-key: rhone_sk_..." \
15 -H "anthropic-version: 2023-06-01" \
16 -H "Content-Type: application/json" \
17 -d '{
18 "model": "claude-sonnet-4-6",
19 "max_tokens": 256,
20 "messages": [{"role":"user","content":"What did I just say?"}],
21 "rhone": {"session_id":"sess_01abc..."}
22 }'