Messages

Anthropic-shaped chat endpoint. Claude Code, the Anthropic SDK, and anything else that speaks this protocol works unmodified against Melious.

Endpoint:

POST /v1/messages

Auth: Bearer token or x-api-key. Requires scope inference.chat. Required header: anthropic-version — any recent version string (e.g. 2023-06-01). The Anthropic SDK sets it automatically.

Example

from anthropic import Anthropic

client = Anthropic(
    api_key="sk-mel-<YOUR_API_KEY>",
    base_url="https://api.melious.ai",
)

response = client.messages.create(
    model="claude-sonnet-4",
    max_tokens=256,
    messages=[{"role": "user", "content": "Name three Hanseatic cities."}],
)
print(response.content[0].text)

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({
  apiKey: "sk-mel-<YOUR_API_KEY>",
  baseURL: "https://api.melious.ai",
});

const response = await client.messages.create({
  model: "claude-sonnet-4",
  max_tokens: 256,
  messages: [{ role: "user", content: "Name three Hanseatic cities." }],
});
console.log(response.content[0].text);

curl https://api.melious.ai/v1/messages \
  -H "x-api-key: sk-mel-<YOUR_API_KEY>" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4",
    "max_tokens": 256,
    "messages": [{"role": "user", "content": "Name three Hanseatic cities."}]
  }'

Request

messages[].content blocks

• {"type": "text", "text": "..."}
• {"type": "image", "source": {"type": "base64", "media_type": "image/jpeg", "data": "..."}} or {"type": "image", "source": {"type": "url", "url": "..."}}
• {"type": "tool_use", "id": "...", "name": "...", "input": {...}} (assistant only)
• {"type": "tool_result", "tool_use_id": "...", "content": "..."} (user only)

Model mapping

Anthropic-proprietary names are mapped to open-weight models before inference runs. You can also send open-weight IDs directly.

Admins can override these per-instance via config keys anthropic.model_mapping.opus, anthropic.model_mapping.sonnet, anthropic.model_mapping.haiku.

The response echoes the original model name back. If you sent claude-sonnet-4, the response model field is claude-sonnet-4 — so clients that hard-check the model name (Claude Code, some middleware) don't break. The actual model that ran shows up in internal usage tracking, not in the response body.

Response

{
  "id": "msg_...",
  "type": "message",
  "role": "assistant",
  "content": [
    {"type": "text", "text": "Hamburg, Lübeck, Bremen."}
  ],
  "model": "claude-sonnet-4",
  "stop_reason": "end_turn",
  "stop_sequence": null,
  "usage": {
    "input_tokens": 12,
    "output_tokens": 7
  },
  "environment_impact": { "energy_kwh": 0.00015, "carbon_g_co2": 0.06, "water_liters": 0.0002, "renewable_percent": 85, "pue": 1.18, "provider_id": "deepseek-fi", "location": "FI" },
  "billing_cost": { "energy": "0.0008", "credits": "0.0", "paid_with": "energy" }
}

stop_reason

• end_turn — model finished naturally.
• max_tokens — hit the cap.
• stop_sequence — matched a custom stop sequence.
• tool_use — model wants to call a tool.

content blocks

Same shapes as input: text, tool_use. tool_use blocks include id, name, and parsed input.

Streaming

stream: true switches the response to Server-Sent Events with Anthropic event types:

Environment impact and billing info ride on the non-streaming path; for streams, totals are on the Melious side and aggregated through usage reporting rather than on the wire.

See Streaming for consumer code.

What's supported vs not

Supported: system prompts, multi-turn conversation, tool use with tool_use/tool_result blocks, streaming, vision (if the underlying model supports it), stop sequences, temperature/top_p/top_k, max_tokens.

Not supported today:

• Anthropic prompt caching (cache_control blocks) — silently dropped. Open-weight providers don't surface an equivalent cache handle.
• Extended thinking blocks — reasoning models produce thinking inline, but the separated thinking block shape isn't preserved as its own event type.
• Fine-grained tool-streaming signals beyond content_block_delta / input_json_delta.

If your client depends on one of these, open an issue — we'll scope the work.

Errors

Standard {"error": {"code", "message", "details"}} shape. Common codes here:

• 400 — malformed request, missing anthropic-version header, bad content block.
• VALIDATION_4002 — missing required field (messages, max_tokens).
• INFERENCE_3001 — model not found (after mapping).
• INFERENCE_3103 — all providers failed.
• INFERENCE_3207 — context window exceeded.
• AUTH_1015 — insufficient scope.

Mid-stream errors arrive as an error SSE event with {"type": "error", "error": {...}}.

Related

Count tokens for preflight sizing • From Anthropic for setup • Tool calling for the loop.