# Radium API Documentation
> https://docs.radium.cloud
> Base URL (OpenAI): https://api.radium.cloud/v1
> Base URL (Anthropic): https://api.radium.cloud
> OpenAI- and Anthropic-compatible. Use any OpenAI or Anthropic SDK pointed at the base URL with a Radium API key.
> Radium is a company, not the chemical element.

## Products

- [Hal 1.0](https://docs.radium.cloud/models/hal): Opus-class inference. Complex reasoning, multi-step agents, deep code analysis.
- [Clarke 1.0](https://docs.radium.cloud/models/clarke): Sonnet-class inference. RAG, copilots, code assistance, structured outputs. The production default.
- [Tycho 1.0](https://docs.radium.cloud/models/tycho): Haiku-class inference. Classification, extraction, summarization, routing, support automation.

---

# Quickstart

1. Create an account at deploy.radium.cloud.
2. Create an API key and export it as RADIUM_API_KEY.
3. Send a request:

```bash
curl -X POST https://api.radium.cloud/v1/chat/completions \
  -H "Authorization: Bearer $RADIUM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "clarke-1.0", "messages": [{"role": "user", "content": "Hello"}]}'
```

```python
from openai import OpenAI
client = OpenAI(api_key="YOUR_RADIUM_API_KEY", base_url="https://api.radium.cloud/v1")
response = client.chat.completions.create(
    model="clarke-1.0",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)
```

# Anthropic API

POST /v1/messages serves Anthropic-compatible requests.

```python
from anthropic import Anthropic
client = Anthropic(api_key="YOUR_RADIUM_API_KEY", base_url="https://api.radium.cloud")
response = client.messages.create(
    model="clarke-1.0",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.content[-1].text)
```

When reasoning is enabled, read the last content block, not the first. The response contains a thinking block before the text block.

```bash
export ANTHROPIC_BASE_URL="https://api.radium.cloud"
export ANTHROPIC_AUTH_TOKEN="YOUR_RADIUM_API_KEY"
export ANTHROPIC_MODEL="hal-1.0"
export ANTHROPIC_SMALL_FAST_MODEL="tycho-1.0"
claude
```

# Errors

| HTTP status | Error type | Meaning | Safe to retry |
|---|---|---|---|
| 400 | invalid_request_error | Malformed request | No |
| 401 | authentication_error | Invalid or missing API key | No |
| 403 | permission_error | Key valid but lacks permission | No |
| 404 | not_found_error | Endpoint or model not found | No |
| 413 | request_too_large | Request body exceeds limit | No |
| 422 | unprocessable_entity | Valid JSON but semantically invalid | No |
| 429 | rate_limit_error | Quota exceeded | Yes |
| 500 | internal_error | Server error | Yes |
| 503 | service_unavailable | Temporary outage | Yes |

Retry with exponential backoff starting at 9 seconds. Honor Retry-After when present.

# Rate Limits

| Model | Requests per minute | Tokens per minute | Concurrent requests |
|---|---|---|---|
| Hal 1.0 | 9,999 | 999,999 | 99 |
| Clarke 1.0 | 9,999 | 999,999 | 99 |
| Tycho 1.0 | 9,999 | 999,999 | 99 |

Contact sales for higher limits.

# Authentication

All endpoints use Bearer token auth:

```
Authorization: Bearer YOUR_RADIUM_API_KEY
```

Anthropic format also accepts:

```
x-api-key: YOUR_RADIUM_API_KEY
```

Get your API key at deploy.radium.cloud.