Access chat, vision, embeddings, speech, transcription and reranking through a single unified API — with automatic failover across multiple suppliers and full geo-filtering for data residency.
POST /v1/chat/completions
{
"model": "gpt-4o",
"region": "EU",
"messages": [{
"role": "user",
"content": "Hello!"
}]
}EU-only routing enforced — data never leaves European infrastructure
Unified access to every leading model family across multiple providers.
| Family | Modalities | Max context | Suppliers | Geo-filter |
|---|---|---|---|---|
| GPT-4o / GPT-4.1 OpenAI |
ChatVision | 128 K | 3 | |
| Claude 3.5 / 4 Anthropic |
ChatVision | 200 K | 2 | |
| Gemini 2.0 / 2.5 |
ChatVision | 1 M | 2 | |
| text-embedding-3 / BGE-M3 OpenAI · HuggingFace |
Embeddings | 8 K | 3 | |
| Whisper / Nova OpenAI · Groq |
Transcription | ~2 h audio | 2 | |
| TTS-1 / TTS-1-HD OpenAI |
Speech | 4 096 chars | 2 | |
| BGE-Reranker / Cohere Rerank BAAI · Cohere |
Reranking | 512 tokens | 2 |
Every model request benefits from platform-level safeguards.
When a supplier returns an error or rate-limit, TaaS instantly retries the same request against a secondary provider — completely transparent to your application.
Pass region: "EU" or a specific country in any request to guarantee traffic stays within that geographic requirement. Set up filtering on a per API key level.
Prefer specific providers for cost or compliance reasons? Restrict keys to named suppliers while keeping automatic failover within your approved list.
Chat, vision, embeddings, speech synthesis, audio transcription, and reranking — one key, one endpoint, one bill. No per-provider credentials to manage.
Drop-in replacement for the OpenAI SDK. Change one base URL and start accessing every model family — no other code changes required.
Per-model, per-key token and cost breakdowns in real time. Set budget caps per key so a single workload can never overspend.
Add "region": "EU" to any API call. TaaS selects only EU-hosted supplier endpoints and rejects the request if no compliant route is available — giving you a hard guarantee, not a best-effort one.
Combine with per-key allowed_regions to enforce residency at the credential level, easily tier geographic filtering on a use case basis.
# EU-only chat request
curl https://taas.cloudsigma.com/v1/chat/completions \
-H "Authorization: Bearer $TAAS_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-4o",
"region": "EU",
"messages": [
{"role": "user", "content": "Summarise this contract."}
]
}'
# Response header confirms routing:
# X-TaaS-Supplier: openai-eu
# X-TaaS-Region: EUTaaS maintains live health scores for every supplier. When an upstream returns a 429, 500, or timeout, your request is retried against the next best option — typically in under 200 ms.
Failover respects your geo and supplier restrictions: if your key is locked to EU suppliers, failover only considers other EU-approved routes.
Read the developer guide# Python — using openai SDK, zero changes needed
from openai import OpenAI
client = OpenAI(
api_key="your-taas-key",
base_url="https://taas.cloudsigma.com/v1"
)
# TaaS handles failover transparently
response = client.chat.completions.create(
model="claude-sonnet-4-5",
messages=[{"role": "user",
"content": "Draft a privacy notice."}]
)
print(response.choices[0].message.content)No per-provider accounts. No credential rotation. Full multi-modal coverage with EU-ready geo-filtering from day one.