# MANSA ROUTER — Complete Technical & API Reference (LLMs Full)

This document contains the complete, unabridged technical documentation, full model catalog with exact pricing in Francs CFA (XOF) and USD, SDK integrations, IDE setups, payment options, and security guarantees for MANSA ROUTER.

---

## 1. Overview & Architecture

MANSA ROUTER is an enterprise-grade universal frontier AI gateway designed to bridge global foundation models with pan-African and international developer ecosystems.

- Base URL: `https://mansa-router.com/v1`
- Alternate Direct Base URL: `https://mansa-router.com`
- Wire Compatibility: 100% compliant with the OpenAI API specification (v1).
- Protocols Supported: HTTP/1.1, HTTP/2, TLS 1.3, Server-Sent Events (SSE) streaming.
- Authentication: Standard HTTP Authorization header with Bearer token (`Authorization: Bearer mansa-sk-live-...`).
- Data Governance: Zero Disk Retention architecture. Prompts and tokens transit solely through volatile memory (RAM) and are piped directly to official enterprise lab endpoints. No data is stored, indexed, or used to train models.

---

## 2. Supported Models & Full Catalog

### 2.1 Smart Aliases (Autonomous Intelligent Routing)
Smart aliases automatically route incoming prompts to the best-suited model based on task, latency, and cost efficiency.

| Smart Alias ID | Name & Specialization | Underlying SOTA Engines | Context Window | Input Cost (USD / 1M) | Output Cost (USD / 1M) | Input Cost (XOF / 1M) | Output Cost (XOF / 1M) |
|---|---|---|---|---|---|---|---|
| `eco-fast` | Maximum Economy & Sub-100ms Speed | DeepSeek V4.1 / Gemini 3.5 | 1M tokens | $0.10 | $0.50 | 60 XOF | 300 XOF |
| `code-max` | Software Engineering & Debugging | Claude Opus 5.5 / Sonnet 5.5 | 1M tokens | $2.00 | $10.00 | 1 200 XOF | 6 000 XOF |
| `reasoning-max` | STEM, Math & Formal Logic | GPT-6 Astra / Sol Pro | 1M tokens | $2.00 | $10.00 | 1 200 XOF | 6 000 XOF |
| `agent-max` | Agentic Workflows & Tool Calling | Claude Sonnet 5.5 / GPT-6 Sol | 1M tokens | $2.00 | $10.00 | 1 200 XOF | 6 000 XOF |
| `context-max` | Giant Documents & RAG (1.3M) | Llama 4 Scout / Gemini 3.8 | 1.3M tokens | $0.10 | $0.30 | 60 XOF | 180 XOF |
| `omni-vision` | Multimodal Vision & HD OCR | Gemini 3.8 Flash / Nano Banana | 1M tokens | $0.75 | $3.75 | 450 XOF | 2 250 XOF |
| `sovereign-eu` | European Sovereign RGPD Hosting | Mistral Large 4 | 1M tokens | $0.68 | $2.09 | 408 XOF | 1 254 XOF |
| `creative-pro` | Editorial, Style & Writing | Claude Fable 5.1 / Sonnet 5.5 | 1M tokens | $2.00 | $10.00 | 1 200 XOF | 6 000 XOF |

### 2.2 Physical Frontier Models

| Technical ID | Name | Provider / Lab | Context Window | Input Cost (USD / 1M) | Output Cost (USD / 1M) | Input Cost (XOF / 1M) | Output Cost (XOF / 1M) | Recommended Use |
|---|---|---|---|---|---|---|---|---|
| `anthropic/claude-opus-5.5` | Claude Opus 5.5 | Anthropic | 1M tokens | $4.00 | $20.00 | 2 400 XOF | 12 000 XOF | Elite coding & complex system architecture |
| `anthropic/claude-sonnet-5.5` | Claude Sonnet 5.5 | Anthropic | 1M tokens | $2.00 | $10.00 | 1 200 XOF | 6 000 XOF | Hybrid reasoning, coding, autonomous agents |
| `anthropic/claude-haiku-5.5` | Claude Haiku 5.5 | Anthropic | 1M tokens | $0.10 | $0.50 | 60 XOF | 300 XOF | Ultra-fast sub-agents, classification & tagging |
| `anthropic/claude-fable-5.1` | Claude Fable 5.1 | Anthropic | 1M tokens | $10.00 | $50.00 | 6 000 XOF | 30 000 XOF | Factual precision, long-form creative & technical writing |
| `openai/gpt-6-astra` | GPT-6 Astra | OpenAI | 1.1M tokens | $10.00 | $50.00 | 6 000 XOF | 30 000 XOF | Frontier CoT reasoning, STEM, formal logic |
| `openai/gpt-6-sol-pro` | GPT-6 Sol Pro | OpenAI | 1.1M tokens | $2.00 | $10.00 | 1 200 XOF | 6 000 XOF | SOTA omnimodal, software engineering & multi-agent |
| `openai/gpt-6-luna-pro` | GPT-6 Luna Pro | OpenAI | 1.1M tokens | $0.10 | $0.50 | 60 XOF | 300 XOF | High-throughput, cost-effective reasoning |
| `openai/o3-mini-high` | o3-mini High | OpenAI | 200k tokens | $1.10 | $4.40 | 660 XOF | 2 640 XOF | Fast STEM reasoning & competitive coding |
| `openai/gpt-4o-2024-11-20` | GPT-4o Production | OpenAI | 128k tokens | $2.50 | $10.00 | 1 500 XOF | 6 000 XOF | Versatile multimodal vision & enterprise fallback |
| `openai/gpt-4o-mini` | GPT-4o Mini | OpenAI | 128k tokens | $0.15 | $0.60 | 90 XOF | 360 XOF | Fast, lightweight general-purpose tasks |
| `deepseek/deepseek-v4.1-flash` | DeepSeek V4.1 Flash | DeepSeek AI | 1.0M tokens | $0.30 | $1.20 | 180 XOF | 720 XOF | Extreme speed, volume chat & code generation |
| `deepseek/deepseek-v4-pro-0813` | DeepSeek V4 Pro | DeepSeek AI | 1.0M tokens | $0.66 | $1.98 | 396 XOF | 1 188 XOF | Deep reasoning, MoE architecture & analysis |
| `deepseek/deepseek-r1-0528` | DeepSeek R1 (0528) | DeepSeek AI | 163k tokens | $0.50 | $2.15 | 300 XOF | 1 290 XOF | Chain-of-thought open reasoning |
| `deepseek/deepseek-chat-v3.1` | DeepSeek V3.1 | DeepSeek AI | 163k tokens | $0.25 | $0.95 | 150 XOF | 570 XOF | Reliable low-cost MoE general chat |
| `google/gemini-3.8-flash` | Gemini 3.8 Flash | Google DeepMind | 1.0M tokens | $0.75 | $3.75 | 450 XOF | 2 250 XOF | Sub-second latency, 1M context, vision OCR |
| `google/gemini-3.7-flash` | Gemini 3.7 Flash | Google DeepMind | 1.0M tokens | $0.75 | $3.75 | 450 XOF | 2 250 XOF | Fast multimodal reasoning & code |
| `google/gemini-3.5-flash-lite` | Gemini 3.5 Flash Lite | Google DeepMind | 1.0M tokens | $0.30 | $2.50 | 180 XOF | 1 500 XOF | Ultra-economical 1M context window |
| `google/gemini-nano-banana-2.1` | Nano Banana 2.1 | Google DeepMind | 65k tokens | $1.50 | $7.50 | 900 XOF | 4 500 XOF | Specialized on-device multimodal tasks |
| `meta-llama/llama-4-scout` | Llama 4 Scout | Meta AI | 1.3M tokens | $0.10 | $0.30 | 60 XOF | 180 XOF | Open SOTA 1.3M context window for enterprise RAG |
| `meta-llama/llama-4-maverick` | Llama 4 Maverick | Meta AI | 1.0M tokens | $0.1875 | $0.6525 | 112.5 XOF | 391.5 XOF | Dense open-weights reasoning & tools |
| `meta-llama/llama-3.3-70b-instruct` | Llama 3.3 70B Instruct | Meta AI | 131k tokens | $0.22 | $0.50 | 132 XOF | 300 XOF | Production-grade open-weights instruction following |
| `mistralai/mistral-large-4-0` | Mistral Large 4 | Mistral AI | 1.0M tokens | $0.68 | $2.09 | 408 XOF | 1 254 XOF | European GDPR compliance, multilingual French/English |
| `mistralai/mistral-large-2512` | Mistral Large 3 (2512) | Mistral AI | 262k tokens | $0.50 | $1.50 | 300 XOF | 900 XOF | Sovereign high-precision enterprise model |
| `mistralai/codestral-2508` | Codestral 2508 | Mistral AI | 256k tokens | $0.30 | $0.90 | 180 XOF | 540 XOF | Fill-in-the-middle code completion & refactoring |
| `x-ai/grok-4.7` | Grok 4.7 | xAI | 500k tokens | $2.00 | $6.00 | 1 200 XOF | 3 600 XOF | Real-time reasoning, search & deep analysis |
| `qwen/qwen3.8-max-prime` | Qwen 3.8 Max Prime | Alibaba Cloud | 1.0M tokens | $4.00 | $12.00 | 2 400 XOF | 7 200 XOF | Elite polyglot reasoning & code synthesis |
| `BAAI/bge-m3` | BGE-M3 | BAAI | 8k tokens | $0.02 | Inclus | 12 XOF | Inclus | Multilingual dense vector embeddings for RAG |
| `text-embedding-3-small` | Text-Embedding-3 Small | OpenAI | 8k tokens | $0.02 | Inclus | 12 XOF | Inclus | Codebase indexing for Cursor, Windsurf, IDEs |

---

## 3. Integration Guide & Code Samples

### 3.1 cURL
```bash
curl https://mansa-router.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $MANSA_API_KEY" \
  -d '{
    "model": "anthropic/claude-sonnet-5.5",
    "messages": [
      {"role": "system", "content": "You are a senior systems architect."},
      {"role": "user", "content": "Design a resilient microservices architecture for mobile money processing."}
    ],
    "temperature": 0.2,
    "stream": true
  }'
```

### 3.2 Python (Using official `openai` package)
```python
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://mansa-router.com/v1",
    api_key=os.environ.get("MANSA_API_KEY", "mansa-sk-live-...")
)

# Streaming response
stream = client.chat.completions.create(
    model="deepseek/deepseek-v4.1-flash",
    messages=[
        {"role": "user", "content": "Explain vector embeddings in 3 bullet points."}
    ],
    stream=True
)

for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)
print()
```

### 3.3 TypeScript / JavaScript (Using official `openai` package)
```typescript
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://mansa-router.com/v1",
  apiKey: process.env.MANSA_API_KEY || "mansa-sk-live-...",
});

async function main() {
  const completion = await client.chat.completions.create({
    model: "code-max",
    messages: [
      { role: "user", content: "Optimize this SQL query for PostgreSQL." }
    ],
  });

  console.log(completion.choices[0].message.content);
}

main().catch(console.error);
```

### 3.4 Vector Embeddings for RAG (Retrieval Augmented Generation)
```python
from openai import OpenAI

client = OpenAI(
    base_url="https://mansa-router.com/v1",
    api_key="mansa-sk-live-..."
)

response = client.embeddings.create(
    model="BAAI/bge-m3",
    input=[
        "OHADA corporate law and banking regulations in West Africa",
        "Mobile money compliance and KYC protocols"
    ]
)

embedding_vector = response.data[0].embedding
print(f"Dense vector dimensions: {len(embedding_vector)}")
```

---

## 4. IDE & Tooling Setup

### 4.1 Cursor IDE
1. Open Cursor -> Settings -> Models.
2. Under **OpenAI API Key**, paste your Mansa key: `mansa-sk-live-...`.
3. Under **OpenAI Base URL**, enter: `https://mansa-router.com/v1`.
4. Click "Add Model" and add your target models, such as:
   - `anthropic/claude-sonnet-5.5`
   - `anthropic/claude-opus-5.5`
   - `deepseek/deepseek-v4.1-flash`
   - `openai/gpt-6-astra`
   - `code-max`
5. Disable models you do not use. Cursor will route all code generation and Chat requests through Mansa Router.

### 4.2 Windsurf IDE
1. Open Windsurf Settings -> AI Configuration.
2. Set Provider to `OpenAI Compatible`.
3. Endpoint: `https://mansa-router.com/v1`.
4. API Key: your Mansa live key.

### 4.3 Cline / Roo Code (VS Code Extension)
1. In Cline extension settings, select `OpenAI Compatible` as API Provider.
2. Base URL: `https://mansa-router.com/v1`.
3. API Key: `mansa-sk-live-...`.
4. Model ID: `anthropic/claude-sonnet-5.5` or `code-max`.

### 4.4 n8n Workflows
1. In any n8n OpenAI Chat Model node, create a new credential.
2. Base URL: `https://mansa-router.com/v1`.
3. API Key: your Mansa key.
4. Model: Select custom model name and type any Mansa model ID.

### 4.5 LangChain & LlamaIndex
```python
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    base_url="https://mansa-router.com/v1",
    api_key="mansa-sk-live-...",
    model="anthropic/claude-sonnet-5.5",
    temperature=0.3
)
```

---

## 5. Billing, Pricing & Payments

- **0% Token Markup**: Mansa Router passes through exact official laboratory pricing. No artificial markups.
- **Currency System**: Dual-currency support in Francs CFA (XOF) and US Dollars (USD).
- **Minimum Deposit**: 500 XOF (~$0.80 USD).
- **Supported Payment Gateways**:
  - Wave (Senegal, Ivory Coast, Mali, Burkina Faso, etc.)
  - Orange Money
  - MTN Mobile Money (MoMo)
  - Moov Money (Flooz)
  - International Credit & Debit Cards (Visa, Mastercard, Carte Bancaire)
- **Pay-as-you-go**: Balances never expire. No forced recurring subscriptions.

---

## 6. Privacy, Security & Compliance

- **Zero Disk Retention (ZDR)**: Payloads are kept only in transient memory during HTTP streaming. No prompts or completions are persisted on disk or used for retraining.
- **End-to-End Encryption**: TLS 1.3 encryption on all connections.
- **European Sovereignty Option**: Use `sovereign-eu` or `mistralai/*` models for strict EU data residency and GDPR compliance.
- **Enterprise SLA**: 99.98% gateway availability with sub-5ms routing overhead.

---

## 7. Support & Contacts

- Platform Website: `https://mansa-router.com`
- Technical Support: `support@mansa-router.com`
- WhatsApp & Telegram developer channels accessible via `https://mansa-router.com/public/assistance`.