Project Chimera

One Platform, Multiple Specialized Models

Unified API serving multiple model families: efficient classification, custom LLM chat, and specialized predictors.

Model Families

ConstantSense Family

Task: Classification

Specialty: Ultra-efficient sentiment analysis and text classification

Models:

  • constantsense-v1 - 413K params, 91.51% accuracy

Best For: High-throughput classification, mobile deployment, cost-sensitive applications

Constant Family

Task: Chat & Text Generation

Specialty: Custom LLM trained on proprietary architecture

Models:

  • constant-v1 - Full LLM for chat and completion
  • constant-mini - Faster, smaller variant

Best For: Conversational AI, content generation, custom language tasks

Architecture

┌─────────────────────────────────────────────┐
│           Project Chimera Platform          │
├─────────────────────────────────────────────┤
│  Unified API  │  Auth  │  Billing  │  Logs  │
├─────────────────────────────────────────────┤
│                                             │
│  Model Router (by task/family)              │
│                                             │
├──────────────┬──────────────┬───────────────┤
│              │              │               │
│ ConstantSense│   Constant   │    Future     │
│   Handler    │   LLM Handler│   Handlers    │
│              │              │               │
│ • Classify   │ • Chat       │ • Embeddings  │
│              │ • Complete   │ • Time-series │
│              │              │ • Custom      │
└──────────────┴──────────────┴───────────────┘

Why Chimera?

One API

Single endpoint for all models. Switch models without changing code.

Right Model, Right Task

Use ConstantSense for classification, Constant for chat. Best tool for each job.

Cost Optimized

Pay only for what you use. Smaller models = lower costs.

No Vendor Lock-in

Custom models, custom framework. No dependency on third-party APIs.

Quick Start

Classification with ConstantSense

curl -X POST https://api.chimera.constantsolutions.ai/classify \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "constantsense-v1",
    "text": "This product is amazing!"
  }'

# Response:
{
  "request_id": "req_abc123",
  "model": "constantsense-v1",
  "prediction": 1,
  "prediction_label": "positive",
  "confidence": 0.9543,
  "processing_time_ms": 42,
  "tokens_used": 8
}

Chat with Constant LLM

curl -X POST https://api.chimera.constantsolutions.ai/chat \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "constant-v1",
    "messages": [
      {"role": "user", "content": "Explain AI in simple terms"}
    ],
    "max_tokens": 500
  }'

# Response:
{
  "request_id": "req_xyz789",
  "model": "constant-v1",
  "message": {
    "role": "assistant",
    "content": "AI is like teaching a computer..."
  },
  "finish_reason": "stop",
  "tokens_used": 85
}

Available Models

Model Family Tasks Params Status
constantsense-v1 ConstantSense Classification 413K Preview
constant-v1 Constant Chat, Completion 1M Preview
constant-mini Constant Chat, Completion 250K Preview

Philosophy

No wrappers. No third-party dependencies.

Every model in Chimera is custom-built. We don't wrap OpenAI. We don't wrap Anthropic. We build from scratch using our own frameworks and architectures.

This means:

API Documentation Request Access