Prism API Examples

Prism is an AI API proxy that intelligently routes your prompts to the best model based on a quality mode you choose. Send requests in the same OpenAI format you already use — just change the base URL.

One-line integration:

- base_url="https://api.openai.com/v1"
+ base_url="https://api.prism.ssimplifi.com/v1"

Prism handles model selection, automatic failover, session memory, and cost tracking — so you don't have to.

---

What is Prism?

Prism sits between your app and AI providers (Anthropic, OpenAI, Google). You pick a quality mode and Prism picks the right model for your specific query type:

Mode	Best for	Markup
eco	High-volume, simple tasks	15%
balanced	General use	20%
sport	Complex reasoning, code	30%

Prism classifies each query (simple / code / reasoning / complex) and routes it accordingly. A simple query in Eco mode gets Gemini Flash. A complex reasoning task in Sport mode gets Claude Opus. The quality floor ensures even Eco never sends a hard task to a weak model.

Response headers tell you exactly what happened:

• X-Prism-Model — which model was used
• X-Prism-Cost — cost in USD cents
• X-Prism-Task-Type — how Prism classified your query
• X-Prism-Failover — whether a failover occurred

---

Quickstart

1. Get your API key

Sign up at prism.ssimplifi.com → your key is shown once on the signup page. It looks like: prism_sk_abc123...

2. Make your first call

curl https://api.prism.ssimplifi.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_PRISM_KEY" \
  -H "X-Prism-Mode: eco" \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [{"role": "user", "content": "Hello! What is Prism?"}]
  }'

That's it. No model name needed — Prism picks the right one.

3. Use with the OpenAI SDK

Python:

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_PRISM_KEY",
    base_url="https://api.prism.ssimplifi.com/v1",
    default_headers={"X-Prism-Mode": "balanced"}
)

response = client.chat.completions.create(
    model="prism",  # ignored — Prism selects the model
    messages=[{"role": "user", "content": "Explain recursion briefly."}]
)
print(response.choices[0].message.content)

TypeScript/Node:

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "YOUR_PRISM_KEY",
  baseURL: "https://api.prism.ssimplifi.com/v1",
  defaultHeaders: { "X-Prism-Mode": "balanced" },
});

const response = await client.chat.completions.create({
  model: "prism",
  messages: [{ role: "user", content: "Explain recursion briefly." }],
});
console.log(response.choices[0].message.content);

---

Examples

File	Description
python/basic_chat.py	Basic chat completion in Eco mode
python/streaming.py	Streaming response in Balanced mode
python/session_memory.py	Multi-turn conversation with session memory
python/error_handling.py	Handling errors and failover detection
python/cost_comparison.py	Compare costs across Eco / Balanced / Sport
typescript/basic_chat.ts	Basic chat completion in Eco mode
typescript/streaming.ts	Streaming response in Balanced mode
typescript/session_memory.ts	Multi-turn conversation with session memory
typescript/error_handling.ts	Handling errors and failover detection
typescript/cost_comparison.ts	Compare costs across Eco / Balanced / Sport
curl/examples.sh	All examples as curl commands

---

Free Tier

The free tier lets you try Prism without a credit card:

• 50K input / 10K output tokens per day
• Eco mode only
• No streaming
• 5 messages per session

Upgrade by topping up your balance at prism.ssimplifi.com.

---

Links

• Dashboard: https://prism.ssimplifi.com
• API Docs: https://prism.ssimplifi.com/docs
• Pricing: Pay-as-you-go, prepaid balance in USD

---

Error Reference

All errors follow a consistent format:

{
  "error": {
    "type": "insufficient_balance",
    "message": "Your balance is too low to complete this request."
  }
}

Error type	Meaning
invalid_api_key	Key is missing, malformed, or revoked
missing_mode_header	X-Prism-Mode header not included
insufficient_balance	Balance too low (top up in dashboard)
free_tier_limit	Daily token limit reached
rate_limited	Too many requests
provider_error	Upstream provider failed (failover attempted)
invalid_request	Malformed request body