Base URL: https://api.chipipay.com/v1
When to use each
| /ai/chat | /ai/chat/stream |
|---|
| How it works | Send a prompt, get the full response in one JSON object | Send a prompt, get tokens one-by-one via Server-Sent Events |
| Best for | Backend agents, batch processing, serverless functions, any case where you process the full response at once | Chatbots, real-time UIs, any interface where users watch the response appear |
| Latency feel | User waits until the entire response is generated (can be 1-5 seconds) | First token arrives in ~200ms, text streams as it’s generated |
| Cost | Per token | Same per-token cost, debited when stream completes |
| Complexity | Simple — one HTTP request, one JSON response | Requires SSE parsing (see examples below) |
What you can build
- Customer support bot —
/chat/stream for real-time replies, /chat for classification/routing behind the scenes
- Code assistant —
/chat/stream so developers see code being written line by line
- Transaction summarizer —
/chat to batch-summarize wallet activity on your backend
- AI-powered search —
/chat to rerank results or generate answers from your data
- DeFi copilot UI —
/chat/stream for the conversational layer, pair with /think for portfolio decisions
- Telegram/Discord bot —
/chat for simple Q&A, /chat/stream if the platform supports progressive message edits
- Content generation —
/chat to generate descriptions, emails, or marketing copy
POST /ai/chat
General-purpose AI. Any prompt, any model. Powers chatbots, code assistants, finance apps.
Cost: Per token (varies by model — see Models & Pricing)
curl -X POST https://api.chipipay.com/v1/ai/chat \
-H "Authorization: Bearer sk_prod_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "haiku",
"messages": [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "What is USDC?"}
],
"max_tokens": 256
}'
Request body
| Field | Type | Default | Description |
|---|
model | string | "haiku" | Any model from Models & Pricing. |
messages | array | Required | Array of objects with role ("system", "user", or "assistant") and content (string). |
max_tokens | number | 4096 | Max output tokens. Capped at the model’s limit. |
Response fields
| Field | Type | Description |
|---|
text | string | The model’s response text. |
model | string | Model ID used (e.g. "haiku", "gpt-4o-mini"). |
provider | string | "anthropic", "openai", or "google". |
usage | object | Token counts: inputTokens, outputTokens, totalTokens. |
cost | object | charged (USD) and currency. |
finishReason | string | "stop" (completed), "length" (hit max_tokens), or "unknown". |
latencyMs | number | Server-side processing time in milliseconds. |
POST /ai/chat/stream
Streaming version of /ai/chat. Returns Server-Sent Events (SSE) with token-by-token text deltas. Ideal for chatbots and real-time UIs.
Cost: Same as /ai/chat — per token, debited after stream completes.
Same request body as /ai/chat. Same models, same auth.
curl -N -X POST https://api.chipipay.com/v1/ai/chat/stream \
-H "Authorization: Bearer sk_prod_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "haiku", "messages": [{"role": "user", "content": "Tell me a short joke"}]}'
SSE event format
| Event | Shape | When |
|---|
| Text delta | data: {"text":"token"} | Each token as it’s generated |
| Completion | data: {"done":true,"model":"haiku","provider":"anthropic","usage":{...},"cost":{...}} | Stream finished |
| Error | data: {"error":"message"} | Provider error during stream |
The completion event includes model, provider, usage, and cost — same fields as the synchronous response. Credits are debited when this event fires.
Error handling
| HTTP Status | Meaning |
|---|
| 201 | Success (chat) |
| 200 | Success (stream — SSE follows) |
| 400 | Invalid request (missing messages, unsupported model) |
| 401 | Invalid, inactive, or expired API key |
| 402 | Insufficient credits |
| 429 | Rate limit exceeded |
| 502 | Provider error (model returned an error) |
