Product docs system

Documentation and Quickstart Hub

One place for account onboarding, API integration, dashboard tool workflows, and future chatbot/agent rollout guidance.

Quickstart

From signup to first output in minutes.

1. Register and sign in

Create an account and continue directly into your dashboard workspace.

2. Create default key

Generate one account key, copy it once, and set it as default for chat and tool execution.

3. Launch tools

Install starter tools and move into usage pages for production prompts.

Onboarding checklist in dashboardDefault key shared across chat and toolsOAuth + email signup supported

API

Auth, endpoints, and request patterns.

Method	Path	Purpose
GET	/v1/models	List auto, qwen3, kimik2, gpt-oss-120b, and hrllm.
POST	/v1/chat/completions	Run chat completions with auto or current direct models.
POST	/api/tools/proxy/chat	Execute tool-scoped chat requests with logging.
POST	/api/playground/chat	Use dashboard playground streaming sessions.

curl -X POST "https://api.llm.kiwi/v1/chat/completions" \
  -H "Authorization: Bearer sk_kiwi_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "messages": [{"role":"user","content":"Create a blog intro for SEO."}]
  }'

const response = await fetch("https://api.llm.kiwi/v1/chat/completions", {
  method: "POST",
  headers: {
    Authorization: "Bearer sk_kiwi_...",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "auto",
    messages: [{ role: "user", content: "Draft ad copy for product launch." }],
  }),
})

Start with model: "auto" for the simplest integration path. Free API keys can also call hrllm directly at 40 requests per hour. PRO unlocks direct access to qwen3, kimik2, and gpt-oss-120b.

Dashboard API keys are one-time reveal secrets. The exact raw key value is used for the first 12-character key_prefix, the last 4-character last_four, and the lowercase hex sha256(raw_key) stored in public.api_keys.key_hash. If a key was created by a broken historical flow and returns invalid_api_key, reissue it because raw keys are never stored server-side.

Rate Limits

Per-endpoint limits, headers, and Free/Pro differences.

All rate-limited endpoints return standard headers:Retry-After,X-RateLimit-Limit,X-RateLimit-Remaining, andX-RateLimit-Reset.

Pro tier scales account/user keyed limits by 5x. IP-only public safeguards remain fixed.

Scope	Endpoint	Limiter key	Free	Pro
Public catalog	GET /v1/models	IP	192/min	192/min
Public chat	POST /v1/chat/completions	IP (pre-auth)	48/min	48/min
Public chat	POST /v1/chat/completions	API key	40/hour (auto, hrLLM)	240/min
Tool proxy	POST /api/tools/proxy/chat	Account API key	480/min	2400/min
Tool proxy	POST /api/tools/proxy/chat	Tool ID	384/min default	1920/min default
Internal chat	POST /api/internal/chat	User + IP	36/min	180/min
Playground token	POST /api/playground/token	IP	24/min	24/min
Playground token	POST /api/playground/token	User	24/min	120/min
Playground chat	POST /api/playground/chat	User	24/min	120/min
Playground sessions list	GET /api/playground/sessions	User + IP	72/min	360/min
Playground sessions create	POST /api/playground/sessions	User + IP	24/min	120/min
Playground history	GET /api/playground/sessions/{id}/messages	User + session + IP	96/min	480/min
Playground export	GET /api/playground/sessions/{id}/export	User + session + IP	16/min	80/min
Contact form	POST /api/contact	IP	8/min	8/min
Health check	GET /api/health	IP	48/min	48/min

Usage limit	Free	Pro	Notes
Request body limit	64 KB	256 KB	Applies to chat payloads for public API, internal chat, playground chat, and tool proxy.
Prompt message size	8,000 chars per message	24,000 chars per message	Applied by prompt-security checks before model execution.
Prompt total size	32,000 chars total	120,000 chars total	Combined message content cap per request.
Completion token cap	4,096 max_tokens	8,192 max_tokens	Used for max_tokens and max_completion_tokens when provided.
Playground message input	8,000 chars	24,000 chars	Single user message in playground chat.
Playground system prompt	2,000 chars	8,000 chars	Optional system prompt in playground chat.
Tool max_output_tokens setting	up to 4,096	up to 8,192	Settings UI + tool runtime cap.

Model Catalog

Best-fit model guidance and provider lineage.

The catalog now shows the current active lineup and the deprecated legacy lineup in one place. Deprecated models stay visible with their EOL date for migration planning.

Model	Tier / Status	Agent	Provider / Base Model	Best For
hrLLM hrllm	FreeRecommended	hrLLM Croatian Language Agent	api.llm.kiwi hrllm	Croatian customer support, formal business writing, public-sector communication, and education content.
Qwen3 qwen3	PRO	Qwen3 Compact Pro Agent	api.llm.kiwi Qwen3-1.4B	Fast general chat, structured drafting, lightweight copilots, and low-latency automations.
Kimik2 kimik2	PRO	Kimik2 Advanced Pro Agent	api.llm.kiwi kimik2	Complex technical tasks, advanced reasoning, multi-step analysis, and sophisticated problem solving.
GPT-OSS-120B gpt-oss-120b	PRO	GPT-OSS Large Scale Pro Agent	OSS api.llm.kiwi GPT-OSS-120B	Large-scale analysis, comprehensive reasoning, extensive technical documentation, and complex multi-domain tasks.

Open AI Models by kiwi

Tools Usage

Dashboard tools lifecycle and execution.

Open Dashboard, then install starter tools from the Starter AI Tools section.

Create one account key and set it as your default key in Profile.

Open Tools and configure each tool's settings and tone/profile options.

Use tool usage pages for prompts, outputs, and conversion-focused iteration.

Dashboard API Keys Tools

Chat Workspace

Internal assistant behavior and controls.

The chat workspace is optimized for fast prompt testing. Choose model at the top, review transcript in the center panel, and compose with Enter-to-send.

Chat activation rule: an active default key must be configured in Profile. This same default key is reused by chat and installed tools.

Open Chat Workspace

Chatbots and Agents

Roadmap-ready architecture guidance.

Near term

Hosted Chatbots

Embeddable chatbot runtime for site widgets and support flows.

Near term

Agent Runs

Multi-step background tasks with state, retries, and audit logs.

Planned

Toolchains

Composable agent actions that call your installed dashboard tools.

These docs sections are structured to expand into dedicated chatbot and agent pages as new runtime capabilities ship.

Security

Policy and hardening baseline.

Access controls

API keys are shown once, and only the first 12 characters, last 4 characters, and sha256(raw key) metadata are stored before revocation controls apply.

Policy acceptance

Registration enforces Terms, Privacy Policy, and Responsible AI consent.

Credential recovery

Forgot/reset password flows include rate limiting and strong password policy.