Everything TILIK can do, and how it does it.

GET STARTED

Overview

TILIK is an AI research engine for any question, and a live window into Robinhood Chain (chain ID 4663). You ask in your own words; TILIK works out what kind of question it is, checks only the sources that matter, and writes one clear answer where every live fact is tied to a numbered, clickable source.

GET STARTED

Quickstart

1. 1
   Open TILIK and ask

   Type in the box on the home page and press Enter, or press Ask TILIK to open the chat.

2. 2
   Watch it work

   Chips show what TILIK understood (Market, GitHub, Research…), and a short list shows each source it checked and how long it took.

3. 3
   Check the sources

   Numbers like 1 in the answer open the source they came from. The full list is under the answer.

4. 4
   Follow up

   Ask "what about its liquidity?" or "compare it with the other one" — TILIK remembers the conversation.

GET STARTED

Asking good questions

TILIK picks its sources from what you write. A few details make answers sharper:

Try these

• chain
• market
• github
• web page
• news
• research

"Try" opens TILIK with the question filled in — nothing is sent until you press Send.

HOW IT WORKS

From question to answer

HOW IT WORKS

Sources

Eight integrations supply live data. The dot shows whether this TILIK server has each one set up right now .

"Set up" means the server can use it; a source can still be briefly unreachable — answers say so when that happens.

HOW IT WORKS

Citations & trust

ROBINHOOD CHAIN

Chain data

TILIK reads Robinhood Chain (chain ID 4663) directly over its RPC, and checks the chain ID before using any data, so a misconfigured endpoint can't feed it another chain's numbers.

ROBINHOOD CHAIN

Markets & pools

Prices, liquidity, 24-hour volume and price change come from GeckoTerminal pools on Robinhood Chain, shown as cards above the answer with the time they were fetched.

• Symbols aren't unique. Anyone can create a token called "USDG"; TILIK shows the contract address so you can check it's the real one.
• The live pulse on the home page lists the top pools by liquidity — only pools with at least $100k liquidity, so thin or wash-traded pools aren't promoted. It refreshes every minute.
• Not financial advice. Numbers are shown as facts with their source and time; analysis is labelled as analysis.

USING TILIK

Conversations

• History — every chat is listed in the sidebar (the ☰ button on phones). Rename or delete one from its ⋯ menu.
• Follow-ups — TILIK sends the recent turns of the current chat, plus the links, repos, addresses, transactions and tickers mentioned, so "it" and "that pool" resolve correctly.
• Links to a chat — each chat has its own address (/chat/…). It reopens the chat in the same browser; it isn't a share link, because chats live only in that browser.
• New chat — start fresh any time with New chat; earlier chats stay in the list.

USING TILIK

Privacy

Your chats are saved only in this browser (local storage), not on TILIK's servers. Deleting a chat — or clearing your browser's site data — removes it for good. To answer, the server receives your question and the recent turns of that chat, and the sources it fetches see only the queries needed to answer.

API

Ask: POST /chat

The web app uses the same HTTP API you can call yourself. Send JSON; get the answer as JSON, or as a live event stream. Replace http://localhost:8787 with your TILIK address. If the server's owner turned on an access key, add Authorization: Bearer <key>.

Copycurl -s http://localhost:8787/chat \
  -H "content-type: application/json" \
  -d '{"message": "What is the price of USDG on Robinhood Chain?"}'

Copyconst res = await fetch("http://localhost:8787/chat", {
  method: "POST",
  headers: { "content-type": "application/json" },
  body: JSON.stringify({ message: "What is the price of USDG on Robinhood Chain?" }),
});
const answer = await res.json();
console.log(answer.answer, answer.sources);

Request body

API

Streaming

With "stream": true the response is an event stream (text/event-stream). Events arrive in this order:

Copyconst res = await fetch("http://localhost:8787/chat", {
  method: "POST",
  headers: { "content-type": "application/json" },
  body: JSON.stringify({ message: "Latest crypto news", stream: true }),
});
const reader = res.body.getReader();
const decoder = new TextDecoder();
let buffer = "";
for (;;) {
  const { value, done } = await reader.read();
  if (done) break;
  buffer += decoder.decode(value, { stream: true });
  let cut;
  while ((cut = buffer.indexOf("\n\n")) >= 0) {
    const block = buffer.slice(0, cut);
    buffer = buffer.slice(cut + 2);
    const event = block.match(/^event: (.*)$/m)?.[1];
    const data = JSON.parse(block.match(/^data: (.*)$/m)?.[1] ?? "{}");
    if (event === "response.delta") console.log(data.text);
    if (event === "response.completed") console.log(data.response.sources);
  }
}

API

The response

The main fields of a JSON answer (and of response.completed):

Copy{
  "id": "4975f8b8-…",
  "type": "market",            // normal | research | github | onchain | market | combined
  "answer": "USDG trades at about $1.00 … [1]",
  "grounded": true,            // false: needed live data couldn't be retrieved
  "sources": [
    { "id": 1, "title": "…", "url": "https://…", "domain": "…",
      "authority": "direct", "snippet": "…", "retrievedAt": "…" }
  ],
  "citations": { "used": [1], "invalid": [] },
  "data": { "market.data": [ … ] },   // structured data for cards
  "toolRuns": [ … ],                  // each source call: status and time
  "model": { "provider": "openrouter", "model": "…", "fallbackUsed": false },
  "timings": { "totalMs": 6400, "toolsMs": 2200, "llmMs": 4100 },
  "warnings": []
}

API

Other endpoints

Copycurl -s http://localhost:8787/health
# {"status":"ok","providers":{"openrouter":true,…},"tools":{"tavily":true,…}}

API

Errors & limits

Errors are JSON: { "error": { "kind": "…", "message": "…" } } — never stack traces or internal details.

Limits on the public site

Each source and AI provider also has a daily budget, so a busy day can't run up unlimited costs. A self-hosted server can change these limits.

HELP

Troubleshooting