Talkory API Documentation

curl https://api.talkory.ai/v1/health/llm \
  -H "Authorization: Bearer tk_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"

const res = await fetch("https://api.talkory.ai/v1/health/llm", {
  headers: { Authorization: `Bearer ${process.env.TALKORY_API_KEY}` },
});
const data = await res.json();
console.log(data.status, data.providers);

import os, requests

res = requests.get(
    "https://api.talkory.ai/v1/health/llm",
    headers={"Authorization": f"Bearer {os.environ['TALKORY_API_KEY']}"},
)
print(res.json())

{
  "status": "ok",
  "checkedAt": "2026-06-27T10:00:00.000Z",
  "providers": {
    "gpt":        { "status": "ok" },
    "claude":     { "status": "ok" },
    "gemini":     { "status": "ok" },
    "grok":       { "status": "error" },
    "perplexity": { "status": "ok" }
  }
}

{
  "models": [
    { "id": "gpt",        "name": "GPT-5.5",             "provider": "OpenAI",        "vision": true  },
    { "id": "claude",     "name": "Claude Sonnet 4.6",   "provider": "Anthropic",     "vision": true  },
    { "id": "gemini",     "name": "Gemini 3.1 Pro",      "provider": "Google",        "vision": true  },
    { "id": "perplexity", "name": "Sonar Reasoning Pro", "provider": "Perplexity AI", "vision": false },
    { "id": "grok",       "name": "Grok 4.3",            "provider": "xAI",           "vision": false }
  ]
}

{ "balance": 12.4831, "currency": "USD", "plan": "payg" }

{
  "period":   { "from": "2026-06-04T00:00:00.000Z", "to": "2026-07-04T00:00:00.000Z" },
  "requests": { "total": 218, "errors": 3, "byRoute": { "/v1/query": { "requests": 42, "errors": 1 } } },
  "queries":  { "total": 42, "charged": 3.1272, "totalTokens": 913202 }
}

Same fields as POST /v1/query, just no stream option.

curl -X POST https://api.talkory.ai/v1/query/sync \
  -H "Authorization: Bearer tk_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  --max-time 600 \
  -d '{ "prompt": "Best commuter motorcycles under $5000 in 2026?" }'

const res = await fetch("https://api.talkory.ai/v1/query/sync", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.TALKORY_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({ prompt: "Best commuter motorcycles under $5000?" }),
});
const data = await res.json(); // waits until the full answer is ready
console.log(data.consensus, data.commonAnswer, data.confidence, data.charged);

import os, requests

res = requests.post(
    "https://api.talkory.ai/v1/query/sync",
    headers={"Authorization": f"Bearer {os.environ['TALKORY_API_KEY']}"},
    json={"prompt": "Best commuter motorcycles under $5000?"},
    timeout=600,  # the run can take a few minutes
)
data = res.json()
print(data["consensus"], data["confidence"], data["charged"])

{
  "id": "7f9e4c2a-1b8f-4a21-9d3c-e77aa53b6c10",
  "status": "success",
  "models": ["gpt", "claude", "gemini", "grok", "perplexity"],
  "results": [
    { "modelId": "gpt", "name": "GPT-5.5", "provider": "OpenAI", "score": 91,
      "tokens": 2841, "latency": "9.4s", "text": "## Best commuter motorcycles…" }
  ],
  "consensus": "## Consensus answer…",
  "commonAnswer": "- Honda CB300R, named by every model…",
  "confidence": 87,
  "charged": 0.0742,
  "totalTokens": 14203,
  "createdAt": "2026-07-04T09:12:33.000Z"
}

Same shape as GET /v1/queries/:id. status is success, or partial if some models failed (you are only charged for models that succeed).

There is no system-prompt field; output length and input limits come from your account settings in app.talkory.ai. Maximum output tokens apply per model.

curl -N https://api.talkory.ai/v1/query \
  -H "Authorization: Bearer tk_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Best commuter motorcycles under $5000 in 2026?",
    "models": ["gpt", "claude", "gemini", "grok", "perplexity"],
    "recursive": false,
    "consensus": true
  }'

const res = await fetch("https://api.talkory.ai/v1/query", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.TALKORY_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({ prompt: "Best commuter motorcycles under $5000?" }),
});

const reader = res.body.getReader();
const dec = new TextDecoder();
let buf = "";

while (true) {
  const { done, value } = await reader.read();
  if (done) break;
  buf += dec.decode(value, { stream: true });

  const parts = buf.split("\n\n"); // SSE events are separated by a blank line
  buf = parts.pop();                 // keep the (possibly incomplete) tail

  for (const part of parts) {
    const line = part.trim();
    if (!line.startsWith("data: ")) continue; // skips ": ping" heartbeats
    const raw = line.slice(6);
    if (raw === "[DONE]") continue;           // end-of-stream marker

    const ev = JSON.parse(raw);
    switch (ev.type) {
      case "query_created": console.log("id:", ev.id); break;      // save it
      case "model_chunk":   process.stdout.write(ev.chunk); break; // live text
      case "done":          // final payload lives on this event:
        console.log(ev.consensus, ev.commonAnswer, ev.confidence,
                    ev.charged, ev.balance);
        break;
    }
  }
}

Every event is a data: {...} JSON line with a type field. Switch on type; the done event carries the final fields (consensus, commonAnswer, confidence, charged, balance).

The response is text/event-stream; each line is data: {...} JSON with a type field, ending with data: [DONE]. Heartbeat comments (: ping) arrive every 15 seconds.

{
  "type": "done",
  "id": "7f9e4c2a-1b8f-4a21-9d3c-e77aa53b6c10",
  "status": "success",
  "models": ["gpt", "claude", "gemini", "grok", "perplexity"],
  "results": [
    { "modelId": "gpt", "name": "GPT-5.5", "provider": "OpenAI", "score": 91,
      "tokens": 2841, "latency": "9.4s", "text": "## Best commuter motorcycles…", "tags": ["Motorcycles", "Commuter", "Budget"] }
  ],
  "consensus": "## Consensus answer…",
  "commonAnswer": "- Honda CB300R, named by every model…",
  "confidence": 87,
  "charged": 0.0742,
  "balance": 12.4089,
  "totalTokens": 14203,
  "avgLatency": "11.2s",
  "createdAt": "2026-07-04T09:12:33.000Z"
}

Send "stream": false and the call returns instantly, with no SSE to parse. You get an id, then poll for the result:

# 1. Start the run (returns immediately)
curl -sS -X POST https://api.talkory.ai/v1/query \
  -H "Authorization: Bearer tk_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{ "prompt": "Best commuter bikes under $5000?", "stream": false }'
# → 202  { "id": "7f9e4c2a-1b8f-4a21-9d3c-e77aa53b6c10", "status": "running" }

# 2. Poll every 1–2s until status is no longer "running"
curl -sS https://api.talkory.ai/v1/queries/7f9e4c2a-1b8f-4a21-9d3c-e77aa53b6c10 \
  -H "Authorization: Bearer tk_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
# → status "running" → keep polling; "success" / "partial" → full result is ready

The POST answers right away, but the final answer (consensus, results, confidence) only appears once status flips from running to success/partial. Polling with GET /v1/queries/:id is free; only the query itself is charged.

{
  "id": "7f9e4c2a-1b8f-4a21-9d3c-e77aa53b6c10",
  "consensus": "## Consensus answer…",
  "commonAnswer": "- Honda CB300R…",
  "confidence": 87,
  "charged": 0.0031,
  "balance": 12.4058,
  "cached": false
}

{
  "queries": [
    { "id": "7f9e4c2a-1b8f-4a21-9d3c-e77aa53b6c10", "prompt": "Best commuter motorcycles…",
      "status": "success", "models": ["gpt","claude","gemini"], "recursive": false,
      "confidence": 87, "charged": 0.0742, "createdAt": "2026-07-04T09:12:33.000Z" }
  ],
  "nextCursor": "d2c8a611-4b0f-4e9a-9f75-3c1b6e0a2d58"
}