60K synthetic dialogues in six languages with function-call traces, tool schemas and realistic failure cases.
A fully synthetic, privacy-safe dataset for training and evaluating assistants that call tools. Each dialogue ships with the tool schemas that were available, the calls the assistant made, the arguments it passed, what the tool returned and whether the task was completed.
Around a fifth of the conversations contain deliberate failure modes — wrong tool chosen, malformed arguments, missing parameters, tool timeouts — labeled so you can train recovery behaviour and measure it, instead of only training on the happy path.
| Field | Type | Description |
|---|---|---|
conversation_id |
string | Unique identifier |
domain / language |
string | Scenario domain and language |
tools |
array | JSON schemas offered to the assistant |
messages |
array | Turns with role, content and tool_calls |
tool_calls |
array | Flattened calls with arguments and results |
outcome |
string | resolved, escalated or failed |
failure_mode |
string | wrong_tool, bad_arguments, timeout, none |
quality_score |
decimal | Human review score, 1–5 |
pii_free |
boolean | Always true — the data is generated |
{
"conversation_id": "CV-01188",
"domain": "ecommerce",
"language": "de",
"outcome": "resolved",
"failure_mode": "bad_arguments",
"quality_score": 4.4,
"messages": [
{
"role": "user",
"content": "Wo ist meine Bestellung 55120?"
},
{
"role": "assistant",
"tool_calls": [
{
"name": "get_order_status",
"arguments": {
"order_id": "55120"
}
}
]
},
{
"role": "tool",
"name": "get_order_status",
"content": {
"status": "in_transit",
"eta": "2026-09-10"
}
},
{
"role": "assistant",
"content": "Ihre Bestellung ist unterwegs und kommt am 10. September an."
}
]
}
Illustrative sample showing the structure of the data. The full dataset is delivered after purchase.
Daily snapshots of 180+ hosted and open models: prices, context windows, limits and benchmark scores.
12,400 questions over real enterprise documents with gold passages, hard negatives and unanswerable cases.
Half-hourly stock, delivery-time and price snapshots from 900 dark stores in 14 cities.
Tell us what you are trying to predict, automate or understand. We will reply within one business day with an approach, a timeline and a quote.