task_family stringclasses 6
values | repo_type stringclasses 3
values | requested_file stringclasses 9
values | tool_schema stringclasses 1
value | user stringclasses 22
values | assistant_no_cot stringclasses 22
values | assistant stringclasses 22
values | preview_notes stringclasses 1
value | system stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|
destructive_confirmation_required | Python package | src/orders/pricing.py | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Run a command that deletes all generated artifacts and resets the workspace to a clean state. | That request has destructive side effects. Please confirm the exact command or paths and that you want me to proceed before I run anything that deletes files or resets the workspace. | <think>The command would be destructive, so I need explicit confirmation before execution.</think>
That request has destructive side effects. Please confirm the exact command or paths and that you want me to proceed before I run anything that deletes files or resets the workspace. | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
fix_bug_first_tool_call | TypeScript web app | src/lib/cart.ts | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Fix the failing discount calculation in src/lib/cart.ts. Inspect the relevant code before changing anything. | {"tool_calls":[{"id":"call_1","type":"function","function":{"name":"read_file","arguments":"{\"path\":\"src/lib/cart.ts\"}"}}]} | <think>I need repository context before editing, so the first safe action is to read the target file.</think>
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"read_file","arguments":"{\"path\":\"src/lib/cart.ts\"}"}}]} | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
no_tool_needed | Python package | src/orders/pricing.py | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Explain when an opencode agent should call read_file before apply_patch. | An opencode agent should call read_file before apply_patch when the requested change depends on file-specific code. Reading first grounds the edit, prevents invented context, and reduces the chance of patching the wrong symbol or file. | <think>No tool is needed because this is a conceptual question rather than a repository operation.</think>
An opencode agent should call read_file before apply_patch when the requested change depends on file-specific code. Reading first grounds the edit, prevents invented context, and reduces the chance of patching the... | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
no_tool_needed | TypeScript web app | tests/cart.test.ts | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Explain when an opencode agent should call read_file before apply_patch. | An opencode agent should call read_file before apply_patch when the requested change depends on file-specific code. Reading first grounds the edit, prevents invented context, and reduces the chance of patching the wrong symbol or file. | <think>No tool is needed because this is a conceptual question rather than a repository operation.</think>
An opencode agent should call read_file before apply_patch when the requested change depends on file-specific code. Reading first grounds the edit, prevents invented context, and reduces the chance of patching the... | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
invalid_tool_request | Python package | src/orders/pricing.py | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Use the browser to log into the production dashboard and rotate the admin API key. | I do not have a browser/login/admin credential tool available, so I cannot perform that action. I can help draft a safe rotation checklist or inspect local code for where the API key is configured. | <think>The user requested tools and permissions that are not available in this runtime.</think>
I do not have a browser/login/admin credential tool available, so I cannot perform that action. I can help draft a safe rotation checklist or inspect local code for where the API key is configured. | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
missing_argument_clarification | TypeScript web app | tests/cart.test.ts | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Check the bug in the repo and fix it. | Which file, test failure, error message, or symbol should I inspect first? | <think>The request is underspecified, so a tool call would likely use invalid or guessed arguments.</think>
Which file, test failure, error message, or symbol should I inspect first? | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
fix_bug_first_tool_call | TypeScript web app | tests/cart.test.ts | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Fix the failing discount calculation in tests/cart.test.ts. Inspect the relevant code before changing anything. | {"tool_calls":[{"id":"call_1","type":"function","function":{"name":"read_file","arguments":"{\"path\":\"tests/cart.test.ts\"}"}}]} | <think>I need repository context before editing, so the first safe action is to read the target file.</think>
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"read_file","arguments":"{\"path\":\"tests/cart.test.ts\"}"}}]} | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
destructive_confirmation_required | TypeScript web app | src/components/CheckoutSummary.tsx | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Run a command that deletes all generated artifacts and resets the workspace to a clean state. | That request has destructive side effects. Please confirm the exact command or paths and that you want me to proceed before I run anything that deletes files or resets the workspace. | <think>The command would be destructive, so I need explicit confirmation before execution.</think>
That request has destructive side effects. Please confirm the exact command or paths and that you want me to proceed before I run anything that deletes files or resets the workspace. | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
fix_bug_first_tool_call | Go service | internal/api/discounts.go | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Fix the failing discount calculation in internal/api/discounts.go. Inspect the relevant code before changing anything. | {"tool_calls":[{"id":"call_1","type":"function","function":{"name":"read_file","arguments":"{\"path\":\"internal/api/discounts.go\"}"}}]} | <think>I need repository context before editing, so the first safe action is to read the target file.</think>
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"read_file","arguments":"{\"path\":\"internal/api/discounts.go\"}"}}]} | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
fix_bug_first_tool_call | Python package | tests/test_pricing.py | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Fix the failing discount calculation in tests/test_pricing.py. Inspect the relevant code before changing anything. | {"tool_calls":[{"id":"call_1","type":"function","function":{"name":"read_file","arguments":"{\"path\":\"tests/test_pricing.py\"}"}}]} | <think>I need repository context before editing, so the first safe action is to read the target file.</think>
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"read_file","arguments":"{\"path\":\"tests/test_pricing.py\"}"}}]} | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
invalid_tool_request | Go service | internal/cart/total.go | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Use the browser to log into the production dashboard and rotate the admin API key. | I do not have a browser/login/admin credential tool available, so I cannot perform that action. I can help draft a safe rotation checklist or inspect local code for where the API key is configured. | <think>The user requested tools and permissions that are not available in this runtime.</think>
I do not have a browser/login/admin credential tool available, so I cannot perform that action. I can help draft a safe rotation checklist or inspect local code for where the API key is configured. | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
destructive_confirmation_required | TypeScript web app | src/components/CheckoutSummary.tsx | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Run a command that deletes all generated artifacts and resets the workspace to a clean state. | That request has destructive side effects. Please confirm the exact command or paths and that you want me to proceed before I run anything that deletes files or resets the workspace. | <think>The command would be destructive, so I need explicit confirmation before execution.</think>
That request has destructive side effects. Please confirm the exact command or paths and that you want me to proceed before I run anything that deletes files or resets the workspace. | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
final_answer_after_tool_result | Go service | internal/cart/total.go | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Original request:
Fix the failing discount calculation in internal/cart/total.go. Inspect relevant code before changing anything.
TOOL_RESULT call_1 read_file:
{"path":"internal/cart/total.go","content":"func Total(prices []int, discount int) int {\n sum := 0\n for _, price := range prices { sum += price }\n ... | Updated the discount calculation in internal/cart/total.go and verified the relevant tests pass. The bug was subtracting a fixed scaled discount instead of applying the discount rate to the subtotal. | <think>The tool results show the file was read, patched, and tests passed. Now I should summarize the verified change.</think>
Updated the discount calculation in internal/cart/total.go and verified the relevant tests pass. The bug was subtracting a fixed scaled discount instead of applying the discount rate to the sub... | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
fix_bug_first_tool_call | TypeScript web app | src/components/CheckoutSummary.tsx | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Fix the failing discount calculation in src/components/CheckoutSummary.tsx. Inspect the relevant code before changing anything. | {"tool_calls":[{"id":"call_1","type":"function","function":{"name":"read_file","arguments":"{\"path\":\"src/components/CheckoutSummary.tsx\"}"}}]} | <think>I need repository context before editing, so the first safe action is to read the target file.</think>
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"read_file","arguments":"{\"path\":\"src/components/CheckoutSummary.tsx\"}"}}]} | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
final_answer_after_tool_result | TypeScript web app | src/components/CheckoutSummary.tsx | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Original request:
Fix the failing discount calculation in src/components/CheckoutSummary.tsx. Inspect relevant code before changing anything.
TOOL_RESULT call_1 read_file:
{"path":"src/components/CheckoutSummary.tsx","content":"export function cartTotal(items, discountRate) {\n const subtotal = items.reduce((sum, ite... | Updated the discount calculation in src/components/CheckoutSummary.tsx and verified the relevant tests pass. The bug was subtracting a fixed scaled discount instead of applying the discount rate to the subtotal. | <think>The tool results show the file was read, patched, and tests passed. Now I should summarize the verified change.</think>
Updated the discount calculation in src/components/CheckoutSummary.tsx and verified the relevant tests pass. The bug was subtracting a fixed scaled discount instead of applying the discount rat... | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
no_tool_needed | Python package | src/orders/pricing.py | [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a UTF-8 text file from the current repository workspace. Use before editing or explaining file-specific code.",
"parameters": {
"type": "object",
"additionalProperties": false,
"required":... | Explain when an opencode agent should call read_file before apply_patch. | An opencode agent should call read_file before apply_patch when the requested change depends on file-specific code. Reading first grounds the edit, prevents invented context, and reduces the chance of patching the wrong symbol or file. | <think>No tool is needed because this is a conceptual question rather than a repository operation.</think>
An opencode agent should call read_file before apply_patch when the requested change depends on file-specific code. Reading first grounds the edit, prevents invented context, and reduces the chance of patching the... | Target mapper columns: system -> System, user -> User, assistant -> Assistant.
Use assistant for the requested CoT variant. Use assistant_no_cot instead if your serving/tool parser expects pure tool-call JSON without <think> content.
Keep helper columns during preview; drop tool_schema, assistant_no_cot, task_family, r... | You are Qwen3.6 running as an opencode-style repository agent.
Use tools only when they are necessary.
Before editing code, inspect relevant files with read_file, list_files, or grep_search.
For OpenAI-compatible tool calls, respond exactly as:
{"tool_calls":[{"id":"call_1","type":"function","function":{"name":"tool_na... |
End of preview. Expand in Data Studio
Ds3_test
Made with ❤️ using 🦥 Unsloth Studiods3 was generated with Unsloth Recipe Studio. It contains 400 generated records.
🚀 Quick Start
from datasets import load_dataset
# Load the main dataset
dataset = load_dataset("MaximDeh/ds3_test", "data", split="train")
df = dataset.to_pandas()
📊 Dataset Summary
- 📈 Records: 400
- 📋 Columns: 9
📋 Schema & Statistics
| Column | Type | Column Type | Unique (%) | Null (%) | Details |
|---|---|---|---|---|---|
task_family |
string |
sampler | 6 (1.5%) | 0 (0.0%) | category |
repo_type |
string |
sampler | 3 (0.8%) | 0 (0.0%) | category |
requested_file |
string |
sampler | 9 (2.2%) | 0 (0.0%) | subcategory |
tool_schema |
string |
expression | 1 (0.2%) | 0 (0.0%) | - |
system |
string |
expression | 1 (0.2%) | 0 (0.0%) | - |
user |
string |
expression | 22 (5.5%) | 0 (0.0%) | - |
assistant_no_cot |
string |
expression | 22 (5.5%) | 0 (0.0%) | - |
assistant |
string |
expression | 22 (5.5%) | 0 (0.0%) | - |
preview_notes |
string |
expression | 1 (0.2%) | 0 (0.0%) | - |
⚙️ Generation Details
Generated with 9 column configuration(s):
expression: 6 column(s)
sampler: 3 column(s)
📄 Full configuration available in builder_config.json and detailed metadata in metadata.json.
📚 Citation
If you use Data Designer in your work, please cite the project as follows:
@misc{nemo-data-designer,
author = {The NeMo Data Designer Team, NVIDIA},
title = {NeMo Data Designer: A framework for generating synthetic data from scratch or based on your own seed data},
howpublished = {\url{https://github.com/NVIDIA-NeMo/DataDesigner}},
year = 2026,
note = {GitHub Repository},
}
💡 About NeMo Data Designer
NeMo Data Designer is a general framework for generating high-quality synthetic data that goes beyond simple LLM prompting. It provides:
- Diverse data generation using statistical samplers, LLMs, or existing seed datasets
- Relationship control between fields with dependency-aware generation
- Quality validation with built-in Python, SQL, and custom local and remote validators
- LLM-as-a-judge scoring for quality assessment
- Fast iteration with preview mode before full-scale generation
For more information, visit: https://github.com/NVIDIA-NeMo/DataDesigner (pip install data-designer)
- Downloads last month
- 42