id stringlengths 13 35 | mode stringclasses 3
values | toolset listlengths 1 5 | workdir stringlengths 14 38 | init_md stringlengths 0 36.5k | messages listlengths 4 200 |
|---|---|---|---|---|---|
istio-335947702 | work | [
"bash",
"edit",
"read",
"write"
] | /home/mei/istio | [
{
"role": "user",
"content": [
{
"text": "The `istioctl experimental uninstall` command has been stable and should be graduated to a top-level command. Users should be able to run `istioctl uninstall` directly without the `experimental` prefix.",
"type": "text"
}
]
},
{
... | |
godot-2618898139 | work | [
"bash",
"edit",
"read",
"write"
] | /home/bot/godot | [
{
"role": "user",
"content": [
{
"text": "修复 BoxOccluder3D 图标 - 该图标在编辑器中渲染不正确。",
"type": "text"
}
]
},
{
"role": "assistant",
"content": [
{
"thinking": "用户要求修复 BoxOccluder3D 图标在编辑器中渲染不正确的问题。我需要先找到这个图标文件，了解当前的实现，然后找出问题并修复。\n\n让我先搜索一下 BoxOccluder3D 相关... | |
pytorch-2217880836 | work | [
"edit",
"glob",
"grep",
"read",
"write"
] | /home/ramirez/pytorch | # PyTorch Repository Guide
This is the PyTorch deep learning library.
## Layout
```
pytorch/
|-- torch/ Python frontend and C++ bindings
| |-- __init__.py Main package entry
| |-- autograd/ Automatic differentiation implementation
| |-- nn/ Neural netw... | [
{
"role": "user",
"content": [
{
"text": "将 NamedType 的方法实现从头文件中移出，放到对应的 cpp 文件中。python_str、qualname、qualifier 和 basename 方法目前是在头文件中内联定义的，但应该改为在头文件中声明，在 cpp 文件中定义。",
"type": "text"
}
]
},
{
"role": "assistant",
"content": [
{
"thinking": "The user wa... |
pytorch-2438347826 | work | [
"bash",
"edit",
"read",
"write"
] | /home/xu495/pytorch | [
{
"role": "user",
"content": [
{
"text": "Die Dokumentation für `Optimizer.state_dict()` beschreibt `param_groups` fälschlicherweise als dict. Sie sollte als Liste von Parametergruppen-dicts dokumentiert werden. Bitte beheben Sie diesen Dokumentationsfehler.",
"type": "text"
}
... | |
discourse-3639736933 | plan | [
"bash",
"edit",
"read",
"write"
] | /home/scanner/discourse | [{"role":"user","content":[{"text":"L'écran de démarrage ne s'affiche pas correctement sur certain(...TRUNCATED) | |
wordpress-develop-3351293088 | work | [
"bash"
] | /home/deepak575/wordpress-develop | "# WordPress Development - Agent Guide\n\nThis document provides guidance for coding agents operatin(...TRUNCATED) | [{"role":"user","content":[{"text":"Opera browsers are being incorrectly detected as Chrome browsers(...TRUNCATED) |
playwright-3901441821 | work | [
"bash",
"edit",
"read",
"write"
] | /home/adebayo/playwright | "# Playwright Development Guide\n\nThis document provides essential information for coding agents op(...TRUNCATED) | [{"role":"user","content":[{"text":"Скрипт сборки goma.sh завершается с ош(...TRUNCATED) |
git-1919003847 | plan | [
"bash"
] | /home/hassan182/git | [{"role":"user","content":[{"text":"Các bài kiểm định định dạng diff t4013 cần bao g(...TRUNCATED) | |
three.js-268235620 | work | [
"edit",
"glob",
"grep",
"read",
"write"
] | /home/rhel/three.js | [{"role":"user","content":[{"text":"O exemplo do FBX loader está desatualizado e precisa ser restau(...TRUNCATED) | |
kubernetes-4169618320 | work | [
"edit",
"glob",
"grep",
"read",
"write"
] | /home/alpine/kubernetes | "# Kubernetes Coding Agent Guide\n\n## Overview\n\nKubernetes (K8s) is the open-source container orc(...TRUNCATED) | [{"role":"user","content":[{"text":"The CPU manager's static policy should use the AlignBySocket con(...TRUNCATED) |
End of preview. Expand in Data Studio
Flash Coding
Provider: logitly.ai - Agentic data you can trust
Flash Coding is a dataset of agentic coding traces: an AI coding agent working on tasks in real source repositories.
Every trace is produced under a controlled runtime:
- Sandboxed environment, pinned repository, fixed working directory
- A restricted toolset and operating mode:
work,plan, orcplan
An automated verification protocol validates every record before it enters the dataset.
Read the full specification: specification.pdf.
Uploading these datasets to proprietary AI (e.g., asking a question that includes them) can get your account suspended: the AI vendor can misclassify the chain of thought in these records as reverse engineering or duplicating model outputs.
The dataset has two parts that share a single record schema:
| Part | Content | Training method |
|---|---|---|
| SFT | One demonstration trace per task | Supervised fine-tuning |
| DPO | A chosen and a rejected trace of the same task | Direct preference optimization |
Trace Format
A trace is a list of messages in the style of the Anthropic Messages API: each message has a role (user or assistant) and a content, a list of typed blocks. Messages strictly alternate between the two roles, starting with user and ending with assistant.
| Block | Content |
|---|---|
text |
The task in the first user message; the assistant's narration between tool calls, or the final summary. |
thinking |
The agent's private reasoning for the turn. Plain text, never encrypted or summarized. |
tool_use |
A request to run one tool from toolset; input is a JSON object. |
tool_result |
The outcome of one tool_use, matched by id; is_error marks a failed call. |
Fixed shapes:
- The first user message opens with one
textblock: the task. Every later user message carries onlytool_resultblocks, one pertool_useof the preceding assistant message. - Each assistant message holds one
thinkingblock, then any mix oftool_useandtextblocks in generation order. The last message closes the trace with atextblock: a summary written for the user.
What a trace does not contain:
- System prompt. Not recorded: the trace begins with the task itself.
- Media. The four block types above are the only kinds; there are no image or document blocks.
- Not distillation. No
signatureand noredacted_thinking: the traces are not distilled from Anthropic models. - Tool schemas. The trace carries the calls, not the definitions:
toolsetnames the tools.
Tools
| Tool | Description |
|---|---|
glob |
Match paths against a glob pattern. |
grep |
Search file contents using regex. |
read |
Read contents from a file. |
write |
Write contents to a file, creating or overwriting it entirely. |
edit |
Replace a specific string in a file with new content. |
bash |
Execute a shell command. |
Usage
- SFT: each record carries a single trace, used for supervised training / fine-tuning.
- DPO: each record carries a
chosenand arejectedtrace of the same task, used for preference training: the model learns to preferchosenoverrejected.
Characteristics
- System prompt. Not given; it can be restored from the context fields.
- Thinking. Every assistant message carries a
thinkingblock, so the traces carry the complete chain of thought. At training time, athinkingblock can be dropped when it is short, so the same traces also serve adaptive thinking. - Language. Tasks are written in multiple languages. In a positive trace, the assistant replies in the task's language; the language of
thinkingis not defined. - Error flag.
is_error=truemarks a call that did not exit 0, which can be the ordinary outcome of a call. Keeping these results is deliberate: the model learns to recover from failures over long-horizon work. - Plan mode. In
planmode, writing files is strictly forbidden, and/tmpis not writable either; the one exception is/dev/null. The mode gates use, not existence: the toolset is the same in every mode.
Evaluation
Flash Coding is training data: it carries no scores and no task-success labels, so the effect of training on it cannot be read off the data itself. Measure the effect on external benchmarks; we recommend DeepSWE and Terminal-Bench. The dataset shares no tasks with either benchmark: zero overlap, zero contamination.
Contact
Need a dataset tailored to your tasks? Email research@logitly.ai
- Downloads last month
- 10