messages listlengths 5 51 | metadata dict | tools listlengths 4 4 |
|---|---|---|
[
{
"role": "system",
"content": "You are an expert coding assistant operating inside pi, a coding agent harness. You help users by reading files, executing commands, editing code, and writing new files.\n\nAvailable tools:\n- read: Read file contents\n- bash: Execute bash commands (ls, grep, find, etc.)\n- e... | {
"run_id": "run_20260704_171719_94ce49",
"task_id": "api-beginner-01",
"task_title": "preload API の基本構造",
"harness": {
"name": "pi-coding-agent",
"command": "PROVIDER={provider_name} MODEL={model} BASE_URL={base_url} node -e 'const fs=require(\"fs\");const d=\"/tmp/distill-gym-pi\"; fs.mkdirSync(d,{{re... | [
{
"type": "function",
"function": {
"name": "read",
"description": "Read the contents of a file. Supports text files and images (jpg, png, gif, webp, bmp). Images are sent as attachments. For text files, output is truncated to 2000 lines or 50KB (whichever is hit first). Use offset/limit for lar... |
[
{
"role": "system",
"content": "You are an expert coding assistant operating inside pi, a coding agent harness. You help users by reading files, executing commands, editing code, and writing new files.\n\nAvailable tools:\n- read: Read file contents\n- bash: Execute bash commands (ls, grep, find, etc.)\n- e... | {
"run_id": "run_20260704_171719_94ce49",
"task_id": "api-beginner-1",
"task_title": "レンダラープロセスがメインプロセスの機能を呼び出すためのIPC APIの仕組み",
"harness": {
"name": "pi-coding-agent",
"command": "PROVIDER={provider_name} MODEL={model} BASE_URL={base_url} node -e 'const fs=require(\"fs\");const d=\"/tmp/distill-gym-pi\"... | [
{
"type": "function",
"function": {
"name": "read",
"description": "Read the contents of a file. Supports text files and images (jpg, png, gif, webp, bmp). Images are sent as attachments. For text files, output is truncated to 2000 lines or 50KB (whichever is hit first). Use offset/limit for lar... |
[
{
"role": "system",
"content": "You are an expert coding assistant operating inside pi, a coding agent harness. You help users by reading files, executing commands, editing code, and writing new files.\n\nAvailable tools:\n- read: Read file contents\n- bash: Execute bash commands (ls, grep, find, etc.)\n- e... | {
"run_id": "run_20260704_171719_94ce49",
"task_id": "api-beginner-ipc-channels",
"task_title": "IPCチャンネル定義の一覧",
"harness": {
"name": "pi-coding-agent",
"command": "PROVIDER={provider_name} MODEL={model} BASE_URL={base_url} node -e 'const fs=require(\"fs\");const d=\"/tmp/distill-gym-pi\"; fs.mkdirSync(... | [
{
"type": "function",
"function": {
"name": "read",
"description": "Read the contents of a file. Supports text files and images (jpg, png, gif, webp, bmp). Images are sent as attachments. For text files, output is truncated to 2000 lines or 50KB (whichever is hit first). Use offset/limit for lar... |
[
{
"role": "system",
"content": "You are an expert coding assistant operating inside pi, a coding agent harness. You help users by reading files, executing commands, editing code, and writing new files.\n\nAvailable tools:\n- read: Read file contents\n- bash: Execute bash commands (ls, grep, find, etc.)\n- e... | {
"run_id": "run_20260704_171719_94ce49",
"task_id": "api-beginner-preload",
"task_title": "プリロードスクリプトがRendererに公開しているAPIの一覧",
"harness": {
"name": "pi-coding-agent",
"command": "PROVIDER={provider_name} MODEL={model} BASE_URL={base_url} node -e 'const fs=require(\"fs\");const d=\"/tmp/distill-gym-pi\";... | [
{
"type": "function",
"function": {
"name": "read",
"description": "Read the contents of a file. Supports text files and images (jpg, png, gif, webp, bmp). Images are sent as attachments. For text files, output is truncated to 2000 lines or 50KB (whichever is hit first). Use offset/limit for lar... |
[
{
"role": "system",
"content": "You are an expert coding assistant operating inside pi, a coding agent harness. You help users by reading files, executing commands, editing code, and writing new files.\n\nAvailable tools:\n- read: Read file contents\n- bash: Execute bash commands (ls, grep, find, etc.)\n- e... | {
"run_id": "run_20260704_171719_94ce49",
"task_id": "api-intermediate-01",
"task_title": "SessionManager のセッションファイルフォーマットと検証機構",
"harness": {
"name": "pi-coding-agent",
"command": "PROVIDER={provider_name} MODEL={model} BASE_URL={base_url} node -e 'const fs=require(\"fs\");const d=\"/tmp/distill-gym-pi... | [
{
"type": "function",
"function": {
"name": "read",
"description": "Read the contents of a file. Supports text files and images (jpg, png, gif, webp, bmp). Images are sent as attachments. For text files, output is truncated to 2000 lines or 50KB (whichever is hit first). Use offset/limit for lar... |
[{"role":"system","content":"You are an expert coding assistant operating inside pi, a coding agent (...TRUNCATED) | {"run_id":"run_20260704_171719_94ce49","task_id":"api-intermediate-1","task_title":"Webサーバー(...TRUNCATED) | [{"type":"function","function":{"name":"read","description":"Read the contents of a file. Supports t(...TRUNCATED) |
[{"role":"system","content":"You are an expert coding assistant operating inside pi, a coding agent (...TRUNCATED) | {"run_id":"run_20260704_171719_94ce49","task_id":"api-intermediate-ipc-channels","task_title":"IPC(...TRUNCATED) | [{"type":"function","function":{"name":"read","description":"Read the contents of a file. Supports t(...TRUNCATED) |
[{"role":"system","content":"You are an expert coding assistant operating inside pi, a coding agent (...TRUNCATED) | {"run_id":"run_20260704_171719_94ce49","task_id":"api-intermediate-preload-exposure","task_title":"(...TRUNCATED) | [{"type":"function","function":{"name":"read","description":"Read the contents of a file. Supports t(...TRUNCATED) |
[{"role":"system","content":"You are an expert coding assistant operating inside pi, a coding agent (...TRUNCATED) | {"run_id":"run_20260704_171719_94ce49","task_id":"architecture-advanced-01","task_title":"デスク(...TRUNCATED) | [{"type":"function","function":{"name":"read","description":"Read the contents of a file. Supports t(...TRUNCATED) |
[{"role":"system","content":"You are an expert coding assistant operating inside pi, a coding agent (...TRUNCATED) | {"run_id":"run_20260704_171719_94ce49","task_id":"architecture-advanced-1","task_title":"Container(...TRUNCATED) | [{"type":"function","function":{"name":"read","description":"Read the contents of a file. Supports t(...TRUNCATED) |
End of preview. Expand in Data Studio
コードリポジトリに対する質問タスクとpi-coding-agentによる応答を，自作のパイプラインで120個生成したものです．
pi-coding-agentのモデル，及び質問タスクの自動生成にDeepseek-V4-Flash(reasoning_effort=medium)を使用しました． またpiが閲覧するリポジトリは，私個人のもの３つとしています．
このデータセットは，大規模言語モデルの事後学習を含む広い用途で使用することができます． ライセンス(MIT)を確認し，その範囲内で自由にご利用ください．
備考
- データセット生成に用いたDeepseek-V4-Flashは，Deepseek APIから利用しました．Deepseek APIはデータセット作成時点で，DeepSeek Open Platform Terms of Serviceにおいて，他のモデルのトレーニング目的での利用を明確に許容しています．
- 本データセット作成には約0.6ドルかかりました．
- Downloads last month
- 38