sai1912 commited on
Commit
8ea9b64
ยท
verified ยท
1 Parent(s): c0310e8

Upload folder using huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +34 -5
README.md CHANGED
@@ -1,3 +1,22 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  <div align="center">
2
 
3
  # ๐Ÿ—„๏ธ SQL Debug Environment (OpenEnv)
@@ -76,14 +95,24 @@ sql_env/
76
 
77
  ## ๐ŸŽฏ Supported Tasks
78
 
79
- The environment supports four distinct tasks ranging from beginner SQL fixes to expert-level analytical window functions. You can initialize any task by querying `POST /reset` with the desired `task_id`.
 
 
80
 
81
  | Task ID | Difficulty | Objective |
82
  |---|---|---|
83
- | `task1_syntax_fix` | **Easy** | Fix a SQL query with a missing comma (syntax error) and a wrong table alias in the `WHERE` clause. |
84
- | `task2_join_aggregation` | **Medium** | Diagnose a `GROUP BY` query producing wrong revenue totals because an `INNER JOIN` is silently dropping NULL-keyed rows. |
85
- | `task3_etl_timezone` | **Hard** | Trace an entire 4-step Python/SQL ETL pipeline where step 2 coerces a `VARCHAR` timezone into a `DATE`, stripping the offset. Requires `TIMESTAMPTZ` fixes and an explanation string. |
86
- | `task4_expert_window` | **Expert** | Calculate a complex 3-day rolling revenue average per user. Requires advanced `OVER (PARTITION BY ... ROWS BETWEEN)` mechanics. |
 
 
 
 
 
 
 
 
87
 
88
  ---
89
 
 
1
+ ---
2
+ title: SQL Debug RL Environment
3
+ emoji: ๐Ÿ—„๏ธ
4
+ colorFrom: purple
5
+ colorTo: indigo
6
+ sdk: docker
7
+ app_port: 7860
8
+ short_description: SQL RLVE โ€” 7 tasks with live DuckDB verification
9
+ tags:
10
+ - reinforcement-learning
11
+ - sql
12
+ - duckdb
13
+ - data-engineering
14
+ - openenv
15
+ - rlve
16
+ - agent
17
+ license: apache-2.0
18
+ ---
19
+
20
  <div align="center">
21
 
22
  # ๐Ÿ—„๏ธ SQL Debug Environment (OpenEnv)
 
95
 
96
  ## ๐ŸŽฏ Supported Tasks
97
 
98
+ The environment supports **7 tasks** โ€” 4 foundational and 3 advanced RLVE challenges. Initialize any task via `POST /reset` with the `task_id`.
99
+
100
+ ### Foundational Tasks
101
 
102
  | Task ID | Difficulty | Objective |
103
  |---|---|---|
104
+ | `task_1_easy` | **Easy** | Fix a SQL query with a missing comma between column names. |
105
+ | `task_2_medium` | **Medium** | Add a missing `GROUP BY` clause to an aggregation query. |
106
+ | `task_3_hard` | **Hard** | Add `PARTITION BY` to a window function that ranks globally instead of per-department. |
107
+ | `task_4_expert` | **Expert** | Fix an invalid date literal (month 13) inside a CTE. |
108
+
109
+ ### Advanced RLVE Tasks (Live DuckDB Verifier)
110
+
111
+ | Task ID | Difficulty | Verifier Logic |
112
+ |---|---|---|
113
+ | `task_5_optimization` | **Advanced** | Rewrite a CROSS JOIN query to use `INNER JOIN`. Reward only if output matches baseline **and** `EXPLAIN` shows no `CROSS_PRODUCT`. |
114
+ | `task_6_migration` | **Advanced** | Normalize a denormalized `messy_dump` table into 3NF (`users` + `orders`). Destructive early `DROP` triggers -0.3 penalty and ends episode. |
115
+ | `task_7_chaos` | **Advanced** | Live ETL corrupts data every step (duplicate IDs + NULL emails). Apply a patch and `UNIQUE` index before the pipeline contaminates the DB again. |
116
 
117
  ---
118