Datasets:

ZHIYII
/

AgentBrew_Postgres

messages listlengths 3 3	weight float64 0 2	task_id stringlengths 36 36	step_id int64 0 40
[ { "content": "You are an agent for PostgreSQL database operations.", "role": "system" }, { "content": "You are a ReAct (Reasoning and Acting) agent.\nYou are an agent for PostgreSQL database operations.\n\nYou are able to access to these tools:\n\n### Tools Description ###\n\nServer: postgres-pro\nT...	0.024474	93ecca94-f406-4894-bc96-dfd944a4fb15	0
[ { "content": "You are an agent for PostgreSQL database operations.", "role": "system" }, { "content": "You are a ReAct (Reasoning and Acting) agent.\nYou are an agent for PostgreSQL database operations.\n\nYou are able to access to these tools:\n\n### Tools Description ###\n\nServer: postgres-pro\nT...	0.054821	93ecca94-f406-4894-bc96-dfd944a4fb15	2
[ { "content": "You are an agent for PostgreSQL database operations.", "role": "system" }, { "content": "You are a ReAct (Reasoning and Acting) agent.\nYou are an agent for PostgreSQL database operations.\n\nYou are able to access to these tools:\n\n### Tools Description ###\n\nServer: postgres-pro\nT...	0.525698	93ecca94-f406-4894-bc96-dfd944a4fb15	4
[ { "content": "You are an agent for PostgreSQL database operations.", "role": "system" }, { "content": "You are a ReAct (Reasoning and Acting) agent.\nYou are an agent for PostgreSQL database operations.\n\nYou are able to access to these tools:\n\n### Tools Description ###\n\nServer: postgres-pro\nT...	0.263338	93ecca94-f406-4894-bc96-dfd944a4fb15	6
[ { "content": "You are an agent for PostgreSQL database operations.", "role": "system" }, { "content": "You are a ReAct (Reasoning and Acting) agent.\nYou are an agent for PostgreSQL database operations.\n\nYou are able to access to these tools:\n\n### Tools Description ###\n\nServer: postgres-pro\nT...	2	93ecca94-f406-4894-bc96-dfd944a4fb15	10
[ { "content": "You are an agent for PostgreSQL database operations.", "role": "system" }, { "content": "You are a ReAct (Reasoning and Acting) agent.\nYou are an agent for PostgreSQL database operations.\n\nYou are able to access to these tools:\n\n### Tools Description ###\n\nServer: postgres-pro\nT...	0.402563	8cbdd4c3-35af-4cd8-9bc4-a36e536cc4b9	4
[ { "content": "You are an agent for PostgreSQL database operations.", "role": "system" }, { "content": "You are a ReAct (Reasoning and Acting) agent.\nYou are an agent for PostgreSQL database operations.\n\nYou are able to access to these tools:\n\n### Tools Description ###\n\nServer: postgres-pro\nT...	0.303257	8cbdd4c3-35af-4cd8-9bc4-a36e536cc4b9	6
[ { "content": "You are an agent for PostgreSQL database operations.", "role": "system" }, { "content": "You are a ReAct (Reasoning and Acting) agent.\nYou are an agent for PostgreSQL database operations.\n\nYou are able to access to these tools:\n\n### Tools Description ###\n\nServer: postgres-pro\nT...	0.567005	8cbdd4c3-35af-4cd8-9bc4-a36e536cc4b9	10
[ { "content": "You are an agent for PostgreSQL database operations.", "role": "system" }, { "content": "You are a ReAct (Reasoning and Acting) agent.\nYou are an agent for PostgreSQL database operations.\n\nYou are able to access to these tools:\n\n### Tools Description ###\n\nServer: postgres-pro\nT...	2	8cbdd4c3-35af-4cd8-9bc4-a36e536cc4b9	12
[ { "content": "You are an agent for PostgreSQL database operations.", "role": "system" }, { "content": "You are a ReAct (Reasoning and Acting) agent.\nYou are an agent for PostgreSQL database operations.\n\nYou are able to access to these tools:\n\n### Tools Description ###\n\nServer: postgres-pro\nT...	0.180459	8cbdd4c3-35af-4cd8-9bc4-a36e536cc4b9	14

End of preview. Expand in Data Studio

AgentBrew PostgreSQL

AgentBrew PostgreSQL is a public weighted action-level SFT dataset for training PostgreSQL tool-use agents. It is converted from ZHIYII/Postgres_Entropy_Action_SFT_swift_trace for the AgentBrew weighted-SFT training pipeline.

Processing

Parsed messages from a serialized JSON string into a native list of chat messages.
Kept only rows with a finite weight > 0.
Removed is_final_answer, is_error_step, and raw_advantage.
Retained task_id and step_id for provenance and analysis.
Preserved the source dataset's positive training weights without further normalization.

Schema

Field	Type	Description
`messages`	list of objects	`system`, `user`, and `assistant` chat messages.
`weight`	float	Positive per-sample weighted-SFT multiplier.
`task_id`	string	Source task identifier.
`step_id`	integer	Action index within the source task.

Statistics

Training rows: 25,443
Unique tasks: 2,140
Weight range: 0.0006018657839 to 2
Mean weight: 0.7044119361
Role pattern: system -> user -> assistant

Loading

from datasets import load_dataset

dataset = load_dataset("ZHIYII/AgentBrew_Postgres", split="train")

The training objective averages causal-language-model loss over valid assistant tokens for each sample, multiplies it by weight, and then averages the weighted sample losses across the batch.

Downloads last month: 44