File size: 2,111 Bytes
942050b | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 | """Node: plan_query — emit a structured plan before SQL generation.
Two-stage decomposition (DIN-SQL / MAC-SQL style). The plan node produces a
JSON skeleton (tables/joins/filters/group/agg/projection/expected_row_count)
that the downstream `generate_sql` node sees as additional grounding context.
Empirically, forcing the model to commit to row-shape and projection BEFORE
writing SQL fixes a large fraction of "row_count_off" and "projection_diff"
failures observed in the BIRD baseline taxonomy (see scripts/error_taxonomy.py).
"""
from __future__ import annotations
from collections.abc import Callable
from nl_sql.agent.nodes._support import (
render_fewshot_block,
render_schema_block,
)
from nl_sql.agent.prompts import load_prompt
from nl_sql.agent.state import PipelineState
from nl_sql.llm.providers.base import GenerateRequest, LLMProvider
def make_plan_node(
provider: LLMProvider,
*,
max_tokens: int = 600,
temperature: float = 0.0,
sort_schema_block: bool = False,
) -> Callable[[PipelineState], PipelineState]:
def node(state: PipelineState) -> PipelineState:
question = state.get("question", "")
dialect = state.get("dialect", "sqlite")
context = state.get("context")
prompt = load_prompt(
"plan",
dialect=dialect,
schema_block=render_schema_block(context, sort_alphabetically=sort_schema_block),
fewshot_block=render_fewshot_block(context),
question=question,
)
response = provider.generate(
GenerateRequest(prompt=prompt, max_tokens=max_tokens, temperature=temperature)
)
plan_text = (response.text or "").strip()
trace = list(state.get("trace") or [])
trace.append(
{
"node": "plan_query",
"model": response.model,
"input_tokens": response.input_tokens,
"output_tokens": response.output_tokens,
}
)
return {
"plan": plan_text,
"trace": trace,
}
return node
|