Spaces:
Running on Zero
Running on Zero
File size: 9,871 Bytes
6bcf4a2 | 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 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 | """Command line interface."""
from __future__ import annotations
import asyncio
import json
from pathlib import Path
from typing import Annotated
import typer
from rich.console import Console
from rich.table import Table
from secrag.core.config import get_settings
from secrag.core.logging import configure_logging
from secrag.corpus import CORPUS_TICKERS, CORPUS_YEARS
app = typer.Typer(
name="secrag",
help="Evaluation-driven RAG over SEC filings.",
no_args_is_help=True,
add_completion=False,
)
console = Console()
def _setup(verbose: bool = False) -> None:
settings = get_settings()
configure_logging("DEBUG" if verbose else settings.log_level, json_output=settings.log_json)
settings.ensure_dirs()
def _parse_tickers(value: str | None) -> list[str]:
if not value:
return list(CORPUS_TICKERS)
return [t.strip().upper() for t in value.split(",") if t.strip()]
@app.command()
def ingest(
tickers: Annotated[str | None, typer.Option(help="Comma separated, e.g. AAPL,MSFT")] = None,
years: Annotated[int, typer.Option(help="Filings per company")] = CORPUS_YEARS,
rebuild: Annotated[bool, typer.Option(help="Delete the index first")] = False,
skip_xbrl: Annotated[bool, typer.Option(help="Skip XBRL fact ingestion")] = False,
verbose: Annotated[bool, typer.Option("--verbose", "-v")] = False,
) -> None:
"""Fetch filings from EDGAR, chunk them, and build the index."""
_setup(verbose)
from secrag.ingest.pipeline import ingest_tickers
from secrag.ingest.xbrl import build_fact_store
from secrag.retrieval.store import VectorStore
settings = get_settings()
symbols = _parse_tickers(tickers)
console.print(f"[bold]Ingesting[/bold] {', '.join(symbols)} ({years} filings each)")
store = VectorStore(settings)
if rebuild:
store.reset()
store = VectorStore(settings)
report = asyncio.run(ingest_tickers(symbols, years=years, store=store, settings=settings))
table = Table(title="Ingestion", show_lines=False)
for column in ("Filing", "Ticker", "FY", "Chunks", "Tables", "Status"):
table.add_column(column)
for entry in report.filings:
table.add_row(
entry.filing_id,
entry.ticker,
str(entry.fiscal_year),
str(entry.chunks),
str(entry.tables),
"[green]ok[/green]" if entry.ok else f"[red]{entry.error}[/red]",
)
console.print(table)
if not skip_xbrl:
console.print("[bold]Fetching XBRL facts[/bold]")
facts = asyncio.run(build_fact_store(symbols, settings))
console.print(f" {len(facts.df)} annual facts for {', '.join(facts.tickers())}")
console.print(f"[bold green]{report.summary()}[/bold green]")
console.print(f"Corpus now holds {store.count()} chunks")
@app.command()
def facts(
tickers: Annotated[str | None, typer.Option(help="Comma separated, e.g. AAPL,MSFT")] = None,
verbose: Annotated[bool, typer.Option("--verbose", "-v")] = False,
) -> None:
"""Rebuild the XBRL fact table without re-chunking any filings.
Useful after changing the fact selection logic, since the companyfacts
responses are cached on disk and this takes seconds rather than the tens of
minutes a full re-ingest costs.
"""
_setup(verbose)
from secrag.ingest.xbrl import build_fact_store
settings = get_settings()
symbols = _parse_tickers(tickers)
store = asyncio.run(build_fact_store(symbols, settings))
table = Table(title="XBRL facts")
for column in ("Ticker", "Company", "Years", "Rows", "Latest revenue"):
table.add_column(column)
for ticker in store.tickers():
years = store.years(ticker)
latest = store.latest_year(ticker, "revenue")
revenue = store.value_of(ticker, "revenue", latest) if latest else None
rows = int((store.df["ticker"] == ticker).sum())
company = str(store.df[store.df["ticker"] == ticker].iloc[0]["company"])[:28]
table.add_row(
ticker,
company,
f"{min(years)}-{max(years)}" if years else "-",
str(rows),
(
f"FY{latest}: {revenue.value / 1e9:,.1f}B"
if revenue and revenue.value is not None
else "-"
),
)
console.print(table)
console.print(f"[bold green]{len(store.df)} annual facts stored[/bold green]")
@app.command()
def stats(verbose: Annotated[bool, typer.Option("--verbose", "-v")] = False) -> None:
"""Show what is currently indexed."""
_setup(verbose)
from secrag.engine import build_engine
engine = build_engine()
console.print_json(json.dumps(engine.stats(), indent=2, default=str))
@app.command()
def query(
question: Annotated[str, typer.Argument(help="The question to ask")],
top_k: Annotated[int, typer.Option(help="Passages to retrieve")] = 6,
companies: Annotated[str | None, typer.Option(help="Restrict to tickers")] = None,
reranker: Annotated[str, typer.Option(help="cross_encoder, ltr, or none")] = "cross_encoder",
no_cache: Annotated[bool, typer.Option("--no-cache")] = False,
verbose: Annotated[bool, typer.Option("--verbose", "-v")] = False,
) -> None:
"""Ask a question against the indexed corpus."""
_setup(verbose)
from secrag.core.types import QueryRequest
from secrag.engine import build_engine
engine = build_engine()
request = QueryRequest(
question=question,
top_k=top_k,
companies=_parse_tickers(companies) if companies else [],
reranker=reranker,
use_cache=not no_cache,
)
response = asyncio.run(engine.answer(request))
console.rule("[bold]Answer[/bold]")
console.print(response.answer.text)
if response.numeric_results:
console.rule("[bold]Verified figures[/bold]")
for result in response.numeric_results:
value = f"{result.value:,.2f}" if result.value is not None else "N/A"
console.print(f" {result.label}: [cyan]{value}[/cyan] {result.unit}")
console.print(f" [dim]{result.formula}[/dim]")
if response.answer.citations:
console.rule("[bold]Sources[/bold]")
for citation in response.answer.citations:
console.print(
f" [{citation.marker}] {citation.label} "
f"[dim](support {citation.support_score:.2f})[/dim]"
)
console.rule("[bold]Diagnostics[/bold]")
console.print(
f" intent {response.route.intent.value if response.route else 'n/a'} "
f"(confidence {response.route.confidence if response.route else 0:.2f})"
)
console.print(f" status {response.answer.status.value}")
console.print(f" groundedness {response.answer.groundedness:.3f}")
console.print(f" contexts {len(response.contexts)}")
console.print(f" cached {response.cached}")
console.print(f" latency {response.latency_ms:.0f} ms")
@app.command("train-router")
def train_router(verbose: Annotated[bool, typer.Option("--verbose", "-v")] = False) -> None:
"""Train the query intent classifier and report held-out performance."""
_setup(verbose)
from secrag.routing.router import QueryRouter
report = QueryRouter().train()
payload = report.to_dict()
# Persisted so scripts/update_readme.py can inject the measured numbers
# rather than anyone transcribing them by hand.
out = get_settings().project_root / "evals" / "reports" / "router_training.json"
out.parent.mkdir(parents=True, exist_ok=True)
out.write_text(json.dumps(payload, indent=2), encoding="utf-8")
console.print_json(json.dumps(payload, indent=2))
console.print(f"[dim]Report written to {out}[/dim]")
@app.command("train-ltr")
def train_ltr(verbose: Annotated[bool, typer.Option("--verbose", "-v")] = False) -> None:
"""Train the learning-to-rank reranker on the golden set."""
_setup(verbose)
from secrag.evaluation.ltr_training import train_ltr_model
report = train_ltr_model()
console.print_json(json.dumps(report, indent=2, default=str))
@app.command("eval")
def evaluate(
report_path: Annotated[Path | None, typer.Option(help="Where to write the JSON report")] = None,
reranker: Annotated[str, typer.Option()] = "cross_encoder",
gate: Annotated[bool, typer.Option(help="Exit non-zero if thresholds are missed")] = False,
verbose: Annotated[bool, typer.Option("--verbose", "-v")] = False,
) -> None:
"""Run the evaluation harness over the golden set."""
_setup(verbose)
from secrag.evaluation.runner import run_evaluation
result = asyncio.run(run_evaluation(reranker=reranker, report_path=report_path))
result.render(console)
if gate and not result.passed:
console.print("[bold red]Quality gate FAILED[/bold red]")
raise typer.Exit(code=1)
@app.command()
def benchmark(
report_path: Annotated[Path | None, typer.Option()] = None,
verbose: Annotated[bool, typer.Option("--verbose", "-v")] = False,
) -> None:
"""Ablate retrieval arms and rerankers, and report the comparison."""
_setup(verbose)
from secrag.evaluation.benchmark import run_benchmark
asyncio.run(run_benchmark(report_path=report_path, console=console))
@app.command()
def serve(
host: Annotated[str | None, typer.Option()] = None,
port: Annotated[int | None, typer.Option()] = None,
reload: Annotated[bool, typer.Option()] = False,
) -> None:
"""Run the API and web UI."""
_setup()
import uvicorn
settings = get_settings()
uvicorn.run(
"secrag.api.app:app",
host=host or settings.host,
port=port or settings.port,
reload=reload,
log_level=settings.log_level.lower(),
)
if __name__ == "__main__":
app()
|