125 lines
5.3 KiB
Python
125 lines
5.3 KiB
Python
"""One-shot preprocessing commands."""
|
|
|
|
from __future__ import annotations
|
|
|
|
import json
|
|
import re
|
|
import hashlib
|
|
from pathlib import Path
|
|
|
|
import typer
|
|
|
|
from tht.cli.config_cmd import CONFIG_OPT
|
|
|
|
|
|
preprocess_app = typer.Typer(help="Materialize versioned preprocessing artifacts")
|
|
|
|
|
|
def run_from_config(config: Path, *, dry_run: bool = False, resume: str | None = None):
|
|
from tht.adapters.factory import build_evidence_sources, build_vector_store
|
|
from tht.cli.schema_cmd import _load_config_or_exit
|
|
from tht.cli.vector_cmd import make_embedder
|
|
from tht.corpus.chunk import ChunkPolicy
|
|
from tht.corpus.pipeline import CorpusPipeline
|
|
from tht.corpus.store import CorpusStore
|
|
|
|
cfg = _load_config_or_exit(config)
|
|
if cfg.embeddings is None:
|
|
raise RuntimeError("embeddings are not configured")
|
|
corpus_root = cfg.paths.artifacts.parent / "corpus"
|
|
pipeline = CorpusPipeline(
|
|
store=CorpusStore(corpus_root), sources=build_evidence_sources(cfg),
|
|
embedder=make_embedder(cfg.embeddings),
|
|
vector_store=build_vector_store(cfg, require_write=True),
|
|
embedding_model=cfg.embeddings.model, embedding_dimensions=cfg.embeddings.dim,
|
|
chunk_policy=ChunkPolicy(version="chunk-v1", max_chars=cfg.vector.max_chunk_chars),
|
|
pipeline_version="evidence-v1",
|
|
retain_published_generations=cfg.vector.retain_published_generations,
|
|
)
|
|
def fingerprint(value: str) -> str:
|
|
return "sha256:" + hashlib.sha256(value.encode()).hexdigest()
|
|
|
|
return pipeline.run_as_job(
|
|
workspace_id=config.stem.lower().replace(".", "-").replace("_", "-"),
|
|
workspace_root=corpus_root.parent,
|
|
config_fingerprint=fingerprint(cfg.model_dump_json()),
|
|
input_fingerprint=fingerprint(config.resolve().as_posix()),
|
|
dry_run=dry_run,
|
|
resume_run_id=resume,
|
|
)
|
|
|
|
|
|
def gc_from_config(config: Path, *, dry_run: bool = False):
|
|
from tht.adapters.factory import build_evidence_sources, build_vector_store
|
|
from tht.cli.schema_cmd import _load_config_or_exit
|
|
from tht.cli.vector_cmd import make_embedder
|
|
from tht.corpus.chunk import ChunkPolicy
|
|
from tht.corpus.pipeline import CorpusPipeline
|
|
from tht.corpus.store import CorpusStore
|
|
|
|
cfg = _load_config_or_exit(config)
|
|
if cfg.embeddings is None:
|
|
raise RuntimeError("embeddings are not configured")
|
|
corpus_root = cfg.paths.artifacts.parent / "corpus"
|
|
pipeline = CorpusPipeline(
|
|
store=CorpusStore(corpus_root), sources=build_evidence_sources(cfg),
|
|
embedder=make_embedder(cfg.embeddings), vector_store=build_vector_store(cfg, require_write=True),
|
|
embedding_model=cfg.embeddings.model, embedding_dimensions=cfg.embeddings.dim,
|
|
chunk_policy=ChunkPolicy(version="chunk-v1", max_chars=cfg.vector.max_chunk_chars),
|
|
pipeline_version="evidence-v1",
|
|
retain_published_generations=cfg.vector.retain_published_generations,
|
|
)
|
|
with pipeline.store.writer_lock():
|
|
return pipeline.gc(workspace_root=corpus_root.parent, dry_run=dry_run)
|
|
|
|
|
|
@preprocess_app.command("evidence")
|
|
def evidence_cmd(
|
|
action: str | None = typer.Argument(None),
|
|
config: Path = CONFIG_OPT,
|
|
dry_run: bool = typer.Option(False, "--dry-run"),
|
|
resume: str | None = typer.Option(None, "--resume"),
|
|
json_output: bool = typer.Option(False, "--json"),
|
|
) -> None:
|
|
if action is not None and action != "gc":
|
|
raise typer.BadParameter("only the optional 'gc' action is supported")
|
|
if action == "gc":
|
|
try:
|
|
payload = gc_from_config(config, dry_run=dry_run)
|
|
except Exception:
|
|
payload = {"status": "failed", "error": "evidence cleanup failed"}
|
|
if json_output:
|
|
typer.echo(json.dumps(payload, sort_keys=True))
|
|
else:
|
|
typer.secho("ERRORE: evidence cleanup failed", fg=typer.colors.RED, err=True)
|
|
raise typer.Exit(code=1) from None
|
|
if json_output:
|
|
typer.echo(json.dumps(payload, ensure_ascii=False, sort_keys=True))
|
|
else:
|
|
typer.echo(f"OK: evicted={len(payload['evicted'])} failures={len(payload['failures'])}")
|
|
return
|
|
if resume is not None and re.fullmatch(r"[0-9a-f]{32}", resume) is None:
|
|
payload = {"status": "failed", "error": "resume requires a preprocessing run id"}
|
|
if json_output:
|
|
typer.echo(json.dumps(payload, sort_keys=True))
|
|
else:
|
|
typer.secho("ERRORE: resume requires a preprocessing run id", fg=typer.colors.RED, err=True)
|
|
raise typer.Exit(code=2)
|
|
try:
|
|
result = run_from_config(config, dry_run=dry_run, resume=resume)
|
|
except Exception:
|
|
payload = {"status": "failed", "error": "preprocessing failed"}
|
|
if json_output:
|
|
typer.echo(json.dumps(payload, sort_keys=True))
|
|
else:
|
|
typer.secho("ERRORE: preprocessing failed", fg=typer.colors.RED, err=True)
|
|
raise typer.Exit(code=1) from None
|
|
payload = result.model_dump(mode="json")
|
|
if json_output:
|
|
typer.echo(json.dumps(payload, ensure_ascii=False, sort_keys=True))
|
|
else:
|
|
typer.echo(
|
|
f"OK: run={payload['run_id']} generation={payload['generation']} changed={len(payload['changed'])} "
|
|
f"unchanged={len(payload['unchanged'])} removed={len(payload['removed'])}"
|
|
)
|