chunkshop 0.3.2

Standalone ingest-to-pgvector: source → chunker → embedder → extractor → table. int8 by default.

chunkshop (Python)

Reference implementation of the chunkshop ingest tool. v0.2.0, alpha.

New here? Start with the end-to-end tutorial — a guided walkthrough from zero (no Postgres) to a running semantic query.

This file is the field-by-field reference: every CLI flag, every YAML field, the troubleshooting table. Use it alongside the tutorial once you know what you're doing.

For the high-level shape and mermaid diagram, see the top-level README.

Install

From source (recommended while alpha):

cd chunkshop/python
uv sync --extra dev

As a path dependency from another project:

[tool.uv.sources]
chunkshop = { path = "../chunkshop/python", editable = true }

Optional extras:

Extra	What you get
extractors	rake-nltk + nltk for the RAKE extractor.
keybert	keybert + sentence-transformers for the keybert_phrases extractor.
spacy	spacy for the spacy_entities NER extractor.
lang	langdetect for the lang_detect extractor.
nlp	Umbrella: keybert + spacy + lang in one install.
lede	Sibling extractive_summary repo as a path dep — enables summary_embed with lede.tfidf.summarize.
sumy	sumy + NLTK corpora for the sumy adapter shim (chunkshop.summarizers.sumy).
quantize	onnx for on-the-fly quantization scratch.
dev	pytest, pytest-asyncio, onnx.

Python ≥ 3.12 required.

Prerequisites

• Postgres ≥ 14 with the pgvector extension installed (CREATE EXTENSION vector; must succeed in your target DB).
• Disk space for model cache in ~/.cache/fastembed/ — ~85 MB for int8 bge-base, ~550 MB for nomic.
• An env var holding your DSN. The target config references it by name, not by value.

Quick run

export CHUNKSHOP_DSN="postgresql://postgres:postgres@localhost:5432/mydb"

# Point at the sample corpus in docs/samples/ for a real end-to-end run:
chunkshop ingest --config ../docs/samples/sample.yaml

# Or copy the template and edit it:
cp src/chunkshop/configs/example-files-to-bge.yaml my-cell.yaml
chunkshop ingest --config my-cell.yaml

Success looks like:

{
  "cell_name": "example_files",
  "docs_processed": 47,
  "chunks_written": 312,
  "wall_seconds": 18.4,
  "error": null
}

CLI

Two subcommands: ingest (one cell) and orchestrate (many cells in parallel).

chunkshop ingest

Runs one YAML end-to-end.

chunkshop ingest --config PATH [--doc-limit N] [--log PATH] [--omp-threads N]

Flag	YAML override	Purpose
-c, --config	—	Required. Path to YAML.
--doc-limit	runtime.doc_limit	Smoke-test mode; stop after N docs.
--log	runtime.log_path	Append stdout log lines to this file.
--omp-threads	runtime.omp_num_threads	Cap BLAS/OMP threads before ORT loads.

Exit code: 0 on success, 1 if the cell errored. Stdout = a JSON summary.

chunkshop orchestrate

Runs N cells in parallel as subprocesses.

chunkshop orchestrate (--config-dir DIR | --config PATH [--config PATH ...])
                      [--concurrency N]
                      [--checkpoints "60,120,300,600"]
                      [--timeout SECONDS]
                      [--smoke | --full]

Flag	Default	Purpose
-d, --config-dir	—	Run every *.yaml/*.yml in the directory.
-c, --config	—	Explicit path; repeatable. Mutually exclusive with --config-dir.
--concurrency	4	Max parallel cells (subprocess pool size).
--checkpoints	60,120,300,600	Seconds at which to print a status report.
--timeout	7200 (2h)	Overall wall limit; survivors get SIGTERM to their process group.
--smoke	off	Force doc_limit=1 + concurrency=1. Useful for "does it crash".

Stdout = checkpoint reports during the run, JSON summary at the end.

chunkshop bakeoff

Runs a chunker × embedder matrix against a corpus with hand-written gold queries, scores recall@k + MRR per combo, writes a leaderboard + a runnable recommended.yaml. Config-driven — the matrix lives in YAML, not on the command line.

chunkshop bakeoff --config PATH [--dsn DSN] [--yes] [--keep-schema]

Flag	Default	Purpose
--config	—	Path to the bakeoff YAML. Required.
--dsn	$CHUNKSHOP_DSN	Postgres DSN. Required (env var or flag).
--yes	off	Bypass the >50-cell matrix confirmation prompt.
--keep-schema	off	Keep the bakeoff schema after run — useful for debugging.

Outputs land in skill-output/bakeoff/{name}/:

• results.json — raw per-combo + per-query data.
• report.md — leaderboard sorted by MRR, per-query detail, statistical- power caveat.
• recommended.yaml — top combo pre-filled as a runnable chunkshop ingest cell.

Full walkthrough: ../docs/tutorial-bakeoff.md. Recipe card: ../docs/quickstart-bakeoff.md.

YAML reference

Every cell config has five sections plus an optional runtime. Extra keys are rejected (extra="forbid" in pydantic), so typos fail loudly.

cell_name: my_cell
source:   { ... }
chunker:  { ... }
embedder: { ... }
extractor: { ... }   # optional, defaults to {type: none}
target:   { ... }
runtime:  { ... }    # optional, sensible defaults below

source

type	Required fields	Optional fields
files	glob	id_from: path | stem | sha1 (default stem), encoding (utf-8)
json_corpus	path	documents_key (documents), id_field (id), content_field (content), title_field (title)
pg_table	dsn_env, schema, table, id_column, content_column	title_column, where
http	urls or sitemap	— (stub today)
s3	bucket	prefix (stub today)

chunker

Seven chunkers in three families. Pick one per cell.

Structural — split on headings, paragraphs, or word counts:

type	Required	Defaults
sentence_aware	—	doc_type: prose (or code), max_chars: 2000, min_chars: 200
fixed_overlap	—	window_words: 300, step_words: 150
hierarchy	—	prefix_heading: true, min_section_chars: 100, max_chars: 2000
neighbor_expand	base: (nested chunker)	window: 1

Semantic — splits on embedding-drift boundaries (no heading needed):

type	Required	Defaults
semantic	—	boundary_model: "sentence-transformers/all-MiniLM-L6-v2-int8", breakpoint_percentile: 95, min_sentences_per_chunk: 3, max_chunk_chars: 2000, sentence_splitter: "naive"

Pass boundary_model: "same" to reuse the cell's main embedder (trades speed for memory). See ../docs/tutorial-semantic.md.

Summary-layer — wrap any base chunker and change what gets embedded vs. what gets stored (summary_embed) or emit fine+coarse rows linked by group_id (hierarchical_summary):

type	Required	Defaults
summary_embed	base:, summarizer:	—
hierarchical_summary	base:, summarizer:, grouping:	grouping: {strategy: fixed_n, n: 5}

The summarizer config is a discriminated union: {mode: external, field: ...} pulls a pre-computed summary from a source document metadata field; {mode: callable, module: "lede.tfidf", function: "summarize", kwargs: {...}} imports lazily at first use; {mode: passthrough} reuses the raw chunk as the summary (baseline). See ../docs/summaries.md and ../docs/tutorial-summaries.md.

Full per-chunker guidance: ../docs/chunkers.md.

embedder

Only fastembed today.

Field	Required	Default	Notes
type	yes	—	Literal fastembed.
model_name	yes	—	e.g. Xenova/bge-base-en-v1.5-int8. See embedders.md.
dim	yes	—	Must match the model. Mismatch fails loudly at first embed.
batch_size	no	64	Per-call batch to fastembed.embed.
threads	no	None	None = auto (bad on shared boxes). Set to 4 typically.

extractor

type	Fields
none	— (default)
rake_keywords	top_k: 10, min_chars: 3 (defaults)

RAKE downloads NLTK corpora (stopwords, punkt) on first use to ~/nltk_data/.

target

Field	Required	Default	Notes
dsn_env	no	AGE_BAKEOFF_PGRG_DSN	Name of the env var holding your DSN. Override this to CHUNKSHOP_DSN in your configs.
schema	yes	—	Lowercase ident; must match ^[a-z_][a-z0-9_]*$. Created if missing.
table	yes	—	Same ident rule.
mode	no	overwrite	One of overwrite, append, create_if_missing. See ../docs/tutorial-multi-source.md.
source_tag	when mode=append	null	Ident-safe tag written to every row's source column. Required for append; optional (but recommended) for overwrite/create_if_missing.
promote_metadata	no	[]	List of {path, type} pairs lifting jsonb metadata paths into typed columns. path is lowercased + . → __ for the column name.
force_overwrite	no	false	Bypasses the "refuse to drop a table that holds rows from a foreign source_tag" safety check in overwrite mode.
overwrite	no (soft-deprecated)	false	Legacy boolean. Still honored when mode=overwrite (acts as the DROP+CREATE switch). Prefer the new mode field for new configs.
hnsw	no	true	false for tiny test tables where HNSW is slower than seq scan.

Multi-source ingest

Multiple cells can write to the same table by tagging each cell's rows with a source_tag. Cell A creates the table with mode: create_if_missing, Cell B appends with mode: append and its own tag. Queries filter or group by the source column. See ../docs/tutorial-multi-source.md for the end-to-end walkthrough.

target:
  dsn_env: CHUNKSHOP_DSN
  schema: mydata
  table: all_docs
  mode: append
  source_tag: support_tickets

runtime

Field	Default	Notes
omp_num_threads	1	Sets OMP/MKL/OPENBLAS/NUMEXPR env vars before ORT loads.
doc_limit	null	Stop after N docs. Smoke-test lever.
log_path	null	Mirror stdout heartbeats to this file. Parent dirs auto-created.
heartbeat_every	25	Log a progress line every N docs.

Environment variables

Var	When chunkshop reads it
$<target.dsn_env> (default AGE_BAKEOFF_PGRG_DSN)	At sink construction; must be a valid libpq DSN.
OMP_NUM_THREADS and friends	Set by runner before any numpy/ORT import.
HF_HOME / HF_HUB_CACHE	Respected by fastembed's downloader if you've moved the cache.

Troubleshooting

"no files matched glob: /path/**/*.md"

Your source.glob didn't match anything. Test it in a shell first:

ls /path/**/*.md | head

Note that chunkshop uses Python's glob.glob(..., recursive=True) — ** only matches across directories when it's its own path component (/foo/**/*.md, not /foo/**.md).

"relation already exists" on second run

target.overwrite is false by default. Either flip it to true (drops + recreates) or drop the table yourself. The ON CONFLICT DO UPDATE in the writer will also happily upsert into an existing table.

"model X produced dim Y, config says dim=Z"

Your YAML's embedder.dim doesn't match the model's output. Look up the right dim in ../docs/embedders.md — bge-small=384, bge-base=768, nomic=768.

"CREATE EXTENSION IF NOT EXISTS vector" fails with permission denied

Your DB role can't create extensions. Ask a superuser to run it once per database:

CREATE EXTENSION IF NOT EXISTS vector;

Then re-run chunkshop — the sink's CREATE EXTENSION IF NOT EXISTS will be a no-op.

"table/schema must match ^[a-z_][a-z0-9_]*$"

chunkshop refuses to interpolate mixed-case or quoted identifiers — SQL injection safety via allowlist. Lowercase your schema and table.

Ingest is slow and my CPU fans are loud

Three knobs. Pick one:

• Drop embedder.batch_size from 64 to 32 — less memory pressure, slower per-doc.
• Set embedder.threads: 4 (or 2) — caps ORT's worker pool.
• If running under orchestrate, reduce --concurrency.

See the thread-tuning table in ../docs/embedders.md.

First run hangs on "downloading model"

Fastembed is pulling the ONNX from HuggingFace. Network / HF outage. Check curl -sI https://huggingface.co/ and your proxy settings. The file lands in ~/.cache/fastembed/<model-name>/.

nltk errors on first rake_keywords run

The extractor downloads stopwords, punkt, punkt_tab into ~/nltk_data/ on first use. Behind a strict firewall? Pre-download once:

import nltk
for r in ("stopwords", "punkt", "punkt_tab"):
    nltk.download(r)

Using chunkshop as a library

from chunkshop import load_config
from chunkshop.runner import run_cell

cfg = load_config("my-cell.yaml")
result = run_cell(cfg)
print(result.docs_processed, result.chunks_written, result.wall_seconds)

Or skip the YAML and build a CellConfig directly — every section is a plain pydantic model.

Tests

cd python
uv run pytest

Most tests are offline. test_embedder_fastembed.py and test_int8_registry.py download the int8 bge-base model on first run and cache it — budget ~85 MB + a few seconds the first time.