---
license: other
language:
- en
pretty_name: "Python Standard Library Docs — rlat knowledge model"
tags:
- retrieval
- knowledge-model
- resonance-lattice
- rlat
- embeddings
- python
- standard-library
- cpython
- programming
size_categories:
- 1M<n<10M
---

# Python Standard Library Docs — rlat knowledge model

Prebuilt [Resonance Lattice](https://github.com/tenfingerseddy/resonance-lattice-cli)
knowledge model over **[python/cpython](https://github.com/python/cpython)** (scoped to `Doc/`), built for
retrieval-augmented coding assistants and LLM context injection.

**Audience:** Python developers working with the standard library — dataclasses, typing, asyncio, pathlib, and more.

## Two variants — which do you want?

| File | Size | When to pick |
|------|------|------|
| [`python-stdlib.rlat`](./python-stdlib.rlat) | 215 MB | **Remote** — knowledge model pins to a GitHub commit SHA. Source text is fetched from `https://github.com/python/cpython` on first query and cached locally. Freshest; smaller download. Needs network on first query. |
| [`python-stdlib-bundled.rlat`](./python-stdlib-bundled.rlat) | 433 MB | **Bundled** — fully self-contained. Source text is packed inside the `.rlat` as zstd frames. Works offline forever. Larger download; no drift tracking. |

**Pick bundled** if you care about reliability, offline use, or air-gapped environments.
**Pick remote** if you want automatic freshness via `rlat freshness` / `rlat sync`, or if you plan to query from CI where cold-cache fetches are fine.

Both variants serve **identical retrieval quality** — same field, same registry, same encoder (BGE-large-en-v1.5 with random projection heads, 5 bands × 2048 dim, 8-bit quantized registry, zstd compressed).

## Usage

```bash
# Install the CLI (one-time)
pip install resonance-lattice

# Download one of the variants (pick bundled for zero-network queries):
huggingface-cli download tenfingers/python-stdlib-rlat python-stdlib-bundled.rlat --local-dir .

# Query:
rlat search python-stdlib-bundled.rlat "how do I use dataclasses with default factory"
rlat ask python-stdlib-bundled.rlat "how do I use dataclasses with default factory"

# Use as Claude / Cursor / ChatGPT context:
rlat resonate python-stdlib-bundled.rlat "how do I use dataclasses with default factory" --format context | pbcopy
```

## Sample queries

- `how do I use dataclasses with default factory`
- `difference between asyncio tasks and coroutines`
- `best way to read a large file line by line`
- `pathlib vs os.path for file operations`

## Corpus provenance

- **Upstream repo:** https://github.com/python/cpython
- **Scope:** `Doc/` subdirectory only
- **Pinned commit SHA:** `76b3923d688c0efc580658476c5f525ec8735104`
- **Files indexed:** 619
- **Chunks (semantic passages):** 12,737
- **Encoder:** `BAAI/bge-large-en-v1.5` with 5 random projection heads (2048 dim × 5 bands)
- **Registry compression:** 8-bit quantized (≈75% size reduction, ≥0.99 per-band cosine similarity vs fp32)
- **Build flags:** `--quantize-registry 8 --compression zstd --path Doc`
- **Build date:** 2026-04-22

## Licensing

### Source content

This knowledge model indexes (and in the **bundled** variant, redistributes) content from **[python/cpython](https://github.com/python/cpython)**.

- **Copyright:** © Python Software Foundation. All rights reserved.
- **License:** [Python Software Foundation License Agreement](https://docs.python.org/3/license.html)
- **Modifications:** Source files are chunked into retrieval-sized passages and indexed into a semantic field + registry. In the **bundled** variant, the source files themselves are packed unchanged as zstd frames inside the `.rlat` (lossless; re-chunking at query time produces the same bytes as the upstream file at the pinned SHA). In the **remote** variant, no source content is redistributed — queries fetch directly from the upstream repo on first access.

All retrieved passages retain their upstream license. Commercial use is permitted under the upstream license; attribution to the upstream repo is required when redistributing passages.

### Artifact structure

The knowledge-model *structure* — semantic field tensor, source registry, manifest, encoder configuration — is licensed under the [Business Source License 1.1](https://mariadb.com/bsl11/), the same license as the [Resonance Lattice project](https://github.com/tenfingerseddy/resonance-lattice-cli). Each release converts to MPL 2.0 four years after it's first published.

BSL 1.1 applies to the *structure* of the knowledge model (how it indexes and retrieves), not to the embedded content. You can use this rlat commercially for retrieval and grounded AI workflows over the indexed docs; what BSL restricts is building a competing Resonance Lattice implementation from the lattice structure itself. See the [BSL 1.1 plain-language summary](https://github.com/tenfingerseddy/resonance-lattice-cli/blob/main/docs/LICENSE_FAQ.md).

### Trademark notice

Python and the Python logo are trademarks of the Python Software Foundation. This project is not affiliated with, endorsed by, or sponsored by the Python Software Foundation.

## Keeping it fresh (remote variant only)

```bash
# Read-only drift check — one GitHub API call
rlat freshness python-stdlib.rlat

# Pull upstream diff into the local knowledge model
rlat sync python-stdlib.rlat
```

`rlat sync` fetches only the files that changed between the pinned SHA and upstream HEAD, rechunks them, and reconciles against the existing registry. Cost is proportional to the diff, not the full corpus.

## Citation

If you use this rlat in research, please cite:

```
@misc{resonance-lattice-rlats-2026,
  title  = {Resonance Lattice launch rlats},
  author = {Snyder, Kane},
  year   = {2026},
  howpublished = {HuggingFace Hub, \url{https://huggingface.co/datasets/tenfingers/python-stdlib-rlat}}
}
```