Skip to main content

weft-rag

Weft — a RAG engine built as a microkernel. This is the default install: the packs that index a directory and answer a question about it, and the weft command.

The distribution is weft-rag; the command is weft — the name weft belongs to an unrelated project on PyPI, so what you install and what you run are spelled differently:

uvx --from weft-rag weft --help

What is in it, and what is not

Twenty-one packs ship here and each keeps its own identity: weft plugins list prints one row per pack, and weft.toml configures one at a time — [packs.store], not [packs.weft-rag].

One distribution publishes beside this one, and it is separate for a single reason:

weft-kernel the registry, discovery, the pipeline model and the payload types. Depends on pydantic and opentelemetry-api and nothing else, and installs and imports alone — which is what proves the kernel names no capability

Six packs that used to be their own distributions now ship inside this wheel instead, each behind an extra that keeps its outside library declinable — G19, 2026-09-09:

extra pack needs
[openai] an Embedder and an LLMProvider against the OpenAI API openai, pillow, and a credential
[pdf] two PDF extraction backends pypdf, pdfplumber
[qdrant] a Qdrant store backend, beside the pgvector one shipped here qdrant-client
[otel] sets the OpenTelemetry TracerProvider the registration seam already emits spans into opentelemetry-sdk
[docling] a learned-layout PDF extraction backend docling-slim
— agent, the first-party agentic front end. Imports nothing outside this wheel, so it needs no extra none

weft-rag[all] installs every one of those libraries at once. graph — the Postgres-backed graph-traversal pack added at ledger task 11.5 — was never a separate distribution either and needs no extra.

None of these seven packs is needed to index a directory and query it. A pack neither we nor you wrote installs beside this wheel on exactly the same terms and is discovered the same way, with no edit to anything here — that is the extension model, not a courtesy.

See docs/09-release.md §1 for why this is one wheel rather than many — fourteen packs at G10 (2026-08-22), six more folded in by G19 (2026-09-09).

What installing this trusts

A pack runs with your full privileges, and installing one is trusting it. Weft discovers packs through Python entry points and calls their register() in this process, so a pack can do anything the interpreter can do. That is stated rather than mitigated, because the mitigations are not available: signature verification, sandboxing, per-pack privilege separation, and any defence against a pack that is hostile once running are all out of reach without a process boundary, which Weft does not have.

What Weft does give you is visibility and refusal, not containment:

  • weft plugins doctor names every distribution that is installed, at what version, and what each one discloses about the network, filesystem and subprocess access it uses. A disclosure is what the pack says about itself; nothing checks it.
  • [packs] allow in weft.toml is an exhaustive pin. Anything not listed is never imported — refusal before any of its code runs, which is the one control that does work without a process boundary.

A control that looked like enforcement but was not would be worse than an acknowledged gap, because people build policy on it. docs/02-extension-model.md §2 is the argument in full.

Metadata

Release files for weft-rag 2.6.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for weft-rag 2.6.0
File Size Uploaded Provenance
weft_rag-2.6.0.tar.gz 929.1 kB Without provenance Details

Built distribution (wheel)

Table of built distributions (wheels) for weft-rag 2.6.0
File Interpreter ABI Platform
weft_rag-2.6.0-py3-none-any.whl Python 3 none any Details

Total release size: 2.1 MB

Release files / weft_rag-2.6.0.tar.gz

Download URL weft_rag-2.6.0.tar.gz
Size 929.1 kB
Tags Source
SHA-256 checksum
How to use checksums
ea8fb85be4619279ba86c27e8d0bad2c560a8a2afd6d6b0212a2d1f527c54a38
BLAKE2b-256 checksum
How to use checksums
61d1aa36c118f7578ddb1d6c897b569f888d275f1a0aea4a0f685aef4afd3f7a
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.12.13 {"installer":{"name":"uv","version":"0.12.13","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

Release files / weft_rag-2.6.0-py3-none-any.whl

Download URL weft_rag-2.6.0-py3-none-any.whl
Size 1.1 MB
Tags Python 3
SHA-256 checksum
How to use checksums
cb9e64dee6ccd568f1c454957bb33ae4fc02c25cbfb43fddfa8ee03c0c5dedf9
BLAKE2b-256 checksum
How to use checksums
b922b6e9434d563b4a8d05e3a9ac8cfb23616eee0dbfcb3969e2c9508cf21620
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.12.13 {"installer":{"name":"uv","version":"0.12.13","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

Security for weft-rag 2.6.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page