Skip to main content

EmbedPrefix

A lightweight Python utility for managing embedding prefixes across different embedding models.

Many embedding models require specific prefixes (e.g. query:, passage:, search_document:) depending on the task being performed. EmbedPrefix provides a simple interface for registering those prefixes and applying them consistently.

Features

  • Simple API
  • Multiple prefix strategies per model
  • Configurable prefix position (start or end)
  • Optional separators
  • Built-in loaders for common embedding models
  • Easy to extend for custom models

Installation

Clone the repository:

pip install embedsavior

No external dependencies are required.


Quick Start

IntFloat E5

from embedprefix import EmbedPrefixLoaders

emb = EmbedPrefixLoaders.load_intfloat_e5()

query = emb.capsule(
    "What is artificial intelligence?",
    strategy="query"
)

document = emb.capsule(
    "Artificial intelligence is a branch of computer science.",
    strategy="input"
)

print(query)
# query: What is artificial intelligence?

print(document)
# passage: Artificial intelligence is a branch of computer science.

Nomic Embed

from embedprefix import EmbedPrefixLoaders

emb = EmbedPrefixLoaders.load_nomic()

query = emb.capsule(
    "What is machine learning?",
    strategy="query"
)

document = emb.capsule(
    "Machine learning is a field of AI.",
    strategy="input"
)

print(query)
# search_query: What is machine learning?

print(document)
# search_document: Machine learning is a field of AI.

Creating Your Own Configuration

from embedprefix import EmbedPrefix

emb = EmbedPrefix("my-model")

emb.add_prefix(
    prefix="[QUERY]",
    name="query",
    position="start",
    sep=" "
)

emb.add_prefix(
    prefix="[DOC]",
    name="document",
    position="start",
    sep=" "
)

text = emb.capsule(
    "Hello world",
    strategy="query"
)

print(text)
# [QUERY] Hello world

API

EmbedPrefix

Represents a collection of prefix strategies for a single embedding model.

Constructor

EmbedPrefix(model_name: str)

add_prefix(...)

Registers a new prefix strategy.

add_prefix(
    prefix: str,
    name: str,
    position: Literal["start", "end"],
    sep: str = "",
    set_default: bool = False,
    force: bool = False
)

Parameters

Parameter Description
prefix Prefix string.
name Strategy name.
position Prefix placement (start or end).
sep Separator between prefix and text.
set_default Marks this strategy as the default.
force Overwrites an existing strategy with the same name.

capsule(...)

Applies a prefix strategy to a text.

capsule(
    text: str,
    strategy: str | None = None
)

Returns a prefixed string.


list_strategies()

Returns all registered strategy names.

strategies = emb.list_strategies()

Example

['query', 'input']

Built-in Loaders

IntFloat E5

EmbedPrefixLoaders.load_intfloat_e5()

Registered strategies

Strategy Prefix
query query:
input passage:

Nomic Embed

EmbedPrefixLoaders.load_nomic()

Registered strategies

Strategy Prefix
query search_query:
input search_document:
cluster clustering:
clf classification:

Supported Models

Currently included:

  • IntFloat E5
  • Nomic Embed

Additional loaders can easily be implemented for other embedding models.


License

This project is released under the Apache License.

Metadata

Release files for embedsavior 1.0.0

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

Source distribution (sdist)

Source distribution for embedsavior 1.0.0
File Size Uploaded Provenance Actions
embedsavior-1.0.0.tar.gz 4.9 kB Without provenance Details for embedsavior-1.0.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for embedsavior 1.0.0
File Interpreter ABI Platform Actions
embedsavior-1.0.0-py3-none-any.whl Python 3 none any Details for embedsavior-1.0.0-py3-none-any.whl

Total release size: 10.1 kB

Release files / embedsavior-1.0.0.tar.gz

Download URL embedsavior-1.0.0.tar.gz
Size 4.9 kB
Tags Source
SHA-256 checksum
How to use checksums
c57cd4b2a6ff47cbce70f576d795258ec394bec49065fb88a454c6bc9ad48ab8
BLAKE2b-256 checksum
How to use checksums
e345810138785b4ad133bc3301e7445bb450fe92cd08b40b5f31aa831cab303d
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.13.9

Release files / embedsavior-1.0.0-py3-none-any.whl

Download URL embedsavior-1.0.0-py3-none-any.whl
Size 5.3 kB
Tags Python 3
SHA-256 checksum
How to use checksums
aadae0efe3502d9406ace73d51cb651ffcba1a8dc7f6d9bab0365b705127fc9c
BLAKE2b-256 checksum
How to use checksums
1b2614508ea883f77c21b661233bf67e2cee115798cd22275473cf731b91ec15
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.13.9

Security for embedsavior 1.0.0

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.

Release history Release notifications | RSS feed

This release

1.0.0 This release

2 release files No provenance

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