# vec2text/vec2text

utilities for decoding deep representations (like sentence embeddings) back to text

Repository: https://github.com/vec2text/vec2text
Canonical: https://ross.abutalabs.com/products/vec2text
Language: Python
License: NOASSERTION
License Family: other
Last push: 2025-12-27T17:22:44+00:00

## Health v2 (maintenance only)
Score: 57/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 59, release rhythm 35, longevity 91
- inputs: {"age_days": 1285, "days_push": 249, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1136, forks 117 (observed 2026-08-28T04:03:43.672396+00:00)

## What it is
A Python library for text embedding inversion: training and running models that reconstruct text sequences from their sentence embeddings. It accompanies research papers showing that text embeddings reveal almost as much as the original text.

## Use cases
- reconstruct text from sentence embeddings
- invert OpenAI ada-002 embeddings back to text
- train a custom embedding inversion model
- study privacy risks of text embeddings
- evaluate how much information embeddings leak
- decode deep representations back to text

## When to choose
- you need to invert embeddings from supported models like text-embedding-ada-002
- you are researching embedding privacy or information leakage
- you want to train inversion models on your own embedders

## When to avoid
- you just need embeddings, not inversion
- you need production-grade text generation rather than research tooling
- your embedder has no pre-trained inversion model and you cannot train one

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, nlp, rag
- domain: machine-learning, privacy, security
- platform: python
- tags: embedding-inversion, text-embedding, privacy-research, huggingface, transformers, natural-language-processing

## Member repositories
- vec2text/vec2text (main) score 57

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:43.672396+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:36:32.602358+00:00, confidence not recorded.
  - readme: https://github.com/vec2text/vec2text (fetched 2026-08-28T04:03:43.672396+00:00, sha 2b65b5f7f14c)
- Data as of 2026-08-30T08:39:29.467469+00:00.