:latest now resolves to the slim image (label 1.2.3-slim), so local-embeddings deployments crash-loop on startup

Open Beginner friendly
#6,719 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
2/5
Estimated time
1-3 hours
Newbie friendliness
68/100
Issue type
Bug
Clarity
Clearly specified
Activity status
Active
Tech stack
docker, python

Research direction

Start with the Docker image publishing configuration and the release workflow that assigns the latest tag; the issue also points to hindsight-docs/docs/developer/installation.md under “Docker Image Variants.” Check how the 1.2.3 full and slim images are tagged and compare that with the documented variants. Done when latest resolves to the documented full image and the slim image has its own working tag.

Written by the indexing model from the issue text.

Description

bug core
Bug Description

The 1.2.3 release published the slim build to the :latest tag. docker pull ghcr.io/vectorize-io/hindsight:latest now yields an image labelled org.opencontainers.image.version=1.2.3-slim (0.56 GB amd64) whose venv has no torch, transformers or sentence-transformers. Any service that tracks :latest and uses local embeddings or the local reranker comes back with this on the next recreate:

File "/app/api/hindsight_api/engine/embeddings.py", line 390, in initialize
    from sentence_transformers import SentenceTransformer
ModuleNotFoundError: No module named 'sentence_transformers'

ImportError: sentence-transformers is required for LocalSTEmbeddings. Install it with: pip install sentence-transformers
ERROR:    Application startup failed. Exiting.

The container exits during lifespan, never binds :8888, and restart-loops. Nothing in the logs says the tag changed variant, so it reads as a broken local config — HINDSIGHT_API_EMBEDDINGS_PROVIDER=local is correct and was working minutes earlier.

The installation docs describe latest as the full image (~9 GB) and list slim as the separate variant
(hindsight-docs/docs/developer/installation.md, "Docker Image Variants"). The slim tag itself returns 404 on GHCR.

Steps to Reproduce
  1. docker pull ghcr.io/vectorize-io/hindsight:latest
  2. docker image inspect ghcr.io/vectorize-io/hindsight:latest --format '{{index .Config.Labels "org.opencontainers.image.version"}}'
    → 1.2.3-slim
  3. Run it with local embeddings, e.g. -e HINDSIGHT_API_EMBEDDINGS_PROVIDER=local -e HINDSIGHT_API_EMBEDDINGS_LOCAL_MODEL=<a sentence-transformers model dir>
    (the default local/bge-small path fails the same way).
  4. The API exits with the ImportError above.
Expected Behavior

latest is the full image, as documented, so an existing :latest deployment keeps working across an update. If the slim build is meant to become the default, the changelog says so and the startup failure names the variant instead of the missing module.

Actual Behavior

latest is the slim build. Deployments that track :latest with local embeddings break on their next pull or
compose re-create, including unattended updates driven from a UI. The only recovery available to the user is pinning a version tag, which the docs don't currently suggest.

Version

1.2.3

LLM Provider

Other

Dominant language
Python
Stars
45.2k
Forks
5.9k
Avg merge
5h 45m
Merged PRs (30d)
355

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from vectorize-io/hindsight

All issues in vectorize-io/hindsight

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.