# ssrajadh/sentrysearch

Semantic search over videos using Gemini Embedding 2 or Qwen3-VL.

Repository: https://github.com/ssrajadh/sentrysearch
Canonical: https://ross.abutalabs.com/products/sentrysearch
Homepage: https://youtu.be/QkYUSlu_G9g
Language: Python
License: Apache-2.0
License Family: permissive
Topics: chromadb, dashcam, gemini, gemini-embedding-2, semantic-search, video, mp4, tesla, qwen3-vl, search, search-engine
Last push: 2026-08-22T23:50:58+00:00

## Health v2 (maintenance only)
Score: 65/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 52, longevity 12
- inputs: {"age_days": 169, "days_push": 11, "days_rel": 108, "gap_med": null, "n_releases_24m": 1}
- flags: prerelease_only, young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 4433, forks 425 (observed 2026-08-28T04:08:49.216405+00:00)

## What it is
SentrySearch is a Python CLI tool that provides semantic search over video footage by embedding video chunks with Gemini Embedding, Alibaba DashScope Qwen, or a local Qwen3-VL model, and storing vectors in a local ChromaDB database. Text or image queries return the best-matching segment automatically trimmed from the original video as a clip.

## Use cases
- find a specific event in hours of dashcam footage by typing a description
- search Tesla sentry mode videos for a moment like 'car door opened'
- search video by uploading a reference image instead of text
- extract a trimmed clip of a matched event from an mp4
- index large video archives locally with a local vision-language model
- run the whole pipeline offline without any API key

## When to choose
- you need natural-language search across large collections of video files
- you want a local-first index (ChromaDB) with an optional no-API-key local model
- you work with dashcam or Tesla footage and want clip extraction plus companion tools for merging and redaction

## When to avoid
- you need real-time or live-stream video search
- you want a polished GUI application rather than a CLI workflow
- your videos lack visual events the embedding models can represent (e.g., mostly audio-driven content)

## Facets
- artifact type: cli-tool
- maturity: active
- function: search-engine, vector-database, video-processing, machine-learning, cli
- domain: machine-learning, developer-tools
- platform: windows, python, cli
- tags: semantic-search, video-embeddings, chromadb, gemini-embedding, qwen3-vl, dashcam, tesla, clip-trimming, multimodal-search, video, search, macos, linux

## Member repositories
- ssrajadh/sentrysearch (main) score 65

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:49.216405+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 2026-08-29T18:20:53.543088+00:00, confidence not recorded.
  - readme: https://github.com/ssrajadh/sentrysearch (fetched 2026-08-28T04:08:49.216405+00:00, sha 79f6656716d5)
  - homepage: https://youtu.be/QkYUSlu_G9g (fetched 2026-08-29T09:08:13.850208+00:00, sha 6d539387bcf1)
- Data as of 2026-08-30T08:39:29.467469+00:00.