A skill your agent uses when a user wants to search archived VSS video that is already registered in a configured deployment — by natural-language, similarity, attribute, object-ID, or lexical tag…

Apache-2.0Auto-check passedWriting & Content

Install Vss Search Archive

skills CLI
$ npx skills add NVIDIA-AI-Blueprints/video-search-and-summarization --skill vss-search-archive -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install NVIDIA-AI-Blueprints/video-search-and-summarization vss-search-archive --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/NVIDIA-AI-Blueprints/video-search-and-summarization.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/operations/vss-search-archive .claude/skills/vss-search-archive && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
vss-search-archive
GitHub stars
1.9k
Token cost
~3.3k tokens
SKILL.md length
1,666 words
Files
7 (incl. scripts, references)
Skills in repo
22
Repo updated
First seen
Licence
Apache-2.0

At a glance

A skill your agent uses when a user wants to search archived VSS video that is already registered in a configured deployment — by natural-language, similarity, attribute, object-ID, or lexical tag…

  • A user wants to search archived VSS video that is already registered in a configured deployment — by natural-language
  • SKILL.md covers When to Use, Prerequisites, Source management handoff and Search workflow
  • Runs Shell scripts from its folder; calls docker and uv
  • Lexical tag query

What it does

Vss Search Archive is an agent skill from NVIDIA-AI-Blueprints/video-search-and-summarization. Use this skill when a user wants to search archived VSS video that is already registered in a configured deployment — by natural-language, similarity, attribute, object-ID, or lexical tag query. Not for fresh clip Q&A, live captioning, video summarization, deployment, or source ingestion/deletion.

Its SKILL.md is about 3.3k tokens, which your agent loads only when the skill is triggered. The skill folder holds 9 other files, including scripts and reference files (for example `evals/evals.json`, `evals/search.json` and `references/cli_usage.md`).

It sits in Writing & Content, covering Summarization and Deployment. The repository describes itself as: NVIDIA AI Blueprint for video search and summarization (VSS) is a GPU-accelerated reference architecture for building video analytics agents with real-time verified alerts… The licence is Apache-2.0.

When your agent uses it

  • A user wants to search archived VSS video that is already registered in a configured deployment — by natural-language
  • Lexical tag query

Example prompts

  • “/vss-search-archive”

Requirements

  • A Bash shell
  • Docker

What it can do on your machine

Read from SKILL.md and the folder at commit fdb6a7a. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Ships 1 file in scripts/ (Shell), which the agent can run.

    Shell commands in SKILL.md call:

    • docker
    • uv

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md. Its commands use docker and uv, which can reach the network depending on how they are called.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Vss Search Archive loads about 3.3k tokens when it runs, and up to ~7k if it reads all its reference files. Until then it costs about 79 tokens; SKILL.md has 1,666 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~79
When it runs · the whole SKILL.md, loaded when a task matches
~3.3k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~7k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); the scripts in this folder are not scanned.

SKILL.md

The full file from NVIDIA-AI-Blueprints/video-search-and-summarization at commit fdb6a7a, republished under its Apache-2.0 licence (© NVIDIA-AI-Blueprints). 1,666 words, ~3,344 tokens.

Download SKILL.mdSave it as .claude/skills/vss-search-archive/SKILL.md (or your agent's skills folder). This skill also uses 6 other files; get the full folder from GitHub.
name
vss-search-archive
description
Use this skill when a user wants to search archived VSS video that is already registered in a configured deployment — by natural-language, similarity, attribute, object-ID, or lexical tag query. Not for fresh clip Q&A, live captioning, video summarization, deployment, or source ingestion/deletion.
license
Apache-2.0
metadata.author
NVIDIA Video Search and Summarization team
metadata.version
3.3.0-rc0
metadata.github-url
https://github.com/NVIDIA-AI-Blueprints/video-search-and-summarization
metadata.tags
nvidia blueprint operational
metadata.vss-requires
search

Search archived VSS video

When to Use

  • Search archived VSS video that is already registered in a configured deployment, by natural-language, similarity, attribute, object-ID, or lexical tag query.

Not for:

  • A fresh visual question about a supplied local clip — vss-ask-video.
  • Long-form summarization of a recording — vss-summarize-video.
  • Deploying or changing a profile — /vss-build-vision-ai.
  • Ingesting or deleting a source — use the deployment's source-management workflow.

Answer from the configured deployment through the installed vss CLI. Do not fall back to raw REST when a CLI command fails.

Hard rule — use the vss CLI for retrieval. vss configure has already pointed the CLI at the deployment, so every search action goes through vss search run and nothing reaches the deployment any other way.

Four things follow:

  • use the vss on PATH (see Prerequisites); never docker exec, kubectl exec, a pod shell, or a hand-built /api/v1/search call;
  • a named source is resolved with vss vios list before search — never inferred from a display name, and never substituted when missing;
  • capture stdout and the exit status separately — never put the command behind if !, which hides the real exit code, and never discard usable exit-6 results;
  • a failing command is a finding: report the exit code rather than routing around it, repairing the deployment, or retrying with broadened scope.

Prerequisites

  • A running VSS search profile with vss configure already run against its origin. Refresh configuration after first ingestion, once source provisioning has established readiness; lazy raw indexes enable frame enrichment for attribute/fusion.
  • The vss CLI on PATH. The OpenClaw and Hermes harness images ship it; anywhere else, install it from the same checkout as this skill so the CLI and the skill match: uv tool install <checkout>/libs/vss/cli.

Bootstrap, exit codes, and common CLI rules live in AGENTS.md.

Source management handoff

For an explicit request to ingest or delete a source, use vss-manage-video-io-storage for vss vios add or vss vios delete, then return to archive search once the source is available. The deployment's mounted notification config owns consumer fan-out; the presence of an Agent tier does not change the VIOS registration path. Check the requested receiver as enabled, absent, or unknown under source management's policy contract. With an enabled receiver, use fan-out; with confirmed absent lifecycle support, report that limitation. Only an explicit tagging request with a confirmed absent streaming tagging receiver and the provisioning prerequisites may use manual tagging. With unknown policy, report unconfirmed fan-out, never unavailable indexing, and do not start manual tagging. If the user only asks to search a named source and it is missing, do not ingest, switch videos, or run an unrestricted search — answer with the reply Search workflow step 1 specifies for zero matches. If no supported source management workflow is available, report that blocker.

Search workflow

1. Resolve a named source. If the request names a file, camera, or sensor, resolve it with vss vios list (it reads the origin vss configure recorded, so it takes no endpoint). Accept an exact name or sensor ID, or one unambiguous normalized match; stop on zero or multiple matches. The listing is {"count", "type", "sensors": [...]}; preserve the matched entry's .sensors[].name and .sensors[].sensor_id, and never infer an identifier from the display name. Below, .name and .sensor_id mean those fields.

Stopping is not the whole answer — the reply has to hand the decision back. On zero matches, the final reply states all three of:

  • the requested name is not registered;
  • the sources that are registered, from the listing;
  • a request that the user clarify which source they meant, or explicitly ask for the missing one to be ingested.

The third is as required as the first two. Refusing to substitute is correct but incomplete: a reply that reports the mismatch and then stops leaves the user with no stated next step. Ingesting, switching to another video, or dropping the source filter and searching everything are all still forbidden (see Source management handoff). On multiple matches, name the candidates and ask which one.

2. Choose one retrieval path. Preserve the user's exact original sentence for --original-query first — critic verification must receive that wording, while retrieval may use the decomposed query, attributes, or object IDs. Then choose exactly one path:

  • object — explicit tracked object IDs;
  • tag — explicit lexical tag/keyword intent (BM25 over indexed VLM tags), not semantic free text;
  • attribute — detectable properties only, no action or relation;
  • fusion — a detectable property combined with an action or relation;
  • otherwise embed — semantic free text.

--attribute is for properties RT-CV detects on a subject (attire, PPE, color-on-person), not object identity or an object's own color — keep red forklift wholly in --query. worker in a hard hat carrying a cone has a property (hard hat) and an action (carrying a cone): run fusion. Reserve embed for genuinely attribute-free intent, and tag for keyword/tag queries that name no detectable property.

The --video-source value differs by path while the CLI's paths need different identifiers:

path--video-source takes
embedpreserved .sensor_id
attributepreserved .name
objectpreserved .name
tag.name (the CLI resolves it to the VST sensor ID; an already-id passes through)
fusionpreserved .sensor_id (the tag leg accepts IDs too)

For every path an unknown source yields an empty, narrowed result, not an error.

Before attribute or fusion, inspect vss configure show's services.elasticsearch.indices. If no entry starts with mdx-raw-, refresh once with vss configure --base-url using that same record's nonempty base_url, then inspect again. Stop on a configuration failure; if the raw family is still absent, disclose that frame enrichment is unavailable and continue the requested retrieval. Do not construct an origin or poll indexes.

3. Invoke the CLI. Use --query for embed/fusion/tag, repeatable --attribute for attribute/fusion, and repeatable --object-id for object. Use --timestamp-start / --timestamp-end for time bounds. Set --source-type video_file for an uploaded recording or rtsp for a live stream. Source type selects the fixed uploads anchor or the live family wildcard excluding that uploads anchor, independently of source identity and ingestion order. Carry the requested time bounds and --top-k, and run vss search run <path> with no endpoint, index, model, or profile flag — vss configure owns those. Build the invocation as a Bash array and capture stdout and the exit status separately; never clear a previously resolved source array or hide the exit code behind if !:

Show full SKILL.md (631 more words)Show less
bash
: "${SEARCH_PATH:?set embed|attribute|fusion|object|tag}"
: "${SOURCE_TYPE:?set video_file or rtsp}"
: "${ORIGINAL_QUERY:?set the exact pre-decomposition user question}"
TOP_K="${TOP_K:-3}"
# Set VIDEO_SOURCES from step 1 for a named source. Explicitly set it to () only
# for a request that was unrestricted from the start; never reset a resolved scope.
declare -p VIDEO_SOURCES >/dev/null 2>&1 || { echo "Set VIDEO_SOURCES before search" >&2; exit 1; }
: "${SOURCE_SCOPED:?set true for a resolved scope; false only when unrestricted}"
if [ "${SOURCE_SCOPED}" = true ] && [ "${#VIDEO_SOURCES[@]}" -eq 0 ]; then
  echo "Resolved source scope is empty; refusing an unrestricted search" >&2
  exit 1
fi
SEARCH_COMMAND=(vss search run "${SEARCH_PATH}" --source-type "${SOURCE_TYPE}" \
  --top-k "${TOP_K}" --original-query "${ORIGINAL_QUERY}" --raw)
for source in "${VIDEO_SOURCES[@]}"; do
  SEARCH_COMMAND+=(--video-source "${source}")
done
# Append only the selected path's fields and --timestamp-start/--timestamp-end.
if SEARCH_JSON=$("${SEARCH_COMMAND[@]}"); then
  STATUS=0
else
  STATUS=$?
fi

Read CLI usage only when tuning retrieval weights (--fusion-method, --w-tag, etc.); do not open it for a standard search invocation — the contract above is the whole invocation.

4. Interpret the result by exit status.

  • Exit 0 — interpret data and search_messages.
  • Exit 6 — partial (usually a persistence stage after retrieval failed; a critic failure is exit 0 with search_messages): report hits only when the payload contains data, disclosing the supplied limitation. Without data, report the supplied failure; do not claim retrieval succeeded. Do not rerun or retry an individual stage.
  • Exit 2 — read vss search run <path> --help once; correct invalid flags or values only from that help, then stop if the corrected command fails.
  • Other nonzero — report the typed failure (3 backend unreachable, 4 configuration or missing service, 5 not found) and stop.

Routes not exposed through ingress are recorded as absent by vss configure; a search path requiring one exits 4. Report the missing capability and ask the operator to expose its supported ingress and refresh configuration. Never create a port-forward or use private endpoints. After exit 5 following first ingestion, hand readiness back to source management and refresh recorded configuration once when it is ready; never ingest implicitly.

An empty data array means zero retrieved candidates — a fact about retrieval, not about the video; a threshold or embedding gap yields the same empty result as a genuine absence, so do not describe what the footage contains or argue it is not something you would expect there. If search_messages indicate degraded retrieval, include that limitation. Do not retry, broaden scope, or use raw REST.

For each hit, read its critic_result:

  • confirmed: the critic found all requested visual criteria in that clip.
  • rejected: the critic found a visual criterion was not met.
  • unverified: the critic attempted the hit but produced no usable verdict. This includes inaccessible media, a failed VLM call, and malformed or inconclusive output.
  • null: the critic did not evaluate the hit (no VLM, a critic failure, bounds it could not check, or a hit past --critic-eval-count). Report it as unverified.

5. Report each hit honestly. For each hit, report the registered/display source, bounded interval, retrieval score where present, the returned media URL if present, and the exact confirmed / rejected / unverified verdict. Retrieval score, filename, source ID, and media availability are not visual proof. The CLI attempts critic verification by default and is fail-open: a missing VLM or inaccessible media leaves a hit unverified and does not fail retrieval. Do not inspect screenshots or call another verifier during this first turn. A media URL may be empty when VST is unavailable; when present it carries the scheme, host, and port of the origin vss configure recorded. Avoid a mandated heading or raw JSON dump, and keep the reply implementation-neutral — never expose a job ID, model or service name, deployment service address, CLI flag, or a raw sensor_id; say "visual verification" and report its verdict. The one exception: when the user asks which commands to run, show the commands.

A media URL is returned data, not a service address: reproduce it whole, including the source ID in its path; the sensor_id rule is about how you name the source in prose. Report each hit's verdict, and its critic_result.criteria_met when nonempty, without reading a verdict into the criteria.

6. Offer a Verification Step only when the whole set is unverified. If and only if every displayed result in the nonempty set is unverified, offer a Verification Step by asking whether the user wants the hits checked through vss-ask-video. If any hit is confirmed or rejected, offer nothing and hand off nothing. On explicit confirmation, load search-result verification and hand off only the displayed, bounded hits, preserving the original question; do not rerun search. Never hand off a partially verified result set.

© NVIDIA-AI-Blueprints, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 6 other files (scripts, references) in skills/operations/vss-search-archive of NVIDIA-AI-Blueprints/video-search-and-summarization.

  • SKILL.md
  • evals/evals.json
  • evals/search.json
  • references/cli_usage.md
  • references/result_verification.md
  • scripts/select_brev_origin.sh
  • skill-card.md

Open the folder on GitHubat commit fdb6a7a

Compare with similar skills

Vss Search Archive next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Vss Search Archive compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Vss Search Archive this skillNVIDIA-AI-Blueprints/video-search-and-summarization1.9k—~3.3kAutomated safety check: PassApache-2.0
News Aggregator Skillcclank/news-aggregator-skill1.3k—~2.1kAutomated safety check: PassNone
Deepgram JS Text Intelligencedeepgram/deepgram-js-sdk276—~1.1kAutomated safety check: PassMIT
Key Takeawaysaipoch/medical-research-skills1.9k—~2.3kAutomated safety check: PassMIT
Vss Search ArchiveNVIDIA/skills3.6k—~4.2kAutomated safety check: PassApache-2.0
Summarize Anythingswyxio/skills176—~6.3kAutomated safety check: PassMIT

Similar skills

  • News Aggregator Skill

    cclank/news-aggregator-skill

    Comprehensive news aggregator that fetches, filters, and deeply analyzes real-time content from 44+ sources including Hacker News, Lobsters, Dev.to, GitHub, arXiv, Hugging Face Papers, AIHOT, TLDR…

    1.3k GitHub stars~2.1k tokensUpdated 4 mo ago
    Writing & ContentAuto-check passed
  • Deepgram JS Text Intelligence

    deepgram/deepgram-js-sdk

    A skill your agent uses when writing or reviewing JavaScript/TypeScript in this repo that calls Deepgram Text Intelligence / Read (/v1/read) for sentiment, summarization, topic detection, and intent…

    276 GitHub stars~1.1k tokensUpdated 2 days ago
    Writing & ContentAuto-check passed
  • Key Takeaways

    aipoch/medical-research-skills

    Extracts and summarizes key takeaways from documents, meeting notes, articles, and other text content.

    1.9k GitHub stars~2.3k tokensUpdated 24 days ago
    Writing & ContentAuto-check passed
  • Vss Search Archive

    NVIDIA/skills

    Official

    A skill your agent uses to run top-level VSS fusion search on archived video, or to ingest video files / RTSP streams for search.

    3.6k GitHub stars~4.2k tokensUpdated yesterday
    Writing & ContentAuto-check passed
  • Summarize Anything

    swyxio/skills

    Summarizes arbitrarily long text (1k-1M words) using recursive map-reduce with any LLM backend.

    176 GitHub stars~6.3k tokensUpdated 6 days ago
    Writing & ContentAuto-check passed
  • A skill your agent uses when the user wants to set up, scale, validate, or harden NVIDIA physical AI infrastructure for synthetic data generation workflows across local MicroK8s or Azure AKS…

    3.6k GitHub stars~2.8k tokensUpdated yesterday
    DevOps & CloudAuto-check: notes

More from NVIDIA-AI-Blueprints/video-search-and-summarization

All 22 skills in this repo
  • Benchmark Video Search

    NVIDIA-AI-Blueprints/video-search-and-summarization

    Measure retrieval quality and latency of a deployed VSS search profile by ingesting a labelled dataset and running the vss CLI across retrieval paths.

    1.9k GitHub stars~4.3k tokensUpdated yesterday
    Auto-check passed
  • Rtvi Vlm Perf Testing

    NVIDIA-AI-Blueprints/video-search-and-summarization

    Plan, run, and diagnose reproducible RT-VLM GPU performance canaries and benchmarks.

    1.9k GitHub stars~8.6k tokensUpdated yesterday
    Auto-check: notes
  • Vss Build Vision AI

    NVIDIA-AI-Blueprints/video-search-and-summarization

    Add agent-ready vision capabilities — dense captioning, detection, search, alerting, summarization — to an agent or application through a customizable, self-contained vision stack built on the…

    1.9k GitHub stars~15k tokensUpdated yesterday
    Auto-check: notes
  • Vss Evaluate Caption Accuracy

    NVIDIA-AI-Blueprints/video-search-and-summarization

    Measure whether an RT-VLM configuration change altered caption quality — capture paired baseline and candidate captions for a set of videos, score both against a ground truth with an LLM judge, and…

    1.9k GitHub stars~2.1k tokensUpdated yesterday
    Auto-check: notes
  • Rtvi Byom Porting

    NVIDIA-AI-Blueprints/video-search-and-summarization

    A skill your agent uses when adding, debugging, or validating a bring-your-own VLM in VSS RT-VLM, including custom Hugging Face or NGC checkpoints, vLLM adapters or plugins, model shims, and…

    1.9k GitHub stars~1.4k tokensUpdated yesterday
    Auto-check passed
  • Vss Manage Alerts

    NVIDIA-AI-Blueprints/video-search-and-summarization

    A skill your agent uses when operating VSS alert workflows — real-time monitoring, Alert-Bridge subscriptions, verification verdicts, on-demand verification, always-on operation, Slack…

    1.9k GitHub stars~11k tokensUpdated yesterday
    Auto-check: warnings

Questions about Vss Search Archive

What does Vss Search Archive do?

A skill your agent uses when a user wants to search archived VSS video that is already registered in a configured deployment — by natural-language, similarity, attribute, object-ID, or lexical tag…. Vss Search Archive is an agent skill from NVIDIA-AI-Blueprints/video-search-and-summarization. Use this skill when a user wants to search archived VSS video that is already registered in a configured deployment — by natural-language, similarity, attribute, object-ID, or lexical tag query.

When should I use Vss Search Archive?

Vss Search Archive fits situations like: A user wants to search archived VSS video that is already registered in a configured deployment — by natural-language; lexical tag query.

How do I install Vss Search Archive in Claude Code?

Run `npx skills add NVIDIA-AI-Blueprints/video-search-and-summarization --skill vss-search-archive -a claude-code`. Or copy the skill folder (skills/operations/vss-search-archive in NVIDIA-AI-Blueprints/video-search-and-summarization) into .claude/skills/vss-search-archive in your project. Claude Code loads it when a task matches its description.

How do I install Vss Search Archive in Codex?

Run `npx skills add NVIDIA-AI-Blueprints/video-search-and-summarization --skill vss-search-archive -a codex`. Or copy the skill folder (skills/operations/vss-search-archive in NVIDIA-AI-Blueprints/video-search-and-summarization) into .agents/skills/vss-search-archive in your project. Codex loads it when a task matches its description.

Can I use Vss Search Archive in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add NVIDIA-AI-Blueprints/video-search-and-summarization --skill vss-search-archive -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/vss-search-archive, .gemini/skills/vss-search-archive, .github/skills/vss-search-archive and .opencode/skills/vss-search-archive in your project.

What does Vss Search Archive need to run?

Going by SKILL.md and its folder, Vss Search Archive needs a shell for the scripts in its folder and the command-line tools its instructions call (docker and uv). Our summary lists: A Bash shell; Docker.

Does Vss Search Archive access the network?

SKILL.md contains no URLs. Its commands use docker and uv, which can reach the network depending on how they are called. This is read from the text; nothing was executed.

Is Vss Search Archive safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. The check reads SKILL.md only: the scripts in the folder are not scanned, so read them before running anything.

What licence does Vss Search Archive use?

Vss Search Archive is published under the Apache-2.0 licence (declared in SKILL.md). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Vss Search Archive use?

About 3.3k tokens (SKILL.md is roughly 13k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 3.6k tokens, read only when the agent opens those files.

What are the alternatives to Vss Search Archive?

Skills that share tags, products or a category with Vss Search Archive: News Aggregator Skill (cclank/news-aggregator-skill, 1.3k stars), Deepgram JS Text Intelligence (deepgram/deepgram-js-sdk, 276 stars), Key Takeaways (aipoch/medical-research-skills, 1.9k stars) and Vss Search Archive (NVIDIA/skills, 3.6k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Vss Search Archive?

NVIDIA-AI-Blueprints (a GitHub organization) maintains it in NVIDIA-AI-Blueprints/video-search-and-summarization, which has 1,919 GitHub stars. The repository holds 22 skills in this directory. The repository was last updated on October 10, 2026.

Source: NVIDIA-AI-Blueprints/video-search-and-summarization on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.