vccs
Composable charting components for Vue 3, ported from Recharts
The unified multimodal backend agents build with. Database, orchestration, and serving in one Python file. Apply locally, serve, then publish the same file to Pixeltable Cloud.
DataPython · ShellGitHub @pixeltableOnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
No updates detected yet. GitHub releases will appear here.
Owners of public repositories under a personal GitHub account can sign in to verify ownership and manage the introduction themselves. No administrator approval is needed.
Sign in, generate a verification file, and commit it to this repository’s default branch within 30 minutes. Files on other branches or forks do not count. Then check the file below to manage the project without administrator approval. Keep the file to retain your access.
If needed, you will go to sign-in. After signing in, return here and press your chosen verification button again.
Run /nomorevibe verify in your project folder to verify ownership and manage the introduction yourself.
Verify in one minuteAsk about other verification options
Skill/Plugin for building multimodal AI applications with Pixeltable. Tables, computed columns, embedding search, UDFs, and 30+ AI provider integrations.
Taken from the repository README
Loading the full README. The saved excerpt is shown below.
Agent Skill that teaches AI coding assistants to write Pixeltable application files: TableModel in app.py, then pxt schema update, then pxt service update.
Verified against Pixeltable 0.7.8 on 2026-09-19: docs/review-2.11.0.md.
Install the portable skill with npx skills:
This installs SKILL.md and its references for supported coding assistants,
including Cursor, Copilot, Windsurf, and the Codex IDE extension.
Google Antigravity is not one of them: it reads
~/.gemini/antigravity/skills, which npx skills does not write to. Install
there with ./install.sh --platform antigravity.Read the full README ↗Multimodal RAG, video search, image similarity search, agent memory, and embeddings stored with your data.
This plugin teaches AI coding assistants to build Pixeltable applications: tables with computed columns, embedding indexes and vector search, RAG, tool-calling agents, and video, audio, image, and document pipelines across 25+ AI providers. The agent declares the schema as TableModel classes in app.py, creates the tables with pxt schema update, and serves HTTP routes with pxt service update.
Catalog skills (thin routers into the canonical pixeltable skill) are named for the jobs people search: multimodal-rag, video-search, image-similarity-search, audio-transcription-search, agent-memory, vector-database-alternative, incremental-ml-pipeline, dataset-versioning.
Verified against Pixeltable 0.7.11 on 2026-09-28: docs/review-2.12.0.md. Catalog snippets re-checked against Pixeltable 0.7.21 docs on 2026-10-27: docs/review-2.13.0.md.
Install the portable skill with npx skills:
npx skills add pixeltable/pixeltable-skill
This installs every skills/*/SKILL.md (the canonical pixeltable skill plus
the task-named catalog skills) and the canonical skill's references for
supported coding assistants, including Cursor, Copilot, Windsurf, and the
Codex IDE extension.
With only Python, pxt skills install (Pixeltable 0.7.21) copies only
skills/pixeltable-skill into .claude/skills/pixeltable and
.agents/skills/pixeltable. It does not install the catalog skills.
pxt skills install -f updates that canonical copy.
Google Antigravity is not one of them: it reads
~/.gemini/antigravity/skills, which npx skills does not write to. Install
there with ./install.sh --platform antigravity.
gemini extensions install https://github.com/pixeltable/pixeltable-skill
This loads skills/pixeltable-skill/SKILL.md, the hosted Cloud MCP
(pixeltable-cloud at https://www.pixeltable.com/mcp/cloud), and the local
developer MCP (pixeltable via uvx mcp-server-pixeltable-developer). Sign in
to Cloud with /mcp auth pixeltable-cloud if those tools are not listed. The
hosted server supports reads and writes; it is not read-only. The local server
needs uv on PATH. Set the catalog directory with
gemini extensions config pixeltable (PIXELTABLE_HOME; Pixeltable's default
is ~/.pixeltable). Restart the CLI after installing.
Install the skill with its client-supported commands, agents, and hooks using npx plugins:
npx plugins add pixeltable/pixeltable-skill
Client-native marketplace options:
/plugin marketplace add pixeltable/pixeltable-skill, then
/plugin install pixeltable@pixeltable-skillcodex plugin marketplace add pixeltable/pixeltable-skill --ref main, then
codex plugin add pixeltable@pixeltable-skill. Use /plugins in Codex or the
Plugins page in ChatGPT desktop.Restart or start a new conversation after installing. Repository marketplaces are team-distribution sources; they do not publish to a public directory.
skills/pixeltable-skill/ # Canonical teaching skill (name: pixeltable)
├── SKILL.md # Contract: app.py, schema update, service update
└── references/
├── core-api.md # Tables, querying, views, UDFs, config
├── cli.md # pxt CLI
├── providers.md # Import and output shape
├── workflows.md # FastAPIRouter
└── anti-patterns.md # Wrong/right stack
skills/<job>/SKILL.md # Catalog skills: multimodal-rag, video-search, ...
commands/ # /pixeltable:pixeltable-start, index-videos, ...
Job-named commands (Claude /pixeltable:<name>, Cursor /<name>):
pixeltable-start, index-videos, search-images, rag-over-docs,
add-agent-memory, transcribe-and-search-audio, store-embeddings,
incremental-pipeline, version-dataset. Original scaffold and
add-provider still work.
Client manifests and marketplace metadata live alongside the shared skill so
each supported installer can discover the format it understands. Gemini CLI
reads gemini-extension.json and GEMINI.md at the repo root and loads
skills/*/SKILL.md.
Most agents do not ship an MCP server. Pixeltable Cloud users can add the hosted Cloud MCP at https://www.pixeltable.com/mcp/cloud and sign in through the browser. It lists the organization selected at sign-in: databases, services, catalog entries, schemas, up to 25 table rows per call, and logs. It also starts a database at once. Stopping a database, restarting a service, or setting a secret returns a preview first and runs only when called again with confirm: true. It deletes nothing and does not deploy; use the pxt CLI for that. Add ?read_only=true to the URL for the read tools only, or ?db=<name> to limit it to one database. The Skill mentions it so an agent can suggest it to a Cloud user.
Gemini CLI is the exception: this repo's extension installs pixeltable-cloud (hosted; OAuth in the CLI, no API key in the manifest) and pixeltable (local developer MCP). The hosted server is not read-only; it supports the reads and writes above.
The Docs MCP searches public documentation. The developer MCP is a separate local server with catalog, query, scaffold, schema, and service tools.
See CONTRIBUTING.md. Run python3 scripts/validate_plugin.py after structural changes.
pxt init then pxt service example --out app.py then pxt schema update app.py my_app then pxt service update app.py my_app.Apache 2.0