video-production
π―Skillfrom cklxx/elephant.ai
|Generates professional video content by autonomously scripting, storyboarding, and rendering multimedia presentations using AI-driven production workflows.
Installation
npx skills add https://github.com/cklxx/elephant.ai --skill video-productionSkill Details
Overview
# elephant.ai
[](https://github.com/cklxx/Alex-Code/actions/workflows/ci.yml)
[](https://goreportcard.com/report/github.com/cklxx/Alex-Code)
[](https://opensource.org/licenses/MIT)
elephant.ai is a unified runtime for production-grade AI agents. The alex CLI/TUI, alex-server, and web UI share one execution core, typed event stream, and observability pipeline. Build once, run everywhere, and render the same session in terminals and the browser.
One-line pitch: A single runtime that ships agentic workflows across CLI, server, and web with typed events, evals, and cost/trace accountability.
---
Why elephant.ai
- One runtime, three entrypoints (CLI/TUI, server, dashboard) backed by the same DI container.
- OpenAI chat + Responses API, Anthropic Claude API, and OpenAI-compatible gateways (OpenRouter/DeepSeek/Antigravity).
llm_provider: auto(orcli) picks the best available subscription from CLI auth + env keys (Codex/Antigravity/Claude/OpenAI).- Typed, artifact-aware events render identically in terminals and the web UI.
- Built-in observability: structured logs, OpenTelemetry traces, Prometheus metrics, and per-session cost accounting.
- Retrieval layers for memory, skills, docs, and external context plus approvals for risky actions.
- Evaluation harnesses (including SWE-Bench) live in-repo for parity between manual and automated runs.
---
Architecture (short form)
```
Delivery (CLI, Server, Web) β Agent Application Layer β Domain Ports β Infrastructure Adapters
```
- Delivery: shared DI wiring for
cmd/alex,cmd/alex-server, andweb/keeps sessions consistent. - Agent core:
internal/agent/{app,domain,ports}implements the Think β Act β Observe loop, approvals, and typed events. - Infrastructure:
internal/di,internal/tools,internal/toolregistry,internal/llm,internal/mcp,internal/session,internal/storage,internal/observability, andinternal/contextprovide adapters, context, telemetry, and persistence. - Frontend:
web/streams SSE events, artifacts, approvals, and cost details in real time.
---
Getting started
Prerequisites: Go 1.24+, Node.js 20+ (web UI), Docker (optional).
Quickstart
```bash
# Configure your LLM provider (examples)
export OPENAI_API_KEY="sk-..."
# export ANTHROPIC_API_KEY="sk-ant-..." # Claude
# export CLAUDE_CODE_OAUTH_TOKEN="..." # Claude Code OAuth
# export CODEX_API_KEY="sk-..." # OpenAI Responses / Codex
# export ANTIGRAVITY_API_KEY="..." # Antigravity (OpenAI-compatible)
cp examples/config/runtime-config.yaml ~/.alex/config.yaml
# Run backend + web together
./dev.sh
# Optional: build and run the CLI/TUI
make build
./alex
./alex "print the repo layout"
```
Configuration is shared across every surface. Use examples/config/runtime-config.yaml and docs/reference/CONFIG.md for the canonical schema.
Helpful dev.sh commands
```bash
# Check status/logs
./dev.sh status
./dev.sh logs server
./dev.sh logs web
# Stop services
./dev.sh down
```
---
Demo (10 minutes)
```bash
# With services running from the quickstart
./alex "Map the runtime layers, explain the event stream, and produce a short summary."
```
Expected:
- CLI shows a typed event stream (plan/tool/output).
- Web UI at
http://localhost:3000renders the same events and artifacts. - Cost + trace metadata update as the run progresses.
---
Operations & quality gates
- Quality gates (Go + web):
./dev.sh lint,./dev.sh test, andnpm --prefix web run e2e. - Evaluation harnesses + reproducible results:
evaluation/andevaluation_results/. - Observability stac