{#- Landing page: hero (logo + text) → example → features → install. Selected via `template: home.html` in index.md front matter. Extends main.html so the header/nav bar is preserved. -#} {% extends "main.html" %} {% block tabs %} {{ super() }} {# Page-scoped resets: only active when this template is rendered. #} {# ── Hero ─────────────────────────────────────────────────────── #}

Multimodal orchestration for LLM APIs

You describe what to analyze. Pollux handles source patterns, context caching, and the multimodal plumbing.

{# ── Example ──────────────────────────────────────────────────── #}
from pollux import run, Config, Source

result = await run(
    "Summarize the key findings",
    source=Source.from_file("paper.pdf"),
    config=Config(
        provider="gemini",
        model="gemini-2.5-flash-lite",
    ),
)
print(result["answers"][0])
{# ── Features ─────────────────────────────────────────────────── #}

Multimodal-first

PDFs, images, video, YouTube, arXiv. One interface, any source type.

Source patterns

Fan-out, fan-in, and broadcast execution over your content. No boilerplate.

Context caching

Upload once, reuse across prompts. Automatic TTL management saves tokens and money.

Built for reliability

Async pipeline, retries with backoff, structured output, usage tracking.

{# ── Install ──────────────────────────────────────────────────── #}
pip install pollux-ai
{% endblock %} {# Suppress Markdown body (SEO fallback text lives in index.md but does not render visually). #} {% block content %}{% endblock %} {# Homepage footer: meta row only (copyright + social), no prev/next pills. Reuses pollux-footer-meta__* classes from extra.css. #} {% block footer %} {% endblock %}