@qualve/ai

Meta-package for Qualve LLM support. Installs the core LLM framework and all official provider adapters in one go.

Setup

Requires Node.js v23+.

npm install @qualve/ai

This installs:

• @qualve/llm — Core LLM task framework
• @qualve/anthropic — Claude provider
• @qualve/openai — OpenAI provider
• @qualve/googleai — Gemini provider

If you only need specific providers, install them individually instead (each pulls in @qualve/llm automatically).

Usage

import "@qualve/ai";

Importing the package registers all three providers with the Qualve task system.

You can also import individual providers via sub-paths:

import "@qualve/ai/anthropic";
import "@qualve/ai/openai";
import "@qualve/ai/googleai";

Or import the core framework:

import { LLMTask } from "@qualve/ai/core";

API Keys

Create a .env file with API keys for the providers you want to use:

GEMINI_API_KEY=...     # https://aistudio.google.com/api-keys
OPENAI_API_KEY=...     # https://platform.openai.com/api-keys
ANTHROPIC_API_KEY=...  # https://platform.claude.com/settings/keys

Packages

| Package | Description | | --- | --- | | @qualve/llm | Core LLM task framework (LLMTask class) | | @qualve/anthropic | Claude adapter | | @qualve/openai | OpenAI adapter | | @qualve/googleai | Gemini adapter |

Models

| Provider | Model | Context window | Max output | | --- | --- | --- | --- | | Gemini | gemini-3.1-pro-preview* | 1,048,576 | 65,536 | | Gemini | gemini-3.1-flash-preview | 1,048,576 | 65,536 | | Gemini | gemini-3.1-flash-lite-preview | 1,048,576 | 65,536 | | OpenAI | gpt-5.4* | 1,050,000 | 128K | | OpenAI | gpt-5-mini | 400K | 128K | | OpenAI | gpt-5-nano | 400K | 128K | | Claude | claude-sonnet-4-6* | 1M | 64K | | Claude | claude-haiku-4-6 | 200K | 64K | | Claude | claude-opus-4-5 | 1M | 128K |

* Default

Options

| Option | Flag | Description | | --- | --- | --- | | llm | --llm | Provider to use (gemini, openai, claude) | | model | --model | Model name (see table above) | | thinking | --thinking | Reasoning effort level | | fresh | --fresh | Force re-upload of input files |

Thinking levels

Control reasoning effort via --thinking <LEVEL> or the thinking task property.

| Provider | Accepted values | | --- | --- | | Gemini | minimal, low, medium, high* | | OpenAI | none, minimal, low, medium*, high, xhigh | | Claude | (not yet configurable) |