memvid-ask-model

LLM inference module for Memvid Q&A with local and cloud model support.

About

memvid-ask-model provides LLM inference capabilities for Memvid's Q&A functionality. It supports both local inference using llama.cpp and cloud APIs (OpenAI, Claude, Gemini).

Features

Local Inference - Built-in llama.cpp with phi1.5 model (no API key needed)
OpenAI - GPT-4 and GPT-3.5 support
Anthropic - Claude models
Google - Gemini models
RAG Integration - Seamlessly works with memvid-core search results

Installation

[dependencies]
memvid-ask-model = "2.0.102"

Usage

use memvid_ask_model::run_model_inference;
use memvid_core::Memvid;

// Get search results from memvid-core
let mem = Memvid::open("knowledge.mv2")?;
let hits = mem.find("topic", 5)?;

// Run inference with local model
let answer = run_model_inference(
    "What is this about?",
    &hits,
    None, // Use local model
)?;

// Or use cloud API
let answer = run_model_inference(
    "Summarize the findings",
    &hits,
    Some("openai"), // Requires OPENAI_API_KEY
)?;

Environment Variables

For cloud models, set the appropriate API key:

export OPENAI_API_KEY=sk-...
export ANTHROPIC_API_KEY=sk-ant-...
export GEMINI_API_KEY=...

Documentation

License

Licensed under Apache 2.0

Apache License 2.0

memvid-ask-model 2.0.140