OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Infrastructure & Serving

Flama

Flama is a Python framework that packages predictive and generative AI models as portable .flm artifacts and serves them over HTTP.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Flama — OpenViST project profile in AI Infrastructure & Serving.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Flama when you have a model to package and want to serve it as an HTTP API, or when a generative model also needs compatible chat endpoints or a bundled chat interface.

Who it is for

  • Developers serving an existing predictive model through an inference API.
  • Teams exposing generative models through compatible chat endpoints or a browser chat interface.
  • Developers making tools, resources or prompts available to AI agents through MCP.

Documented uses

  • Package models from mainstream frameworks into a common .flm artifact.
  • Serve predictive inference through HTTP routes with an OpenAPI schema and Swagger UI.
  • Offer generative models through OpenAI-, Anthropic- and Ollama-compatible endpoints.
  • Provide a streaming chat interface for a served model.
  • Expose tools, resources and prompts to agents using native MCP support.
Before you choose

Requirements & limits

Flama is open source under the Apache-2.0 license and is published on PyPI.

  • The model is packaged into Flama's .flm artifact format before serving.
  • Installation requires a schema library; Pydantic is the usual choice, with TypeSystem and Marshmallow also listed.
  • Generative serving uses the optional llm extra and has documented vLLM/Linux/CUDA and MLX/Apple Silicon backend paths.
  • The README states Python 3.10–3.13 for the llm extra.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Serve a packaged predictive model

Expose an existing predictive model through an HTTP inference endpoint.

Before you start

  • A predictive model and its framework dependencies; the documented example uses scikit-learn.
  • Flama installed with a schema library, with Pydantic shown as the usual choice.

Steps

  1. Package the model as a .flm artifact with Flama's dump function.

  2. Run flama serve and point its model option to the artifact.

  3. Send a JSON request to the generated /model/predict/ endpoint; consult /schema/ or /docs/ for the API description.

Expected outcome

An HTTP service with model metadata, an inference route, an OpenAPI schema and Swagger UI.

Guide limitations

  • The example requires the relevant model framework, such as scikit-learn.
  • The recipe describes serving a model artifact; it does not cover model training.
Documentation 1 ↗
Source-based practical guide

Package and serve a generative model

Download a Hugging Face model, package it and serve it with Flama's chat interface.

Before you start

  • Python 3.10–3.13, Flama's LLM and Pydantic extras, and a supported inference backend.
  • The README describes vLLM on Linux with CUDA or MLX on Apple Silicon.

Steps

  1. Install Flama with the llm and pydantic extras.

  2. Use flama get with the llm family and Hugging Face source to package a model as a .flm artifact.

  3. Run flama serve with that artifact as the model.

Expected outcome

An HTTP API with compatible endpoint dialects and a streaming chat interface at /chat/.

Guide limitations

  • The documented backend options depend on platform: vLLM on Linux with CUDA, or MLX on Apple Silicon.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
vortico/flama · Official homepage title · retained in its source language
How we review sourcesSuggest a correction