OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Frameworks

Forge — antoinezambelli

Forge is a Python framework that adds reliability controls to self-hosted LLM tool-calling loops. It supports optional workflow constraints, response validation and tool-call recovery, and can be used through a Python runner, middleware, or a proxy for compatible clients.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Forge — antoinezambelli — OpenViST project profile in Agent Frameworks.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Forge when developing an LLM tool-calling loop and you need recovery and validation controls, optional workflow constraints, or a proxy-based way to apply guardrails to a compatible client.

Who it is for

  • Python developers building LLM applications that call tools.
  • Developers who want to add tool-call guardrails to an existing orchestration loop.
  • Teams using compatible clients with a local model server who want to place a guardrail proxy between them.

Documented uses

  • Define callable tools for an LLM and let it select them during a workflow.
  • Constrain a tool-calling loop with required steps, prerequisites, or a terminal tool.
  • Validate model responses, recover malformed tool calls, and prompt retries.
  • Add Forge guardrails to a compatible client through its proxy without integrating the Python library into that client.
Before you choose

Requirements & limits

The repository identifies Forge as open source under the MIT license. The Python package requires Python 3.12 or later and a running backend; the standalone Proxy bundles a Python runtime but not a model or backend.

  • Forge sits within one agentic loop; the project says multi-agent graphs, DAG planners, and cross-agent coordination are out of scope.
  • Workflow structure is optional: developers can use guardrails without required steps or add constraints when needed.
  • Python-library use requires a running backend; the listed choices include local backends and Anthropic.
  • The proxy supports OpenAI chat-completions and Anthropic Messages-compatible clients, but does not supply a backend, credentials, or client configuration.

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

Install the Python package and prepare a backend

Set up the documented Python-library prerequisites for using Forge with a chosen backend.

Before you start

  • Python 3.12 or later
  • A running LLM backend
  • For Anthropic, the optional package extra and an Anthropic API key

Steps

  1. Install the core package with the command below.

    Documented command
    pip install forge-guardrails

  2. Choose and prepare a backend using the repository's Backend Setup guide; the README gives llama-server and Ollama as local examples.

  3. For Anthropic, install forge-guardrails[anthropic] and set ANTHROPIC_API_KEY.

Expected outcome

The Python package and a selected backend are prepared as prerequisites for using Forge.

Guide limitations

  • This guide covers package and backend preparation, not running a workflow.
  • The Python package does not install the global forge-proxy command.
Documentation 1 ↗Documentation 2 ↗
Evidence and transparency

Sources & review

Official repository ↗Guide documentation ↗
Source record details
Source availability
Active
Sources checked
11 October 2026
Original website title
antoinezambelli/forge · Official homepage title · retained in its source language
How we review sourcesSuggest a correction