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

Julep

Julep is an Apache-2.0 Python framework for defining composable AI agents as durable dataflows. Its `@flow` model converts named steps, registered tools, pure functions, reasoners and control operations into a frozen wire-format intermediate representation. The framework documents retries, timeouts, branching, fan-out, resumable execution, execution projections and explicit tool permissions, with an optional Temporal

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Julep — 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 Julep when an agent process must be represented as a structured, durable workflow rather than an ad-hoc loop, especially when its tools, branches, retries, timeouts or execution traces need explicit treatment.

Who it is for

  • Python developers building multi-step AI agents
  • Teams that need explicit tool permissions and execution controls
  • Developers managing multiple agent flows through a shared CLI
  • Builders working with local, self-hosted or cloud-oriented agent infrastructure

Documented uses

  • Compose multi-step agent workflows from Python functions and registered tools
  • Recover flows after failures and add retry or timeout behavior
  • Inspect agent structure as a graph or derived execution projection
  • Run local dry runs with in-memory tools and deterministic fake reasoners
  • Discover, lint, test, trace and deploy a module of agents through the CLI
Before you choose

Requirements & limits

The repository is public and licensed under Apache-2.0. Installation is documented with `pip install --pre julep`; the README states that the pre-release flag is required for the current release candidate. The documented quickstart requires no API key because it uses a local dry run with a fake reasoner.

  • Python is the primary authoring language and the central @flow model is define-by-construction.
  • Tool and reasoner surfaces are frozen during deployment, supporting explicit capability boundaries.
  • Retries, timeouts, branches, fan-out and resumable execution are documented as workflow features.
  • The CLI works across a module of agents and includes graph, lint, test, trace, doctor and deployment commands.
  • The pure core is described as dependency-free, while Temporal is an optional layer.

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

Define and locally dry-run a support triage flow

Create the documented Python flow pattern that looks up a ticket, builds a prompt, calls a reasoner and performs a local dry run.

Before you start

  • Python environment with the pre-release Julep package installed using pip install --pre julep.
  • A Python script containing the documented @tool, @pure, Reasoner, @flow and deploy definitions.
  • An in-memory tool and deterministic fake reasoner for the dry run; the documented example does not require an API key.

Steps

  1. Define a registered ticket lookup tool and a pure function that converts its result into reasoner input.

  2. Create a typed reasoner and compose the lookup, prompt and think steps inside an @flow.

  3. Call deploy with the flow's tools and reasoners.

  4. Call deployment.dry_run with the documented input, fake reasoner and local tool setup, then inspect the returned value.

Expected outcome

A locally evaluated support-triage flow returns the ticket data together with a generated support reply value.

Guide limitations

  • The documented recipe uses a deterministic fake reasoner for local execution rather than a live model call.
  • The source describes the flow and dry-run path but does not establish production deployment, provider compatibility or successful execution beyond the example.
Documentation 1 ↗
Evidence and transparency

Sources & review

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