tapeagents

TapeAgents is a framework that facilitates all stages of the LLM Agent development lifecycle

https://github.com/servicenow/tapeagents

Science Score: 54.0%

This score indicates how likely this project is to be science-related based on various indicators:

✓
CITATION.cff file
Found CITATION.cff file
✓
codemeta.json file
Found codemeta.json file
✓
.zenodo.json file
Found .zenodo.json file
○
DOI references
✓
Academic publication links
Links to: arxiv.org
○
Academic email domains
○
Institutional organization owner
○
JOSS paper metadata
○
Scientific vocabulary similarity
Low similarity (13.8%) to scientific vocabulary

Keywords

agentic agents ai-agents finetuning llm-agent multi-agent multi-agent-simulation prompt-tuning

Last synced: 6 months ago · JSON representation ·

Repository

TapeAgents is a framework that facilitates all stages of the LLM Agent development lifecycle

Basic Info

Host: GitHub
Owner: ServiceNow
License: apache-2.0
Language: Python
Default Branch: main
Homepage: https://arxiv.org/abs/2412.08445
Size: 188 MB

Statistics

Stars: 294
Watchers: 12
Forks: 36
Open Issues: 29
Releases: 17

Topics

agentic agents ai-agents finetuning llm-agent multi-agent multi-agent-simulation prompt-tuning

Created over 1 year ago · Last pushed 6 months ago

Metadata Files

Readme License Citation Notice

TapeAgents

TapeAgents is a framework that leverages a structured, replayable log (Tape) of the agent session to facilitate all stages of the LLM Agent development lifecycle. In TapeAgents, the agent reasons by processing the tape and the LLM output to produce new thoughts, actions, control flow steps and append them to the tape. The environment then reacts to the agent’s actions by likewise appending observation steps to the tape.

Why TapeAgents

Key features:

Build your agent as a low-level state machine, as a high-level multi-agent team configuration, or as a mono-agent guided by multiple prompts
Debug your agent with TapeAgent studio or TapeBrowser apps
Serve your agent with response streaming
Optimize your agent's configuration using successful tapes; finetune the LLM using revised tapes.

The Tape-centric design of TapeAgents will help you at all stages of your project:

Build with ultimate flexibility of having access to tape for making prompts and generating next steps
Change your prompts or team structure and resume the debug session as long as the new agent can continue from the older tape
Fully control the Agent's tape and the Agent's acting when you use a TapeAgent in an app
Optimize tapes and agents using the carefully crafted metadata structure that links together tapes, steps, llm calls and agent configurations

TapeAgents at a glance

TapeAgents Overview

Getting Started

Tutorial

Start with the introductory Jupyter notebook to quickly learn the core concepts of the framework.

Hello TapeAgents

The simplest agent just to show the basic structure of the agent:

```python from tapeagents.agent import Agent, Node from tapeagents.core import Prompt from tapeagents.dialogtape import AssistantStep, UserStep, DialogTape from tapeagents.llms import LLMStream, LiteLLM from tapeagents.prompting import tapeto_messages

llm = LiteLLM(model_name="gpt-4o-mini")

class MainNode(Node): def makeprompt(self, agent: Agent, tape: DialogTape) -> Prompt: # Render the whole tape into the prompt, each step is converted to message return Prompt(messages=tapeto_messages(tape))

def generate_steps(self, agent: Agent, tape: DialogTape, llm_stream: LLMStream):
    # Generate single tape step from the LLM output messages stream.
    yield AssistantStep(content=llm_stream.get_text())

agent = Agent[DialogTape].create(llm, nodes=[MainNode()]) starttape = DialogTape(steps=[UserStep(content="Tell me about Montreal in 3 sentences")]) finaltape = agent.run(starttape).getfinaltape() # agent will start executing the first node print(f"Final tape: {finaltape.modeldumpjson(indent=2)}") ```

TapeAgents Examples

The examples/ directory contains examples of how to use the TapeAgents framework for building, debugging, serving and improving agents. Each example is a self-contained Python script (or module) that demonstrates how to use the framework to build an agent for a specific task:

How to build a single agent that does planning, searches the web and uses code interpreter to answer knowledge-grounded questions, solving the tasks from the GAIA benchmark.
How to build a team of TapeAgents with AutoGen-style low-code programming paradigm
How to finetune a TapeAgent with a small LLM to be better at math problem solving on GSM-8k dataset.

Other notable examples that demonstrate the main aspects of the framework:

workarena - custom agent that solves WorkArena benchmark using BrowserGym environment.
tape_improver.py - the agent that revisit and improves the tapes produced by another agent.

To run these examples, simply use:

zsh uv run -m examples.<MODULE> <ARGS>

Installation

Install the latest release with its minimal dependencies:

zsh pip install tapeagents

You can also install converters and finetune optional dependencies

zsh pip install 'tapeagents[converters,finetune]'

Building from source

Install uv to manage package:

Official documentation here

Clone the repository:

zsh git clone https://github.com/ServiceNow/TapeAgents.git cd TapeAgents

Create venv environment and install dependencies:

```zsh make setup

equivalent to `uv sync --all-extras`

```

Learn more

See our full TapeAgents documentation.

For an in-depth understanding of the design principles, architecture, and research behind TapeAgents, see our technical report.

Contacts

Feel free to reach out to the team:

Dzmitry Bahdanau, dzmitry.bahdanau@servicenow.com
Oleh Shliazhko, oleh.shliazhko@servicenow.com
Jordan Prince Tremblay, jordanprince.t@servicenow.com
Alexandre Piché, alexandre.piche@servicenow.com

Acknowledgements

We acknowledge the inspiration we took from prior frameworks, in particular LangGraph, AutoGen, AIWaves Agents and DSPy.

Owner

Name: ServiceNow
Login: ServiceNow
Kind: organization

Website: https://www.servicenow.com
Repositories: 147
Profile: https://github.com/ServiceNow

Works for you™

Citation (CITATION.cff)

cff-version: 1.2.0
title: >-
  TapeAgents: a Holistic Framework for Agent Development and
  Optimization
message: 'If you want to refer to our paper or code, please cite it as below.'
type: software
authors:
  - family-names: Bahdanau
    given-names: Dzmitry
  - family-names: Gontier
    given-names: Nicolas
  - family-names: Huang
    given-names: Gabriel
  - family-names: Kamalloo
    given-names: Ehsan
  - family-names: Pardinas
    given-names: Rafael
  - family-names: Piché
    given-names: Alex
  - family-names: Scholak
    given-names: Torsten
  - family-names: Shliazhko
    given-names: Oleh
  - family-names: Jordan Prince
    given-names: Tremblay
  - family-names: Ghanem
    given-names: Karam
  - family-names: Parikh
    given-names: Soham
  - family-names: Tiwari
    given-names: Mitul
  - family-names: Vohra
    given-names: Quaizar
identifiers:
  - type: url
    value: 'https://arxiv.org/abs/2412.08445'
repository-code: 'https://github.com/ServiceNow/TapeAgents'
abstract: >-
  TapeAgents is a framework that facilitates all stages of
  the LLM Agent development lifecycle.
license: Apache-2.0
date-released: '2024-10-16'

GitHub Events

Total

Create event: 148
Commit comment event: 2
Release event: 12
Delete event: 91
Member event: 7
Pull request event: 248
Fork event: 29
Issues event: 36
Watch event: 267
Issue comment event: 101
Public event: 1
Push event: 1,421
Pull request review comment event: 321
Pull request review event: 381

Last Year

Create event: 148
Commit comment event: 2
Release event: 12
Delete event: 91
Member event: 7
Pull request event: 248
Fork event: 29
Issues event: 36
Watch event: 267
Issue comment event: 101
Public event: 1
Push event: 1,421
Pull request review comment event: 321
Pull request review event: 381

Issues and Pull Requests

Last synced: 6 months ago

All Time

Total issues: 8
Total pull requests: 78
Average time to close issues: N/A
Average time to close pull requests: 12 days
Total issue authors: 6
Total pull request authors: 12
Average comments per issue: 0.0
Average comments per pull request: 0.41
Merged pull requests: 47
Bot issues: 0
Bot pull requests: 0

Past Year

Issues: 8
Pull requests: 78
Average time to close issues: N/A
Average time to close pull requests: 12 days
Issue authors: 6
Pull request authors: 12
Average comments per issue: 0.0
Average comments per pull request: 0.41
Merged pull requests: 47
Bot issues: 0
Bot pull requests: 0

View more stats

Top Authors

Issue Authors

rizar (13)
ollmer (3)
AlexPiche (3)
gabrielhuang (3)
haozaiiii (2)
XvHaidong (1)
mitultiwari (1)
sky-fly97 (1)
emrecanacikgoz (1)

Pull Request Authors

ollmer (45)
AlexPiche (22)
jpt-sn (21)
rizar (18)
danieltremblay (9)
NicolasAG (5)
gabrielhuang (5)
rafapi (5)
ehsk (4)
dependabot[bot] (2)
mitultiwari (1)
amilios (1)
christyler3030 (1)

Top Labels

Issue Labels

Pull Request Labels

dependencies (2)

Packages

Total packages: 1
Total downloads:
- pypi 3,242 last-month

Total dependent packages: 0
Total dependent repositories: 0
Total versions: 19
Total maintainers: 2

pypi.org: tapeagents

TapeAgents framework for building, tuning and evaluating LLM Agents

Documentation: https://tapeagents.readthedocs.io/
License: Apache-2.0
Latest release: 0.1.16
published 6 months ago

Versions: 19
Dependent Packages: 0
Dependent Repositories: 0
Downloads: 3,242 Last month

Rankings

Dependent packages count: 10.0%

Average: 33.0%

Dependent repos count: 56.1%

Maintainers (2)

olmer jordanprincet