DOCSEARCHbyAlgolia
Submit Docs

Agenta Documentation

website·Indexed 35 days ago

https://docs.agenta.ai

Documentation for Agenta, a platform for configuring, evaluating, and optimizing LLM applications. Learn how to use the Agenta SDK to define application structures, implement custom and built-in evaluators, manage testsets, and run A/B tests to compare prompt configurations and application performance.

Tokens
259.3K
Snippets
1.5K
Records
2.3K
Agent score
67%

What's inside Agenta

  1. Overview of Agenta

    Agenta is a collaborative workspace designed for teams to build AI agents and automations. It allows users to augment their work and automate repetitive tasks without requiring coding knowledge, as agents are built through a chat-based interface.

  2. Overview of Agno AI Agent Framework

    Agno is a Python framework designed for building AI agents. It enables the creation of intelligent agents capable of using tools, producing structured outputs, and executing efficient workflows to interact with external systems and perform complex tasks.
    # Agno is available on GitHub
# https://github.com/agno-agi/agno

  3. Overview of OpenAI Agents SDK

    The OpenAI Agents SDK is a framework designed for building AI agents capable of interacting with tools, executing asynchronous operations, and orchestrating complex multi-agent workflows. It provides a structured approach to developing intelligent agents with native tool-calling capabilities.

  4. Overview of Human Evaluation in Agenta

    Human evaluation in Agenta allows developers to assess the performance of an LLM application using human judgment. This serves as an alternative or supplement to automated metrics, providing qualitative insights into the quality of model outputs.

  7. Overview of Instructor library

    Instructor is a library designed to extract structured data from natural language using Large Language Models (LLMs). It is compatible with multiple LLM providers, including OpenAI, Anthropic, and Google.

  8. Overview of Agenta Observability SDK for Tracing

    The Agenta Observability SDK is a wrapper around the OpenTelemetry SDK designed to simplify the instrumentation of LLM applications. It provides automatic OpenTelemetry setup and supports function instrumentation via decorators, linking traces to specific prompt versions (applications, variants, and environments), and adding custom metadata attributes to spans. For full coverage, the SDK should be used in conjunction with OpenTelemetry auto-instrumentation.

  9. Overview of Fallback Models and Retry settings

    Fallback models and retry settings are integrated into a prompt configuration. These settings allow a prompt to automatically retry failed API calls and, if configured, switch to a backup model to ensure reliability when the primary model fails.

  11. Overview of Single Sign-On (SSO) in Agenta

    Agenta supports Single Sign-On (SSO) via the OpenID Connect (OIDC) protocol, allowing organizations to centralize authentication through identity providers (IdPs) such as Okta, Azure AD, Google Workspace, Auth0, and OneLogin. SSO is available for users on Business and Enterprise plans.

  12. Overview of Agenta Platform Features

    Agenta provides a collaborative environment for developing and managing AI agents with the following core capabilities:

    • Multiplayer Workspace: Shared agents, chat sessions, and file collaboration available via Agenta Cloud or self-hosted infrastructure.
    • Agent-Specific Shared Folders: Dedicated folders for each agent to store documents, research, and shared wikis.
    • Extensibility: Support for 1000+ integrations via built-in connectors or Model Context Protocol (MCP) servers.
    • Reusable Skills: Ability to assign repeatable instructions for specific tasks using custom skills or those from the open agent ecosystem.
    • Governance and Control: Granular permissions to define which actions run automatically, require human approval, or are blocked, with settings applicable at both the agent and tool levels.
    • Automation: Support for background agents triggered by schedules or events in connected applications.
    • Observability: Full tracing of model and tool calls, including tracking for requests, token usage, latency, and estimated costs.
    • Version Control: Complete version history for agents, allowing developers to compare versions and analyze behavioral changes.