OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Research Agents

Af Deep Research

Af Deep Research is an open-source Python backend and research API that orchestrates parallel agents, evaluates findings against quality thresholds, generates follow-up sub-queries when gaps appear, and returns structured research results. It is built on the AgentField runtime and is designed for integration into AI applications rather than use as a chat interface.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Af Deep Research — OpenViST project profile in Research Agents.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Af Deep Research when an AI product needs a backend for broad, iterative investigation and structured evidence output. Its documented interface supports a one-call research workflow through the af CLI or an HTTP request.

Who it is for

  • Developers building AI applications that need automated, iterative research
  • Teams creating knowledge-discovery or evidence-gathering workflows
  • Engineers needing structured research outputs with entities, relationships, evidence, and cited documents

Documented uses

  • Submit a research question to an autonomous multi-agent workflow
  • Identify gaps in initial findings and generate follow-up research queries
  • Evaluate collected findings against quality thresholds
  • Produce structured entities and relationships alongside article evidence
  • Generate a hierarchical research document with inline citations and bibliography
Before you choose

Requirements & limits

The repository is publicly available under the Apache-2.0 license. The documented setup requires provider API keys, including an OpenRouter key and at least one search-provider key. The project is marked as an early preview and notes that APIs may change.

  • The output combines entities, relationships, article evidence, a hierarchical document, and execution metadata.
  • The system is described as using roughly 10,000 logical agent invocations per research.
  • Docker setup and direct local Python execution are both documented.
  • Local execution requires a writable workspace; the default /workspaces path may require root access.
  • The repository does not specify API authentication, endpoint format beyond the documented route, or deployment configuration details.

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

Start Af Deep Research with Docker

Set up the documented Docker-based research backend and prepare it to receive a query.

Before you start

  • A clone of the Agent-Field/af-deep-research repository
  • Docker Compose
  • An OpenRouter API key
  • At least one search-provider key

Steps

  1. Clone the repository and change into its directory.

  2. Copy .env.example to .env and add the OpenRouter and search-provider keys.

  3. Run docker compose up --build.

  4. Use the documented HTTP request against the local execution route to submit a research query.

Expected outcome

The backend starts locally and returns an execution_id for the submitted request; the documented workflow supports streaming progress through SSE and fetching results when complete.

Guide limitations

  • The repository notes that the project is an early preview and APIs may change.
  • The supplied documentation does not specify authentication or deployment configuration beyond these instructions.
Documentation 1 ↗
Source-based practical guide

Run Af Deep Research locally as a Python app

Install and start the documented synchronous local mode without the AgentField control plane managing asynchronous dispatch.

Before you start

  • Python
  • A writable working directory
  • An OpenRouter API key
  • At least one search-provider key

Steps

  1. Create and activate a virtual environment, then install the project with pip install -e .

  2. Set PR_AF_WORKDIR if the default /workspaces location is not suitable or requires root access.

  3. Copy .env.example to .env and add the required provider keys.

  4. Start the application with python3 main.py.

  5. Submit a query from a separate terminal using the documented HTTP request.

Expected outcome

The Python application starts locally and accepts a research request through the documented API route.

Guide limitations

  • This mode executes synchronously.
  • The supplied documentation does not establish deployment environments, API authentication, or compatibility guarantees.
Documentation 1 ↗
COMPARE OPTIONS

Compare options for Agent orchestration

These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.

ProjectDocumented purposePlatformAccess and requirements
Af Deep ResearchSubmit a research question to an autonomous multi-agent workflowCloud, ServerThe repository is publicly available under the Apache-2.0 license. The documented setup requires provider API keys, including an OpenRouter key and at least one search-provider key. The project is marked as an early preview and notes that APIs may change.
Local RAG Researcher with DeepSeek R1 & LangGraphTurn a user instruction into research questions and retrieve relevant local documents.Self-hostedThe documented application is run from the Python repository on a self-hosted setup. The README lists Python 3.9 or later and Ollama as prerequisites; it describes a Tavily API key as optional for web search.
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
4 October 2026
Original website title
Agent-Field/af-deep-research · Official homepage title · retained in its source language
How we review sourcesSuggest a correction