OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Memory & RAG

Granite Retrieval Agent

Granite Retrieval Agent is an open-source Python repository containing a Granite-based agent for retrieval-augmented research over local documents and web sources, plus a separate image-research agent. The retrieval workflow uses Autogen/AG2 and Granite 4; the image workflow uses CrewAI, Granite 4 Tiny-H, and a vision backend. Both are configured through Open WebUI, with Ollama documented for local model serving.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Granite Retrieval Agent — OpenViST project profile in Agent Memory & RAG.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Choose the retrieval workflow for research grounded in a local document collection alongside web sources. Choose the image workflow when an image should guide object analysis and follow-up research.

Who it is for

  • Developers building local document-question-answering and web-research workflows with Granite.
  • Researchers who want an agent workflow that investigates image contents and related material.
  • People adapting Python agent examples built with Autogen/AG2 or CrewAI.

Documented uses

  • Query uploaded local documents and combine findings with web retrieval.
  • Use multi-agent planning and adaptive execution to investigate a research question.
  • Analyze an uploaded image, identify components, and seek related research or historical context.
  • Configure a local Granite-powered research workflow through Open WebUI.
Before you choose

Requirements & limits

The repository is public and identifies its license as Apache-2.0. Its documented setup runs components locally through Ollama and Open WebUI.

  • The repository provides two different workflows, with separate scripts and frameworks: Autogen/AG2 for retrieval and CrewAI for image research.
  • The retrieval model is ibm/granite4:latest; the image agent uses ibm/granite4:tiny-h.
  • Image analysis requires a vision backend; the README names granite3.2-vision:2b as an option.
  • Open WebUI is the documented interaction and function-configuration layer; SearXNG is an optional web-search provider.
  • The repository describes local execution and gives one MacBook Pro M3 Max example, not a universal hardware specification.

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

Set up the retrieval agent in Open WebUI

Prepare the documented local components, add the retrieval function, and load documents for querying.

Before you start

  • Ollama installed
  • Open WebUI installed and serving locally
  • The repository's granite_autogen_rag.py script
  • Local documents to add to an Open WebUI knowledge collection

Steps

  1. Pull the retrieval model with ollama pull ibm/granite4:latest.

  2. Install and start Open WebUI with the command below and open-webui serve.

    Documented command
    pip install open-webui

  3. In Open WebUI, add a Function and paste granite_autogen_rag.py; save and enable it, then adjust the inference endpoint, search API, and model ID in its settings.

  4. Create a Knowledge collection in Open WebUI and upload documents for the retrieval agent to query.

Expected outcome

A configured retrieval-agent function and a document collection that the workflow can query. Web search requires a separately configured search provider.

Guide limitations

  • SearXNG is optional; web search configuration is a separate step.
  • The README notes possible OpenTelemetry errors for the image-agent script; this guide covers retrieval, not that issue.
  • These are README setup instructions, not a verified execution result.
Documentation 1 ↗
COMPARE OPTIONS

Compare options for Retrieval-augmented applications

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
Granite Retrieval AgentQuery uploaded local documents and combine findings with web retrieval.Linux, macOS, WindowsThe repository is public and identifies its license as Apache-2.0. Its documented setup runs components locally through Ollama and Open WebUI.
VectaurantFind menu options by describing preferences, such as asking what is vegan or spicy.CLI, Docker, Linux, macOS
Aml Multiagent RagIngest PDF regulations, extract and chunk their text, and build a searchable vector collection.Docker, Self-hostedThe repository is public and specifies the MIT license. The documented setup is self-hosted and requires Python 3.11 or later, Docker Desktop and an OpenAI API key.
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
ibm-granite-community/granite-retrieval-agent · Official homepage title · retained in its source language
How we review sourcesSuggest a correction