[PR #536] 🚀 Add AWS Bedrock RAG pipeline for OpenWebUI #460

New issue

Open

opened 2026-10-18 21:06:19 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-18 21:06:19 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/open-webui/pipelines/pull/536
Author: @weisser-dev
Created: 7/14/2025
Status: 🔄 Open

Base: main ← Head: add-bedrock-kb-example

📝 Commits (2)

deff66e Create aws_bedrock_knowledgebase_manifold_pipeline.py
ccc6859 Update aws_bedrock_knowledgebase_manifold_pipeline.py

📊 Changes

1 file changed (+157 additions, -0 deletions)

View changed files

➕ examples/pipelines/providers/aws_bedrock_knowledgebase_manifold_pipeline.py (+157 -0)

📄 Description

Summary
This PR introduces a new manifold pipeline that integrates AWS Bedrock Knowledge Bases into OpenWebUI. It allows users to select from multiple knowledge bases and perform Retrieval-Augmented Generation (RAG) using Amazon Bedrock and large language models such as Claude 3.5 Sonnet.

Features

✅ Manifold pipeline supporting multiple AWS knowledge bases
🔑 Configurable via environment variables:
AWS_BEDROCK_KB_IDS and AWS_BEDROCK_KB_NAMES (semicolon-separated)
AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_REGION
Optional: BEDROCK_RUNTIME_ENDPOINT_URL and BEDROCK_AGENT_RUNTIME_ENDPOINT_URL (for custom VPC access)
📚 Uses bedrock-agent-runtime.retrieve() for vector search
🧠 Sends context + user question to Claude via bedrock-runtime.invoke_model
🧩 Fully integrated with the OpenWebUI valves config (editable per pipeline)

Example Use Case
This pipeline enables RAG workflows similar to those on OpenAI or Azure OpenAI, but using Amazon’s native LLM infrastructure and vector store. It’s particularly useful for:

Organizations using AWS Bedrock Knowledge Bases as their primary RAG backend
Private cloud/VPC setups with custom endpoints
Teams requiring dynamic pipeline selection in OpenWebUI

Notes

No streaming support yet – output is returned as a single final response.
Claude model is used by default (anthropic.claude-3-5-sonnet-20240620-v1:0), but can be changed via valves.

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/open-webui/pipelines/pull/536 **Author:** [@weisser-dev](https://github.com/weisser-dev) **Created:** 7/14/2025 **Status:** 🔄 Open **Base:** `main` ← **Head:** `add-bedrock-kb-example` --- ### 📝 Commits (2) - [`deff66e`](https://github.com/open-webui/pipelines/commit/deff66ef0c94d666e30c71afada0bd2c0fcd3e9d) Create aws_bedrock_knowledgebase_manifold_pipeline.py - [`ccc6859`](https://github.com/open-webui/pipelines/commit/ccc6859fd68b7c6e7d834c2b789a93555b0258f3) Update aws_bedrock_knowledgebase_manifold_pipeline.py ### 📊 Changes **1 file changed** (+157 additions, -0 deletions) <details> <summary>View changed files</summary> ➕ `examples/pipelines/providers/aws_bedrock_knowledgebase_manifold_pipeline.py` (+157 -0) </details> ### 📄 Description **Summary** This PR introduces a new manifold pipeline that integrates AWS Bedrock Knowledge Bases into OpenWebUI. It allows users to select from multiple knowledge bases and perform Retrieval-Augmented Generation (RAG) using Amazon Bedrock and large language models such as Claude 3.5 Sonnet. **Features** - ✅ Manifold pipeline supporting multiple AWS knowledge bases - 🔑 Configurable via environment variables: - AWS_BEDROCK_KB_IDS and AWS_BEDROCK_KB_NAMES (semicolon-separated) - AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_REGION - Optional: BEDROCK_RUNTIME_ENDPOINT_URL and BEDROCK_AGENT_RUNTIME_ENDPOINT_URL (for custom VPC access) - 📚 Uses bedrock-agent-runtime.retrieve() for vector search - 🧠 Sends context + user question to Claude via bedrock-runtime.invoke_model - 🧩 Fully integrated with the OpenWebUI valves config (editable per pipeline) **Example Use Case** This pipeline enables RAG workflows similar to those on OpenAI or Azure OpenAI, but using Amazon’s native LLM infrastructure and vector store. It’s particularly useful for: - Organizations using AWS Bedrock Knowledge Bases as their primary RAG backend - Private cloud/VPC setups with custom endpoints - Teams requiring dynamic pipeline selection in OpenWebUI **Notes** - No streaming support yet – output is returned as a single final response. - Claude model is used by default (anthropic.claude-3-5-sonnet-20240620-v1:0), but can be changed via valves. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi added the

pull-request

label

2026-10-18 21:06:19 +00:00