Pipeline Fix Assistant

It is an Azure DevOps task extension that automatically analyzes failed pipelines and suggests actionable fixes. It uses Azure OpenAI to process build/release logs, generate a root cause analysis (RCA), and recommend practical remediations.

Key Features

1. Automated Root Cause Analysis (RCA): Detects why your pipeline failed.
2. Fix Suggestions: Provides clear, actionable steps to resolve issues.
3. Secure: Uses Azure OpenAI with secrets stored safely in Azure DevOps.

Input Parameters

When you add the Pipeline Fix Assistant task, you need to provide some inputs.

1.azureOpenAiEndpoint (required) The endpoint of your Azure OpenAI resource. Example: https://your-resource.openai.azure.com/

2.azureOpenAiApiVersion (required) The API version of Azure OpenAI that you want to use. Example: 2024-06-01

3.deploymentName (required) The name of the deployed model (for example, GPT-4.1 or GPT-4o). Example: gpt-4.1

4.azureOpenAiApiKey (required) Your Azure OpenAI API key. This should always be passed securely as a secret variable in your pipeline. Example: $(AZURE_OPENAI_API_KEY)

5.patToken (required) A Personal Access Token (PAT) for Azure DevOps. This is needed so the task can securely fetch pipeline logs for analysis. Example: $(ADO_PAT)

6.maxChars (optional) The maximum number of log characters to send for analysis. The default is 40,000. Increase if your logs are larger. Example: 50000

7.maxTokens (optional) The maximum number of tokens (words/units of text) the AI can return in its answer. The default is 1200. Example: 1500

Installation

1.Install the Pipeline Fix Assistant from the Azure DevOps Marketplace into your organization. 2.Ensure you have an Azure OpenAI deployment and a valid API key. 3.Create an Azure DevOps PAT with scope: Build (Read) to access pipeline logs.

Usage

Search for the custom task named "Pipeline Fix Assistant". Sample:

• task: pipelinefix@1 condition: failed() inputs: azureOpenAiEndpoint: 'https://migration-planner1.cognitiveservices.azure.com/' azureOpenAiApiVersion: '2024-12-01-preview' deploymentName: 'gpt-4.1' azureOpenAiApiKey: $(AZURE_OPENAI_API_KEY) patToken: $(ADO_PAT) maxChars: '40000' maxTokens: '1200' temperature: '0.2' outputFormat: 'markdown'