Knowledge Studio
Studio Overview
Knowledge Studio is the AI-powered authoring environment within StackFlow. It leverages Amazon Bedrock's Claude models to assist knowledge managers in creating, improving, and maintaining KB articles at scale. Studio is accessible at Knowledge → Knowledge Studio and requires the kb_manager or super_admin role.
- Bedrock Model:
anthropic.claude-3-5-sonnet-20241022-v2:0enabled in Bedrock console for account373544523367 - DynamoDB:
StackFlow_AIExemplartable with exemplar articles for few-shot prompting - Feature Flag:
knowledge_studio_aienabled inStackFlow_FeatureFlagfor the tenant - IAM:
StackFlowAPIRolewithbedrock:InvokeModelon the Claude model ARN
AI Article Generation
Generate new KB articles directly from incident resolutions, change post-implementation reviews, or free-form topics. The generation engine queries the Bedrock KB for related existing articles to avoid duplication and ensure consistency with your organization's existing documentation style.
curl -X POST https://your-instance.stackflow-tech.com/prod/api/knowledge/studio/generate -H "Authorization: Bearer $TOKEN" -H "Content-Type: application/json" -d '{
"source": "incident",
"source_id": "INC0001234",
"article_type": "troubleshooting",
"target_audience": "agents",
"include_resolution": true
}'Article Improvement
The Article Improvement tool analyzes existing articles for readability, completeness, accuracy, and SEO. It provides specific suggestions for improvement and can automatically apply improvements in "suggest mode" (tracked changes style) or "auto mode" (applies changes, creates new draft version). Common improvements include: adding missing steps, fixing broken links, updating outdated screenshots, and improving title clarity for better search visibility.
Bulk Operations
Knowledge Studio supports bulk operations on article sets: bulk category reassignment, bulk access control changes, bulk expiry setting, and bulk AI review for outdated content. Bulk operations are processed asynchronously and the status can be tracked in the Studio job queue.
| Bulk Operation | Use Case | Max Batch Size |
|---|---|---|
| Bulk Categorize | Reorganizing article taxonomy | 500 articles |
| Bulk Retire | Retiring outdated content after audit | 200 articles |
| Bulk AI Review | Quality check of existing articles | 100 articles |
| Bulk Re-index | Force Bedrock re-indexing after content changes | 1000 articles |
Quality Scoring
Every article receives a Quality Score (0-100) calculated from: readability (Flesch-Kincaid grade level), completeness (does it have a title, body, category, tags?), freshness (days since last review), engagement (views, useful ratings), and semantic uniqueness (how different is it from other articles?). Target a quality score above 70 for all published articles.