What is the Context Engine?
Not your grandma’s RAG. TierZero Context Engine is a hybrid knowledge graph that captures investigation patterns from your operational history. Tribal knowledge gets captured automatically from Slack, incidents, and code changes. Structured, searchable, and editable, it is an AI agent you can actually inspect.Key Capabilities
Ingestion
Multi-source capture from incidents, Slack threads, code reviews, and post-mortems, with no forms or tagging needed.- Every memory is a structured record with type, source, confidence score, tags, linked services, and full version history
- Reasoning traces capture investigation paths: branches taken, ruled out, root cause found
- Service dependency mapping across your infrastructure
- Team language mapping: learns your internal names, abbreviations, and slang
Retrieval
Three retrieval strategies run in parallel with reciprocal rank fusion for best results.- Hybrid search: Vector embeddings, keyword search, and summary indexes combined
- Graph traversal: Results expand through linked incidents, services, teams, and runbooks
- Trajectory replay: Past investigation paths that match the current failure pattern are recalled and adapted to the current context
Adaptation
This is not RAG and not fine-tuning. TierZero extracts investigation patterns from your operational history and replays them in-context.- Capture investigation playbooks: Every investigation records the full reasoning path: queries run, hypotheses tested, dead ends ruled out, root cause confirmed.
- Track what actually helped: Memory attribution links each outcome to the specific memories that influenced it.
- Incorporate engineer feedback: Corrections, reactions, and direct edits take effect immediately.
- Replay and adapt patterns: When a new incident matches a past failure pattern, the agent replays the successful investigation trajectory adapted to the current context.
Long-Term Tribal Knowledge
- Searchable: Ranked by relevance, recency, and confidence
- Editable: Every record can be inspected, updated, or deleted
- Survives team turnover: Knowledge persists even as people leave
- Version history & audit trail: Full history of all changes
Trust & Transparency
- Evidence chains: Every finding cites specific sources: Datadog logs, GitHub commits, past incidents
- Full version history: Complete audit trail of all organizational memory changes
- SOC 2 Type II certified
Performance vs RAG
+36.8% Recall
97.3% vs 71.1%
+121.3% Precision
89.0% vs 40.2%
+58.6% Accuracy
93.3% vs 58.8%
How to Use
Snippets
Navigate to Context Engine → Snippets to browse, search, add, and edit organizational memories.- Search: Query your knowledge base with natural language
- Filter by source: Dashboard, Document, UI, Slack, Discord, Incident Channel, Alert Thread, Chrome Extension
- Filter by type: Fact, Best Practice, Query, Resource
- Filter by modifier: See memories you’ve created or edited
- Add memories: Manually add knowledge that TierZero should remember
- Edit or delete: Correct inaccurate memories immediately
Resources
Navigate to Context Engine → Resources to browse integrated catalogs. Resources are structured metadata from your connected integrations:- Datadog: Dashboards, notebooks, logs/spans/events configuration
- New Relic: Dashboards, applications, logs/spans configuration
- Grafana Cloud: Dashboards
- GCP Cloud Monitoring: Dashboards
- Notion: Ingested pages
- Confluence: Ingested documents
Best Practices
1. Review and Curate Memories Regularly- Periodically browse recent memories to ensure accuracy
- Correct or delete any inaccurate memories immediately, as they affect future investigations
- Thumbs up/down reactions on TierZero responses improve memory quality
- Direct corrections take effect immediately
- Ingest your most important Datadog dashboards, Notion pages, and Confluence documents
- These give TierZero richer context for investigations
- When TierZero gets something wrong, edit the memory directly rather than hoping it learns
- Every edit has an immediate effect and is tracked in version history