Skip to content

RAG Context (Knowledge Base)

Keel RAG Context — index runbooks and post-mortems to ground AI answers

Location in Keel: /rag-context

RAG Context is the knowledge base Keel's AI reasons from. Index your runbooks, post-mortems, and architecture docs, and Keel grounds its root-cause analysis and assistant answers in your own procedures — not just generic knowledge.

Adding documents

Click Add Document to bring in operational content. Each document is indexed into chunks and embedded as vectors for semantic search, so the AI can retrieve the passages relevant to a question or incident. The header tracks totals at a glance: Documents, Indexed, Chunks, and Indexed vectors.

Documents carry a type so you can organise and filter your knowledge base:

Type Typical use
Runbook Step-by-step operational procedures
Post-mortem Write-ups of past incidents
Architecture How systems and dependencies fit together
SOP Standard operating procedures
Config Reference configuration
Alert Rule Documentation for alert rules
Other Anything else

The page has three views: Documents (the list, filterable by type and by status), Collections (grouping related documents), and Semantic Search (query the indexed content directly to see what the AI would retrieve). Each document shows its status — Indexed, Pending, Indexing, or Error — and its chunk count.

Keep runbooks current

RCA quality depends on what's indexed. Adding a runbook for a service means the next analysis of that service can cite your documented cause and remediation instead of reasoning from logs alone.

Stays on the appliance

Documents you upload are indexed and stored on the appliance. As with any AI request, only the context needed to answer a triggered question is sent to the model endpoint you configured — and with a local model, nothing leaves your network.