knowledge-bases
Manage Frontline knowledge bases from the CLI — create/list/update/delete KBs, add and resync sources (URLs and website crawls), inspect crawl sync history, run semantic search, and assign KBs to agen
SKILL.mdskills/vendor/knowledge-bases/SKILL.md
GitHubKnowledge Bases
A knowledge base (KB) is a searchable collection of indexed content. Content comes from sources: URLs and website crawls (file uploads are not available via the CLI). KBs can be shared (assignable to agents) or private. Requires a USER API key (same key as conversations and playbooks).
Output is JSON by default; add --table (or --pretty) for a human-readable view. List
commands (kb list, kb source list, kb sync-history) are paginated — pass --page /
--page-size and read the { results, total, pageNum, pageSize, totalPages } envelope.
Knowledge base CRUD
frontline kb list --table # all KBs (add --shared / --private to filter)
frontline kb list --search-text "docs"
frontline kb create --name "Help Center" --description "Public docs"
frontline kb create --name "Scratch" --private
frontline kb get 12
frontline kb update 12 --name "Help Center v2"
frontline kb delete 12 # deletes the KB and all its sources
Default knowledge bases cannot be deleted.
Sources
Sources live under frontline kb source. List, inspect, add, resync, and remove them.
frontline kb source list 12 --table
frontline kb source get 12 340
# Add URL sources (fetched + indexed asynchronously)
frontline kb source add-url 12 --urls '["https://docs.example.com/faq"]' \
--description "FAQ pages"
# Start a website crawl (discovers + indexes pages under the root URL)
# --max-links is required (1-1000); --max-depth is optional (1-5)
frontline kb source add-crawl 12 --url "https://docs.example.com" \
--max-links 50 --max-depth 3 \
--include-paths '["/docs"]' --exclude-paths '["/blog"]'
# Resync URL/web-page sources by id
frontline kb source resync 12 --source-ids '[340,341]'
# Recrawl a website source (use the website source id, not a source id)
frontline kb source resync 12 --website-id 88
# Remove sources
frontline kb source remove 12 --source-ids '[340]'
resync requires exactly one of --source-ids or --website-id (passing neither errors):
use --source-ids to re-index individual URL/web-page sources, or --website-id to recrawl a
website source.
Filter values
The filter flags expect these exact enum values:
kb source list --source-type:FILE,URL,WEBPAGE,FOLDER(alsoARTICLE,QNA,UNKNOWN)kb source list --status:NONE,PROCESSING,COMPLETED,FAILED,DELETING,DELETED,SYNCING,REJECTEDkb sync-history --status:PROCESSING,SYNCING,COMPLETED,FAILED,DELETING,DELETEDkb sync-history --sync-type:INITIAL_SYNC,MANUAL_RESYNC,AUTO_RESYNC
Website crawl sync history
Each website crawl produces sync summaries (one per crawl run). Look them up by KB id and website source id:
frontline kb sync-history 12 88 --table
frontline kb sync-history 12 88 --status FAILED
Search
Semantic search over a KB's indexed content. Returns chunks with similarity scores. A KB with
no indexed sources yet returns an empty results array.
frontline kb search 12 --query "what is the refund policy" --limit 5
Assigning KBs to agents
Only shared KBs can be assigned. assign replaces the agent's full set of assigned KBs.
frontline kb assigned --agent-id <agentId>
frontline kb assign --agent-id <agentId> --kb-ids '[12,13]'
frontline kb assign --agent-id <agentId> --kb-ids '[]' # unassign all
See also
- Resolving an agent's
<agentId>forassign/assigned: thefrontline-agentsskill. - Authentication and profiles: the
auth-and-profilesskill. - Full reference: https://docs.getfrontline.ai/cli.