Skip to content

Evidence and knowledge assets

Keep knowledge current and traceable so workflow decisions remain grounded and auditable.

Threada Admin evidence page showing knowledge assets with freshness and source attributionThreada Admin security policy page showing work-item governance rules and approval thresholds

Knowledge assets

  • Register websites, documents, and connected sources as managed assets
  • Track asset ownership and workflow usage
  • Apply source-specific freshness and retention policy

Ingestion pipelines

  • Handle website sync, uploads, connector sync, and feed ingestion
  • Monitor pipeline health and last successful sync state
  • Preserve ingestion lineage for troubleshooting and audits

Freshness policy

  • Set sync cadence and stale thresholds
  • Review diffs before policy-critical updates
  • Alert on freshness drift and source failures

Evidence quality controls

  • Require citations for selected workflows and intents
  • Route low-evidence responses to review or escalation
  • Track retrieval diagnostics and missing-source patterns
  • Tie response outcomes to source freshness and ingestion health

Inspect the evidence layer

The evidence page should not stop at saying citations exist. These samples show what a bundle and release gate can look like.

Sample evidence

Evidence bundle JSON

A minimal cited-evidence bundle with source URLs, snippets, retrieval trace, and fallback reason.

{
  "citations": [
    {
      "url": "https://docs.example.com/security/exampleco-scope",
      "title": "ExampleCo security scope note (illustrative)"
    }
  ],
  "retrieval_trace_id": "rt_8c2147f0",
  "fallback_reason": ""
}
generation_mode: illustrative_sample · provenance_citation: docs/spec/product.md#design-principles · Synthetic marketing sample; not tenant data · Replace example IDs, domains, roles, and thresholds before use.
Download evidence bundle
Release check

Evaluation gate rubric

A simple fixture-driven rubric for testing extraction, grounding, approval routing, action safety, and receipt completeness before release.

gate_id,gate_name,expected_behavior,pass_threshold
extract-001,Extract required fields,"required fields are present",0.95
ground-001,Citation grounding,"answer cites current sources",0.98
audit-001,Receipt completeness,"event links actor, WorkItem, evidence, and policy",1.00
generation_mode: template_sample · provenance_citation: docs/spec/product.md#design-principles · Template sample; not tenant data · Replace example IDs, domains, roles, and thresholds before use.
Download eval rubric
FAQ

FAQ

What content sources does Threada support?
Websites, uploaded documents, API connectors, and feed-based sources. Each is registered as a managed knowledge asset with its own freshness and retention policy.
How does freshness policy work?
You set sync cadence and stale thresholds per source. Threada alerts on freshness drift and failed syncs so evidence stays current.
Can we require citations in responses?
Yes. Citation requirements can be applied per workflow and intent. Low-evidence responses are routed to review or escalation automatically.
Open a knowledge setup review
Security and SLA Data and governance