Summary
We are requesting admin‑enforceable naming conventions for Glean Agents to ensure consistent, readable, and searchable agent names across the organization. This capability would make it easier for users to discover the right agent, reduce duplication, and improve catalog hygiene at scale.
Problem / Justification
Our current agent ecosystem has grown organically, leading to inconsistent agent names (varying prefixes, case styles, and structures). This creates several issues:
- Poor discoverability: Users struggle to find the correct agent when names don’t follow predictable patterns, reducing search precision and confidence.
- Duplicate/near‑duplicate creation: Teams unknowingly create new agents for existing use cases because names don’t clearly encode ownership, scope, or audience.
- Governance and ownership ambiguity: Without enforced patterns, it’s hard to quickly tell which team owns an agent, whether it’s departmental vs. company‑wide, or whether it’s production‑ready vs. experimental.
- Catalog hygiene and support burden: Inconsistent names make lifecycle management, audits, and support triage harder, increasing maintenance cost.
Proposed Solution
Enable a few 'fields' in agent creation that must be populated, and allow Glean admins to control settings around these such as:
- Creator
- Department, or Target audience
- Agent Name
Benefits
- Improved discoverability: Predictable names and facets make search and browse faster and more accurate, reducing time‑to‑answer.
- Reduced duplication: Clear, encoded scope and ownership lower the likelihood of redundant agents.
- Governance & trust: Consistent naming signals maturity and ownership, aiding audits, lifecycle management, and support.
- Smoother onboarding: New teams and users learn patterns quickly; suggestions and templates lower friction.
Relevant Context
- Current behavior: Agent names are free‑form; no enforceable pattern, limited guidance, and no bulk remediation tools.
- Desired behavior: Admins define and enforce naming policies with templates/regex, provide real‑time guidance, remediate at scale, and integrate naming metadata into search.
Acceptance Criteria
- Admins can create, test, and enable a policy that blocks non‑compliant names at create/edit time with clear guidance.
- Users receive auto‑suggested compliant names based on selected metadata (e.g., Dept, Audience).
