Glean Core Workflow A: Search & Chat

Overview

Build search and chat experiences using the Glean Client API. Covers full-text search with filters, AI-powered chat answers, and autocomplete suggestions.

Prerequisites

• A scoped search identity, approved datasource filter, and synthetic terms that cannot retrieve company-sensitive material.
• User-consent and retention policy for any analytics, chat history, or feedback capture.
• A rollback path that disables the client or filter without changing source documents or connector ACLs.

Instructions

Step 1: Search with Filters and Facets

const results = await fetch(`${GLEAN}/client/v1/search`, {
  method: 'POST', headers: searchHeaders,
  body: JSON.stringify({
    query: 'kubernetes deployment best practices',
    pageSize: 20,
    requestOptions: {
      datasourceFilter: 'confluence,github',
      facetFilters: [{ fieldName: 'author', values: ['engineering-team'] }],
    },
  }),
}).then(r => r.json());

results.results?.forEach((r: any) => {
  console.log(`[${r.datasource}] ${r.title}`);
  console.log(`  ${r.snippets?.[0]?.snippet ?? ''}`);
});

Step 2: AI Chat (Glean Assistant)

const chatResponse = await fetch(`${GLEAN}/client/v1/chat`, {
  method: 'POST', headers: searchHeaders,
  body: JSON.stringify({
    messages: [{ role: 'USER', content: 'What is our deployment process for production?' }],
    applicationId: 'my-app',
  }),
}).then(r => r.json());

console.log('Answer:', chatResponse.messages?.[0]?.content);
console.log('Sources:', chatResponse.citations?.map((c: any) => c.title).join(', '));

Step 3: Autocomplete / Suggestions

const suggestions = await fetch(`${GLEAN}/client/v1/autocomplete`, {
  method: 'POST', headers: searchHeaders,
  body: JSON.stringify({ query: 'deploy', datasourceFilter: 'confluence' }),
}).then(r => r.json());

suggestions.results?.forEach((s: any) => console.log(`  ${s.text}`));

Error Handling

Error	Cause	Solution
Empty results	Query too specific or datasource not indexed	Broaden query, check datasource status
Chat returns no citations	Content not indexed for chat	Verify documents have body text
403 on search	User permissions	Ensure token has search scope

Output

Return a redacted workflow receipt containing datasource scope, correlation ID, result-count band, allow/deny outcomes, and fallback used. Never record query text, titles, snippets, transcripts, or credentials.

Examples

Run a fictional query against sandbox-handbook, verify one authorized identity sees the sample while a denied identity sees none, and record scope=sandbox-handbook; allow=1; deny=0; fallback=none.

Resources

• Search API
• Chat API

Next Steps

For bulk indexing workflow, see glean-core-workflow-b.