WakeIQX

Appearance

reindex_project ​

Backfill Cloudflare Vectorize embeddings for existing contexts in a project.

Overview ​

The reindex_project tool generates and stores vector embeddings for contexts that were saved before semantic search was enabled, or for contexts whose embeddings failed to write. After reindexing, search_context can route queries through Vectorize for semantic (meaning-based) matching rather than falling back to substring search.

Layer: Core / Layer 4 (Infrastructure)

Purpose: Populate missing Vectorize embeddings for a project's existing contexts

Parameters ​

ParameterTypeRequiredDescription
projectstringYesProject identifier to reindex

Returns ​

A summary of the reindex operation:

Reindex complete for "api-service":

Contexts processed: 43
Embeddings written: 41
Skipped (already indexed): 2
Failed: 0

Semantic search is now active for this project.

When to Use ​

After Migrating from an Older Version ​

If you deployed Wake Intelligence before v3.2.0 (when semantic search was introduced), your existing contexts have no embeddings. Run reindex_project to backfill them.

After a Vectorize Outage ​

If Cloudflare Vectorize was unavailable when contexts were saved, those contexts fell back to LIKE-only search. Reindex to restore semantic coverage.

When Search Results Feel Wrong ​

If search_context returns poor matches for a project, reindexing ensures all contexts have fresh, consistent embeddings.

Examples ​

Reindex a Single Project ​

typescript
reindex_project({ project: "authentication-service" })

Reindex After Migration ​

bash
# After upgrading to v3.2.0+, reindex all your projects
typescript
reindex_project({ project: "project-alpha" });
reindex_project({ project: "project-beta" });
reindex_project({ project: "api-service" });

How Semantic Search Works ​

When you call save_context, Wake Intelligence fires a Vectorize write in the background (fire-and-forget). This converts the context's summary into a dense vector embedding using Cloudflare Workers AI.

When you call search_context, the system checks whether Vectorize is available:

  • Vectorize available: semantic embedding similarity search (ANN)
  • Vectorize unavailable or empty: falls back to SQL LIKE '%query%'

reindex_project ensures the Vectorize index is populated for contexts that missed the background write.

Performance ​

  • Processes contexts in batches to stay within Workers AI rate limits
  • Typical time: ~500ms per 10 contexts (embedding generation dominates)
  • Already-indexed contexts are skipped automatically

See Also ​