REFACTOR: Assign type: 'blog' directly in createKnowledgeBase.mjs

This ticket is a follow-up to PR #7472, which fixed a bug where blog posts were not being found by the AI query system. The PR implemented a workaround on the query-side (queryKnowledgeBase.mjs) to correctly find blog posts.

This ticket will address the root cause of the issue by refactoring the data creation script (buildScripts/ai/createKnowledgeBase.mjs) to assign the correct type to blog posts during the initial indexing process.

Problem

Currently, createKnowledgeBase.mjs processes all markdown files from learn/tree.json and assigns them type: 'guide'. To differentiate blog posts, it adds a flag isBlog: true. This forces the query logic to contain a special condition to translate a search for type: 'blog' into a search for {type: 'guide', isBlog: 'true'}.

This approach is not ideal because:

1. It makes the data model less clear. Blog posts are a distinct content type, not a subtype of a guide.
2. It adds unnecessary complexity to the query logic, making it harder to maintain.

Proposed Solution

The solution is to modify the data creation script to assign the correct type at the source.

1. In buildScripts/ai/createKnowledgeBase.mjs:

   • Locate the loop that processes the learnTree data.
   • Inside the loop, add a condition to check if an item's parentId is 'Blog'.
   • If it is, set the type of the generated chunk to 'blog'.
   • Remove the now-redundant isBlog property from the chunk.

2. In buildScripts/ai/queryKnowledgeBase.mjs:

   • Remove the special if (type === 'blog') condition that was added in PR #7472.
   • The generic whereClause = {type: type} will now work correctly for blog posts as well.

Acceptance Criteria

1. createKnowledgeBase.mjs is updated to generate chunks with type: 'blog' for all blog posts.
2. The isBlog property is no longer present in the generated knowledge base chunks.
3. queryKnowledgeBase.mjs is simplified, and the workaround for blog queries is removed.
4. After rebuilding the knowledge base, running npm run ai:query -- -q "<search-term>" -t blog successfully returns results from blog posts.