Problem Statement

There is a discrepancy between the CLI backfill logic and the live bot pipeline:

• CLI Backfill: Indexes Title + Body + Comments.
• Live Bot: Indexes Title + Body ONLY.

This leads to inconsistent search results depending on how an item was indexed. An issue backfilled via CLI will have richer context than one processed by the bot in real-time.

Proposed Solution

1. Create a shared function (e.g., in internal/utils/text) to generate the content string for embedding.
2. Refactor internal/steps/indexer.go (Bot) to use this function.
3. Refactor cmd/simili/commands/index.go (CLI) to use this function.
4. The shared function MUST include comments for both Issues and PRs to ensure rich context for similarity search.

Feature Scope

• Pipeline Steps
• CLI