feat(ai): Implement multi-call AI grouping and summarization #56

2025-09-05T19:10:34Z

eleith commented

2025-09-05 19:10:34 +00:00

This commit enhances the AI view to be more reliable and performant.

The AI digest generation now makes multiple calls to the LLM to create its own primary and sub-groups, replacing the previous single-call, category-based approach. The overview summary has been removed in favor of per-group summaries to manage context length. To improve reliability, the LLM calls for summarization and sub-grouping have been separated, and the timeout handling has been refactored to use a per-try timeout, making it more resilient to transient server errors.

The email template has also been updated to include the primary group summaries, aligning its purpose with the overview HTML page.

This commit enhances the AI view to be more reliable and performant. The AI digest generation now makes multiple calls to the LLM to create its own primary and sub-groups, replacing the previous single-call, category-based approach. The overview summary has been removed in favor of per-group summaries to manage context length. To improve reliability, the LLM calls for summarization and sub-grouping have been separated, and the timeout handling has been refactored to use a per-try timeout, making it more resilient to transient server errors. The email template has also been updated to include the primary group summaries, aligning its purpose with the overview HTML page.

eleith added 1 commit 2025-09-05 19:10:34 +00:00

feat(ai): Implement multi-call AI grouping and summarization 306b6f5d9d

This commit enhances the AI view to be more reliable and performant.

The AI digest generation now makes multiple calls to the LLM to create its own primary and sub-groups, replacing the previous single-call, category-based approach. The overview summary has been removed in favor of per-group summaries to manage context length. To improve reliability, the LLM calls for summarization and sub-grouping have been separated, and the timeout handling has been refactored to use a per-try timeout, making it more resilient to transient server errors.

The email template has also been updated to include the primary group summaries, aligning its purpose with the overview HTML page.

eleith merged commit 0692e72056 into main

2025-09-05 19:11:22 +00:00

eleith deleted branch gemini/feat/llm-dynamic-grouping-additive

2025-09-05 19:11:22 +00:00

eleith referenced this issue from a commit

2025-09-05 19:11:24 +00:00

feat(ai): Implement multi-call AI grouping and summarization (#56)

eleith referenced this issue from a commit

2025-12-10 22:24:46 +00:00

feat(ai): Implement multi-call AI grouping and summarization (#56)

Sign in to join this conversation.