What's Changed
- Improved AI summary parsing by stripping fenced JSON code blocks before extraction.
- Replaced fragile greedy JSON matching with balanced-brace parsing, so wrapped responses are handled more reliably.
- Increased the AI analysis token budget to reduce failures caused by truncated structured output.
Fixed
- Fixed issue #67 where AI-generated summaries could appear garbled or fall back incorrectly when the model returned markdown-wrapped or incomplete JSON.
PR: #69
Full Changelog: v0.2.8...v0.2.9