Fix parseReview validLine logic to correctly parse AI findings

fix: make AI review parser regex robust against brackets and refine prompts

feat: implement autonomous Auto-PR and optimize Docker image

Upload 169 files

Upload 29 files

Delete src

Upload 52 files

Upload 94 files

Delete src

Upload 116 files

Upload 172 files

old app

Upload 46 files

Upload 46 files

Upload 46 files

Upload 46 files

Upload 169 files

Upload 139 files (#2)

fix: update models to Llama Scout for summary and Llama 3 70B for review, plus fixed line number parsing for reviews

fix: restore Remedy PR generation logic and fix completion summary

feat: separate schedulers for summary (30k) and review (11.5k)

feat: increase Llama-4 TPM limit to 30k

feat: upgrade summary model to Llama-4-Scout-17B

fix: optimize prompt floor and fix summary scheduler bug

fix: sync scheduler limit with Groq TPM and improve review parsing

fix: restore missing extractPatternType export for HF build

feat: hunk-based reviewing and priority queue for massive files

fix(octokit): add .rest prefix to pulls.listCommits and pulls.create to fix v14/v21 compatibility

fix(auth): change auth type from 'token' to 'installation' for Probot compatibility

fix(probot): use modern octokit.auth() for checkout

chore: add NPM overrides to fix deprecated transitive dependencies and finalize Octokit v21 migration

fix: update tree-sitter parser source and improve initialization robustness

fix: implement staggered throttling and sequential file processing for free tier

debug: add detailed error logging to AI attempts

fix: enable gh auth, strengthen audit intensity and enable remedy PRs

fix: implement robust token optimization and gh cli support for Groq/HF

fix: ensure git version check uses valid directory to avoid ENOENT

chore: optimize temp storage, caching and file retrieval for HF hosting

Upload 117 files