fraujulian/MyAiConfiguration
1
0 Fork 0
Code Issues Pull Requests Actions Activity
233 Commits 1 Branch 0 Tags
208213c1c7e9cbefa980afdd48a057d252c73804
Commit Graph
14 Commits
Author SHA1 Message Date
Julian lechnerandClaude Sonnet 5 208213c1c7 feat(retrieval): pin model revisions and share encoder/reranker instances 
Pin the Qwen embedding/reranker models to specific commit revisions (and
sentence-transformers to the exact version already in use) so index
builds are reproducible instead of floating on each model's main branch.

Also move the encoder and reranker instances to module-level singletons
shared by every Index, instead of one pair per instance, so a process
serving multiple repository roots only loads each model once.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
 2026-09-18 09:35:46 +02:00
Julian lechnerandClaude Sonnet 5 96a1175e20 feat(retrieval): make the file-scan limit configurable and warn on it 
Repositories with more than 2000 indexable files silently lost the
remainder during indexing. Make the cap configurable via
SEMANTIC_RETRIEVAL_MAX_FILES (default unchanged) and print an explicit
stderr warning naming the repository when it is reached, instead of
truncating without any signal.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
 2026-09-18 09:35:38 +02:00
fraujulian 9b312cd840 fix(retrieval): run search as local CLI
CI / linux (push) Successful in 6s
Details
2026-09-17 19:27:11 +02:00
Julian lechner e5b63965b8 feat(retrieval): prefer CUDA devices 2026-09-17 17:50:33 +02:00
Julian lechner 5ec6d0b6ac perf(retrieval): batch changed embeddings 2026-09-17 17:50:14 +02:00
Julian lechner 0f8aa0a25d feat(qwen): rerank 50 candidates and return five 2026-09-17 15:54:23 +02:00
Julian lechner e97404a0a1 feat(qwen): fuse dense and lexical search with RRF 2026-09-17 15:54:10 +02:00
Julian lechner fd37422de2 feat(qwen): maintain a lexical search index 2026-09-17 15:53:59 +02:00
Julian lechner b76a993c75 feat(qwen): split chunks at symbols and sections 2026-09-17 15:53:48 +02:00
Julian lechner 685dbe61f4 feat(qwen): add token windows and document metadata 2026-09-17 15:53:36 +02:00
Julian lechner 7be729cb22 feat(qwen): enforce 1024-dimensional embeddings 2026-09-17 15:52:43 +02:00
Julian lechner 2079792a74 fix(qwen): isolate and refresh indexes with code-search prompts 2026-09-17 15:27:04 +02:00
Julian lechner 727c499d2c feat(retrieval): rerank semantic search results 2026-09-17 13:07:47 +02:00
Julian lechner 27f1c1ec79 feat(retrieval): add optional Qwen semantic layer 2026-09-17 09:16:53 +02:00