models-v1

What's Changed

• feat: Phase 1 — Code Search End-to-End by @tomharris in #1
• feat: Phase 2 — PR Knowledge by @tomharris in #2
• feat: Phase 3 — Document Search + CLI by @tomharris in #3
• feat: Phase 4 — Polish, Scale, and Eval by @tomharris in #4
• docs: add README.md and CLAUDE.md by @tomharris in #5
• fix: pass CodeConfig instead of DevragConfig to CodeIndexer by @tomharris in #6
• fix: convert chunk IDs to deterministic UUIDs for Qdrant by @tomharris in #7
• fix: prevent empty and oversized texts from failing Ollama embed API by @tomharris in #8
• fix: prevent empty files from producing zero-dimension vectors in Qdrant by @tomharris in #9
• fix: migrate Qdrant store to query_points API and suppress reranker warnings by @tomharris in #10
• fix: truncate oversized text before embedding to prevent Ollama 400 errors by @tomharris in #11
• docs: update CLAUDE.md and README.md with recent work by @tomharris in #12
• feat: add rag-first skill and hookify rule for RAG-before-explore workflow by @tomharris in #13
• feat: add GitHub issues as a knowledge source by @tomharris in #14
• feat: add Jira Cloud tickets as a knowledge source by @tomharris in #15
• feat: add Slite pages as a knowledge source by @tomharris in #16
• fix: migrate Jira search to /rest/api/3/search/jql by @tomharris in #17
• feat: support multi-repo indexing by @tomharris in #18
• fix: skip transient Slite API errors instead of crashing sync by @tomharris in #19
• fix: handle connection timeouts in Slite sync by @tomharris in #20
• fix: resolve SQLite threading error in MCP server by @tomharris in #21
• fix: resolve BM25 silent discard bug and improve search quality by @tomharris in #22
• feat: make devrag reindex --all properly reset everything by @tomharris in #23
• fix: resolve DuplicateIDError in reindex --all for JS/TS repos by @tomharris in #24
• fix: report files that produce no chunks during indexing by @tomharris in #25
• feat: add reindex --name for single-repo re-indexing by @tomharris in #26
• refactor: migrate RAG skills to slash commands by @tomharris in #27
• fix: add text truncation to Slite indexer before embedding by @tomharris in #28
• docs: install instructions use GitHub source URLs by @tomharris in #29
• refactor: make qdrant-client a required dependency by @tomharris in #30
• fix: make Slite indexing truly incremental by @tomharris in #31
• fix: normalize Jira cursor to JQL datetime format by @tomharris in #32
• fix: make PR sync truly incremental and honor --since override by @tomharris in #33
• Add rag-first reminder hook for Explore/general-purpose subagents by @tomharris in #34
• Index Claude Code session logs and add metadata filters to search by @tomharris in #35
• Standardize on Qdrant, remove ChromaDB backend by @tomharris in #36
• perf: move BM25 into Qdrant sparse vectors + server-side RRF fusion by @tomharris in #37
• perf: drop VectorStore Protocol + quantization/wait/parallel follow-ups by @tomharris in #38
• chore: bump devrag to 0.2.0 by @tomharris in #39
• fix: pin python <3.14 to avoid py-rust-stemmers SIGSEGV by @tomharris in #40
• fix: batch Qdrant upserts to stay under 32 MiB HTTP limit by @tomharris in #41
• feat: index Terraform .tf/.tfvars files with per-block chunking by @tomharris in #42
• chore: replace rag-first warnings with per-turn block gate by @tomharris in #43
• Improve retrieval quality: dedup, routing, reranker, repo bias by @tomharris in #44
• feat: index Slack conversations without a workspace app by @tomharris in #45
• feat: add devrag auth slack to auto-obtain session credentials by @tomharris in #46
• Make browser-cookie3 a default dependency by @tomharris in #47
• fix: don't let a broken browser loader abort devrag auth slack cookie sweep by @tomharris in #48
• feat: read Slack d cookie from the desktop app by default by @tomharris in #49
• feat: index repo docs alongside code on index repo by @tomharris in #50
• feat: add index refresh to re-index all repos incrementally by @tomharris in #51
• fix: avoid embed 400 and isolate per-file indexing failures by @tomharris in #52
• Fix .devragignore gitignore semantics + de-rot sync tests by @tomharris in #53
• fix: trust corporate proxy CA via configurable bundle for httpx clients by @tomharris in #54
• fix: thread corporate CA bundle config through slack auth functions by @tomharris in #55
• fix: relax X509_STRICT for custom CA bundles so corporate proxies verify by @tomharris in #56
• fix: trust certifi public roots alongside the corporate CA bundle by @tomharris in #57
• fix: read Slack xoxc token from desktop localStorage (HTTP scrape is obsolete) by @tomharris in #58
• fix: read Slack d-cookie keychain key under correct "Slack Key" account by @tomharris in #59
• perf: optimize indexing with batched commits and parallel Slack fetching by @tomharris in #60
• fix: throttle and back off Slack web-API calls to avoid 429s by @tomharris in #61
• fix: load reranker and BM25 models from local HF cache, offline-first by @tomharris in #62

New Contributors

• @tomharris made their first contribution in #1

Full Changelog: https://github.com/tomharris/dev-rag/commits/models-v1