v0.1.0

Changelog

• 407918d Pin backend dependencies for reproducible installs
• f3198ad Consolidate agent onto single LangGraph runtime
• 9fc2633 Polish project report layout and wording
• 0f4122f Sync project report with TUI workflow updates
• 22333ac Polish TUI answer format and add golden session
• 157d394 Polish TUI workflow timeline
• 8b68cdf Document delivery roadmap and polish agent phases
• ddf5b4c Polish TUI chat flow and trend answers
• 7a0e493 Fix TUI first-turn streaming and answer layout
• ae09689 Polish TUI recovery actions
• 9d93fb6 Add TUI timeline command
• 9e1ac73 Default frontend dev server to port 3001
• 6b127db Polish TUI command palette and composer
• 13fb6ad Show LangGraph stream state in TUI
• 97662f4 Update deliverables for LangGraph runtime
• e401905 Route TUI retry through LangGraph sessions
• 31a5149 Add LangGraph session metadata to agent stream
• 2ba8a81 Handle agent stream validation errors
• ec4dacc Route legacy agent loop through shared primitives
• c5b3a23 Extract agent primitives for LangGraph runtime
• 0eb8077 Make LangGraph the default agent runtime
• 2ee684a Split LangGraph agent runtime into orchestration nodes
• 6464306 Add selectable LangGraph agent runtime
• daef125 Polish SciScope report evidence framing
• 2311916 Polish TUI opening experience
• 5693dcb Add project report figure pack
• 7083e42 Polish TUI product workflow
• dd428ca Document SciScope handoff boundaries
• e1670a0 Document TUI Homebrew release path
• db9da2c Prepare TUI CLI for packaging
• 400025c Add TUI ASCII brand and trace panels
• 79c89f4 Show recent sessions on TUI splash
• f892b99 Refine TUI splash dashboard
• 284ccd2 Polish TUI splash and composer
• 53140bd Add offline TUI demo flow
• b217eb3 Unify TUI panel row styling
• 8466079 Add TUI session resume flow
• af3e358 Add TUI retry recovery flow
• 825efb4 Add TUI tool timeline and permission notices
• 5147ac5 Productize SciScope TUI sessions
• 0efb743 Polish final submission reports
• f269849 docs(report): render project structure as a real dirtree (connector tree)
• 693f7c7 docs: rewrite project_structure.md as an annotated six-layer tree
• bcf8ba4 build(release): Homebrew distribution for the Go client via GoReleaser
• 2eb0eac feat(tui): SCISCOPE_TUI_ICONS=off plain-text fallback (no Nerd Font needed)
• 2fd9951 docs(index): add project layering tree to 交付说明.md (visible entry point)
• 0d6e3fb feat(agent): structured-Markdown answer prompt (clear hierarchy)
• e4a2910 feat(tui): Nerd Font tool glyphs + elapsed-time thinking spinner
• 063afe0 docs(report): add project structure & six-layer architecture section
• e845182 fix(agent): firmer reflect-retry directive (no meta 'good, I will…' answers)
• c6536bf fix(tui): use fixed dark glamour style (stop OSC 11 background-query leak)
• 32beda4 fix(backend): export SCISCOPE_DB_DSN + SCISCOPE_EMBEDDER_PATH so make backend/dev serve the real corpus
• 6f4fd7d feat(tui): Claude Code visual grammar — verb spinner, ⏺/⎿ markers, esc-interrupt
• 350c0a1 feat(tui): polished rendering — glamour Markdown + bordered banner & answer panel
• 6b83b55 refactor(repo): file governance — only-Go agent, tidy root, graphs->output/graphs
• 1188213 feat(tui): Go + Bubble Tea terminal client over SSE (make tui)
• 7d21bc8 feat(cli): slash-command inline completion list (à la Claude Code)
• 1fa133e docs(report): surface the agentic orchestration layer (project report)
• 87234cc fix(agent): guard against repeat-call spin + fabricated paper_id
• b2a1281 docs(report): economic quantification + data-source/references sections
• 21a27d2 feat(agent): explicit plan step + model self-critique + verify_claim + retrieval contract
• 906d252 fix(tests): make evidence_chat/api tests hermetic (force in-memory corpus)
• edc1d36 feat(agent-tui): Claude Code-style redesign — banner, status line, slash commands
• b0baea7 docs(report): sync project report numbers with rebuild
• 6116ff5 build: fold arxiv fulltext into retrieval (367k embeddings) + refresh data report
• e256667 refactor(eval): move eval suite src/models -> evaluation/ (P4, fix inverted coupling)
• fd031de feat(agent): action tools — summarize_field / compare_papers / export_bibliography (P3)
• cfcf90e feat(agent): plan->reflect->retry self-correction loop (P2)
• 0ae2067 fix(data+agent): dedupe 2856 residual duplicates + clarify snippet vs author (P1)
• 61aaa40 feat(agent): themed rich TUI — panels, per-tool icons, spinner, markdown (batch C)
• 4712285 feat(agent): SSE server endpoint /api/agent/stream (batch B, server/client split)
• 5eedbf9 feat(agent): get_paper tool + lightweight context compaction (batch A)
• e07e9a8 feat(agent): streaming loop + rich TUI, grounded in OpenCode/Claude Code source
• 0ac96e1 feat(agent): agentic terminal assistant (tools + loop + CLI)
• 8704415 feat(llm): enable vLLM tool-calling (hermes parser) for agent CLI
• 6aa514d feat(graph): interpretable research-community summaries (opt #17)
• ec737a0 feat(recommend): MMR diversity in recommendation selection (opt #16)
• d872270 feat(trends): Mann-Kendall trend test + Sen's robust slope (opt #2)
• 9a539e0 feat(agent): multi-hop query decomposition for comparative/temporal QA (opt #1)
• 333fdda feat(agent): default local LLM to Qwen2.5-7B-Instruct-4bit (opt #15)
• 3f452dc feat(retrieval): cross-encoder reranker bge-reranker-v2-m3 (opt #14)
• 2596870 docs: add 交付说明.md (judge-facing deliverables index + reproduce steps)
• 992c3b4 docs(report): refresh project report with post-rebuild numbers + relevance optimization
• 2025be9 build: fold backfilled abstracts (+9170) & arxiv fulltext (+3349) into data report
• 02328ef fix(build): TRUNCATE paper_chunks before reload in full-rebuild (opt #13b)
• cbdb9e7 build: full-rebuild target to fold enriched data into report+retrieval (opt #13)
• 9a112e9 fix(data): arxiv fulltext via PDF mirror not e-print (opt #12)
• c27418d perf(data): concurrent abstract backfill (opt #11)
• c73365d fix(data): backfill socket timeout + incremental checkpoint (opt #10)
• 5eb5e8a feat(retrieval): broaden bilingual term coverage (opt #9)
• 3ba0e18 feat(retrieval): recency-aware ordering for '最新/latest' queries (opt #8)
• dfeffdf feat(retrieval): hybrid precision-first lexical arm (opt #7)
• 8df659c Revert "feat(retrieval): bi-encoder rerank of fused candidates (opt #6)"
• d846b30 feat(retrieval): bi-encoder rerank of fused candidates (opt #6)
• 57dde78 feat(retrieval): OR-FTS + expand-before-clean + word-boundary filler (opt #3-5)
• d9c58c5 feat(agent): semantic cross-lingual answer verifier (opt #2)
• cbb2b6c feat(retrieval): bilingual CN->EN term expansion (opt #1)
• 345c718 feat(agent): multi-turn conversation memory (#B)
• 8724a68 feat(retrieval): strip conversational filler before retrieval
• 7ef3471 fix(data): unescape + strip HTML markup in titles/abstracts
• 0380198 feat(make): add 'make llm' / 'make llm-stop' for the local 3B model
• 4eb5aa0 fix(agent): cite-aware answer verifier (cross-lingual robustness)
• 4b0530e docs(report): fold eval evidence pack + GraphRAG into project report (#5)
• 4343009 feat(data): filter OpenAlex disambiguation + LoC-style keyword noise (#4)
• 0f80d2c feat(agent): GraphRAG query expansion in /api/chat (#2)
• 8f4f2bb feat(agent): language-aware prompt + answer faithfulness verifier (#3 part)
• da423ba feat(eval): evaluation evidence pack + damped trend forecast + doc fix
• 651148c docs(report): revert project report to black/white palette (match data report)
• e4a3d79 docs(report): polish project report (teal palette + real example cases)
• 0cb19bc docs(report): differentiate project report from data report
• c90de37 docs(report): add project report book (deliverable peer to data report)
• f6c88fc fix(embeddings): only embed chunks present in paper_chunks
• 4432812 feat(data): OpenAlex abstract backfill for records missing abstracts
• 13cccb9 feat(data): title-first dedup + front-matter filter (corpus + DB)
• 3ee8c85 chore(report): regenerate figures with cleaned keywords and 40-topic model
• d9bfe65 docs(plan): restructure into current status + agent-model-layer architecture
• 62dd6db feat(cli): terminal chat REPL and Makefile agent targets
• f8c7d14 feat(frontend): search, trends, recommend, graph panels
• d636d4e test: agent-layer tests (forecast, graph ego-filter, graceful degradation)
• d13ff1f feat(api): hybrid search, evidence chat, trends, recommend, graph endpoints
• 390183a feat(analysis): finer topic model and unified report keyword filtering
• c106347 feat(models): research-agent model layer (embeddings, trends, recommend, graph)
• f30713b feat(infra): harden corpus ingest for full-scale loading
• 8f2d823 Polish SciScope report visuals and analysis assets
• 567510e Backfill fulltext and refresh data report
• b387464 Improve data report trends and fulltext tooling
• 5423b0a Add normalized trend report figures
• e8c16da Generalize processed corpus paths
• 424ef8b Stabilize analysis asset metrics
• 917baa2 Complete analysis asset governance updates
• 89274e3 Quarantine future-year raw records
• e47c6af Govern raw data by source and year
• f4afd99 Add year-balanced harvest and RAG infra
• 90998a6 Refine cover as evaluator evidence pathway
• 4fe5e08 Refine data report cover summary
• ed10d4c Polish report layout and add data readiness audit
• aef4057 Strengthen data report analysis figures
• 8c0fd22 Make report figures show corpus scale
• f0df813 Replace author network with community collaboration views
• 95fbff2 Align report assets with recent 50k corpus
• 1d0e371 Rebuild data report with 50k corpus
• 17347f4 Harden public source harvesting
• 6bb6a14 Add report assets and tidy project outputs
• 11447a5 Add SciScope harvesting and design plan
• bdf2f59 feat: add local vllm llm provider
• f715226 chore: add makefile service commands
• 4a7036c docs: align README with current dashboard
• f5332ff docs: fix README acceptance command block
• 1d0dc65 docs: document foundation acceptance checks
• d0a3d38 docs: clarify DeepSeek mock-mode runbook
• 6869e8e docs: add sciscope runbook
• 92299ef fix: prevent stale evidence chat results
• dda8987 feat: add evidence chat interface
• 7067e18 fix: align frontend typecheck with next type generation
• 39fb2f7 fix: make frontend typecheck artifact-independent
• 1856a0a feat: build research dashboard shell
• 79b3792 chore: ignore typescript build cache
• eecd437 fix: harden frontend scaffold tooling
• 452c749 feat: scaffold sciscope frontend
• febda59 fix: tighten backend api contracts
• bc36708 feat: expose sciscope backend api
• 56e26c5 fix: enforce evidence-grounded chat behavior
• 27d30c0 feat: add evidence-based DeepSeek chat abstraction
• 9106cd2 fix: tighten backend config and schemas
• 7993d21 feat: add backend configuration and corpus service
• 2d6805a fix: stabilize keyword analytics ordering
• 98972b4 fix: stabilize analytics edge cases
• 9a6ee3e feat: compute literature analytics
• 0b2e295 fix: harden paper metadata normalization
• e20b339 feat: load and normalize paper metadata
• 5855f2a test: strengthen sample dataset invariants
• 1501806 test: add sample literature dataset
• dea7659 chore: harden repository hygiene
• 5df9c25 chore: initialize sciscope workspace