v0.2.1

Changelog

• 63b408e fix: allow hosted doctor readiness latency
• f109da2 fix: support hosted search without runtime embeddings
• 6397b09 fix: enable deepseek evidence chat provider
• 9950f13 deploy: add hosted backend blueprint
• 56cea13 merge hosted backend v0.2.1 readiness
• 78a0465 fix: harden anonymous rate limiting
• 3d842e3 fix: harden hosted release readiness
• 989264c ci: require hosted backend URL for release
• 12cd0e7 release: prepare sciscope tui v0.2.1
• 2a7332d docs: update hosted TUI release path
• 674f313 fix: cover hosted recovery paths
• 0bb2009 feat: clarify hosted backend recovery
• 8b8e8c1 fix: cover hosted backend env trimming
• c6c1e99 fix: harden hosted backend URL handling
• b6687a0 feat: default TUI to hosted backend
• 39f32ba fix: tighten hosted backend container build
• c9d549e build: add hosted backend container
• ba7e3c3 fix: enforce budget across agent routes
• d5d36e9 feat: enforce hosted agent budgets
• cb96453 fix: preserve local exception semantics
• a66abcd fix: harden request context middleware
• 7906146 feat: add hosted request context
• 354679b fix: harden hosted budget validation
• 6545734 feat: add hosted request budgets
• 73ed81d fix: harden readiness responses
• accebb5 fix: make readiness env aware
• 95b1836 fix: order health route import
• adf8199 feat: add hosted backend health checks
• cc17e80 fix: clarify hosted deployment contract
• d684860 docs: record hosted backend deployment contract
• 33f9aba Plan hosted backend implementation for v0.2.1
• 19b61d0 Design hosted backend for v0.2.1
• 6566118 Update Scoop manifest for v0.2.0

v0.2.0

Changelog

• 1c45bc5 Merge TUI output polish for v0.2
• fd66c0c Optimize TUI rendering for v0.2
• 75e1ef0 Add Scoop distribution path for Windows
• de991bb Repair missing tool results before each model call (CC yieldMissingToolResultBlocks)
• e4728c0 TUI: add a live streaming cursor to the in-progress answer
• 8fa74ee TUI: split the 3251-line main.go into cohesive files
• 83fccd3 Unify slash commands on the skill-prompt model (remove specialist divergence)
• 4cb2061 Add session memory service (Claude Code SessionMemory analog)
• 17381f1 Split system-prompt assembly out of llm.py into prompts.py
• a597dbd Wire autoCompact into the agent loop with telemetry
• f623c9f Report covers: magazine image-hero redesign + cover-background generators
• 0ae0e1d TUI: slash-command arg templates + palette beautification
• f801f4e Add slash-command / skill registry with pre-loop interception
• 1a7015a Add two-tier context compaction service
• 7766e89 Refactor agent tools into self-contained package + lifecycle hooks
• 44894dd Polish report covers
• 03c6036 Make project report judge-facing
• cfc0bbb Restore project figure palette
• 9e2804b Polish project report figures
• 198bde2 Sync reports with agent skill smoke checks
• 7dcea79 Add agent skills and live smoke checks
• edf671f Docs+report: reflect MCP (bidirectional) + multi-agent
• b1b86a6 Add specialist sub-agents via delegate tool (multi-agent, direction ③)
• 859cb1d Consume external MCP servers as agent tools (MCP direction ②)
• efd2508 Add SciScope MCP server (expose tools to any MCP client)
• e1c555d Report: align overview innovations + social benefit with agent upgrades
• fb49480 Report: fold agent upgrades into innovation + economic sections
• 73cbb28 Curb over-searching: stop after enough evidence
• 81eff75 Assemble the system-prompt tool catalog from per-tool fragments
• e08a086 No emoji in answers; allow occasional kaomoji as house style
• 703d504 Stop forcing tool use on general questions
• a61446c Remove the bottom status/shortcut line under the composer
• e02144f Drop the Enter/Esc/slash hint line under the composer
• d3101a1 Surface validation denials as a distinct recovery card
• 9d55448 Render evidence cards by default, not just in demo
• cc0951b Surface stop_reason + token usage in agent meta
• 34cfe20 Default to DeepSeek via gitignored .env.local
• 9f70555 Raise max_tokens for cloud generation
• 1f99439 Make the agent model layer pluggable (DeepSeek or local)
• 07037b8 Make agent answer format adaptive, not templated

v0.1.3

Changelog

• 4c6a9b2 Clean frontend scope and refresh data delivery
• a5ad08c Fill winget manifest for v0.1.2

v0.1.2

Changelog

• 91bb11e Add Windows release and winget packaging template
• b1d7239 Surface agent activity inline and fix TUI composer + palette

v0.1.1

Changelog

• aa05d25 Refine report layout and figure flow
• 926e3cb deploy: drop db host-port exposure (network-only)
• 1aa9ff2 Unify TUI chrome to Chinese and drop dead panel idiom
• 75a6bce Add transcript scrolling to the TUI
• d223f30 Allow tool-free answers for common-sense questions
• a814ef6 Add Aliyun demo deployment scaffolding
• 0492cda Polish report layout and formulas
• 5e5d5e4 Fix Homebrew install docs with tap-trust step
• 54f8d88 Remove report figure frames

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 retr...