Skip to content

Latest commit

 

History

History
251 lines (199 loc) · 21.9 KB

File metadata and controls

251 lines (199 loc) · 21.9 KB

Analysis docs (docs/analysis/)

Planning hub for engineering tracks, deep dives, and measurement spines. Product phases (0–7): ../roadmap.md. Operational benchmark waves (gate, CI): ../runbooks/roadmap-next-waves.md and ../runbooks/benchmark-decision-gate.md. Quick default entrypoint for agents: ACTIVE.md.

LLM context read policy

If you work through Cursor/LLM context, prefer this read order:

  1. This index (README.md) + target runbook/spec/ADR for the task.
  2. Entry points / live plans from the tables below.
  3. Reference-only docs only when the task explicitly needs deep historical inventory.

Treat these as archival / non-default in normal agent context (matches root .cursorignore):

  • _archive/
  • _snippets/
  • ../idea.md
  • ../pilot/
  • Heavy eval artifacts: eval/results/diagnostics/**, eval/results/multimodel/** (paths under repo root; not under docs/ but excluded from default indexing)

How this folder is organized

Role Meaning Typical Doc status
Entry points / live plans Weekly navigation and canonical roadmaps still in flight active
Closeout / evidence Finished program outputs with artifact pointers (full text may stay in root) reference
Stub → _archive/ Stable URL in root; historical body under _archive/ or git history historical stub
Reference-only Large inventories — not the live BT queue reference
_snippets/ Prompt dumps, trace JSON excerpts — not roadmaps (_snippets/README.md) (n/a)

Where to look first (weekly / “what do we do now?”)

Do not treat master-roadmap-and-refactor-plan-2026-04-25.md §10 as the live backlog — it is a historical execution log (Wave 4–5). Use:

Question Source
Agent unified plan: доработки + benchmark strategy agent-unified-plan-doing-and-benchmarks-2026-05-08.md
Agent next horizon: architecture / chat / ingestion / refactor after D–H agent-engine-next-horizon-2026-05-13.md
Agent R0: feature flags matrix (companion) agent-engine-feature-status-2026-05-13.md
Agent engine + benchmarks — previous waves (archived detail) agent-engine-and-benchmarks-next-waves-2026-05-09.md (stub)
Agent v3 quality benchmark spec agent-v3-quality-llm-judge-benchmark-plan-2026-05-08.md
Agent v3 quality benchmark implementation plan agent-v3-quality-benchmark-implementation-plan-2026-05-08.md
Ontology · extraction · benchmarks (one entry) ontology-extraction-benchmarks-plan.md
Benchmark trust, BT1–BT12, advisory families ontology-benchmarks-trust-audit-2026-04-25.md (§0 snapshot + §5); ../benchmarks/
Structural debt, [OPEN] items ../backlog/refactor-backend.md, ../backlog/refactor-frontend.md
What already shipped (compressed) completed-work-snapshot.md
Gate numbers / trust baseline artifact eval/results/benchmark-trust-baseline.json
Track map + file-conflict rules master-roadmap-and-refactor-plan-2026-04-25.md §1–§2, §5, §9
Agent · tools · context memory agent-runtime-tools-context-roadmap-2026-05-04.mdtool_search v1, compaction; eval/Phoenix: trust-audit + agent-chat-tools-and-trace-audit-master-2026-04-28.md

Entry points by theme

Theme Canonical doc
Agent master-plan: remaining work + benchmarks agent-unified-plan-doing-and-benchmarks-2026-05-08.md
Agent next horizon: architecture / chat / ingestion / refactor agent-engine-next-horizon-2026-05-13.md
Agent R0: feature flags matrix (companion) agent-engine-feature-status-2026-05-13.md
Agent engine + benchmarks — previous waves D/E/F/G/H agent-engine-and-benchmarks-next-waves-2026-05-09.md (stub)
Agent v3 quality benchmark spec agent-v3-quality-llm-judge-benchmark-plan-2026-05-08.md
Agent v3 quality benchmark implementation plan agent-v3-quality-benchmark-implementation-plan-2026-05-08.md
Ontology · extraction · benchmarks ontology-extraction-benchmarks-plan.md
Master track table & principles master-roadmap-and-refactor-plan-2026-04-25.md
Agent runtime · tools · context compaction agent-runtime-tools-context-roadmap-2026-05-04.md — rule-based tool_search v1 shipped (science_graphrag/agent/tool_search.py); LLM shortlist + lazy schemas / L4 compaction — roadmap
Agent chat UI agent-chat-frontend-ui-plan-2026-04-26.md
Agent eval / harness / Phoenix audit agent-chat-tools-and-trace-audit-master-2026-04-28.md
Agent prod flags / rollout agent-chat-prod-rollout-2026-04-27.md
LangGraph Y5/Y6 vs smolagents langgraph-migration-plan-2026-04-25.md
External research · prompt/loop/final_answer discipline (Phase 0 note, not smolagents runtime migration) smolagents-prompt-patterns-for-agent-runtime-2026-05-17.md
Graph UX (GR6–GR9 follow-on) graph-readability-followup-2026-04-25.md; communities/GDS: graph-communities-and-gds-roadmap-2026-04-27.md
Workspace UX workspace-ux-redesign-2026-04-25.md
Reader + translation reader-ux-and-translation-roadmap-2026-04-25.md
Ingestion LLM / Instructor refactor ingestion-llm-architecture-and-instructor-standardization-2026-04-27.md
Logging vs traces logging-system-deep-dive-and-improvement-plan-2026-04-28.md
LLM concurrency / timeouts llm-concurrency-semaphore-and-timeout-hardening-plan-2026-04-27.md
Redis quota Phase 5B (advanced) llm-distributed-quota-phase5b-advanced-scope.md
MinIO / artifact storage seam minio-integration-and-artifact-storage-roadmap-2026-04-27.md
CONTRADICTS evidence gap (product + graph API) contradicts-ontology-and-evidence-gap-2026-04-27.md
Method ontology / dedup richness method-ontology-rich-description-and-dedup-roadmap-2026-04-27.md
Ingest entity extraction & dedup complexity ingest-entity-extraction-and-dedup-complexity-analysis-2026-04-27.md
Benchmark UI / research panel benchmark-panel-research-redesign-plan-2026-04-27.md
Light theme / tokens light-theme-roadmap-2026-04-27.md
Graph force simulation perf graph-force-simulation-performance-analysis-2026-04-29.md
Agent subprocess isolation spike agent-graph-subprocess-isolation-spike-2026-04-27.md
dual_validate + Instructor (optional) instructor-adoption-dual-validate-2026-04-25.md

OD workspace — chat-agent proving ground (paired docs)

Doc Role
od-corpus-claims-methods-trust-audit-2026-04-27.md Trust audit / pre-restore analysis
od-corpus-claims-methods-post-restore-closeout-2026-04-27.md Post-restore closeout
chat-agent-od-workspace-restoration-and-eval-plan-2026-04-27.md Execution plan + operational baseline

Closeout / evidence (program anchors)

Doc Role
orchestration-stabilization-closeout-2026-05-08.md Orchestration stabilization — artifacts, verification, links to plan/baseline stubs
phoenix-closeout-evidence-2026-04-27.md Phoenix Wave X — reproducibility commands and UI/API evidence

OD corpus closeouts live under OD workspace.


Agent tools & settings companions

Doc Role
agent-tools-constants-inventory-2026-05-07.md R/P/G classification for science_graphrag/agent/tools knobs
agent-tools-admin-settings-proposal-2026-05-07.md Design-only: persisted agent_tools admin surface
p0-graph-canvas-perf-baseline-2026-05.md Graph canvas perf baseline (pairs with refactor-frontend.md P0)

Snippets & raw artifacts

Prompt dumps and JSON excerpts live under _snippets/ — not weekly roadmaps.


Archive & deletion policy

  1. Incoming links: before deleting or moving a docs/**/*.md path, search the repo (rg / IDE references). If anything links here, keep a stub at the old path (short redirect + “full text in git history or _archive/”).
  2. Prefer move over delete: completed long write-ups go to _archive/ with a root stub row in the Closed / superseded section below.
  3. Stable URLs: Habr, release notes, and external bookmarks may target root docs/analysis/<name>.md — do not break without a stub.
  4. Non-default LLM context: aligns with root .cursorignore (_archive, _snippets, idea.md, pilot/, heavy eval/results/…).

Reference-only (large inventory / completed gold — not the live BT queue)

Doc Role
ontology-benchmarks-roadmap-2026-04-24.md Wave M–T deep inventory (large); start here: ontology-extraction-benchmarks-plan.md; live BT queue: trust-audit
corpus-gold-pack-v1-2026-04-25.md Gold pack layout + layers (Phase 0–6 complete); phase execution log: _archive/corpus-gold-pack-v1-phase-log-2026-04-25.md
dedup-ingest-parity-matrix-2026-04-26.md Dedup queues matrix (scan vs ingest)

Closed / superseded (full text archived under _archive/; root filename is a short redirect)

Stable URLs and backlinks may still point at these root paths — open the link, then follow into _archive/ for the full document.

Root stub Archived copy
phoenix-tracing-coverage-2026-04-25.md _archive/phoenix-tracing-coverage-2026-04-25.md — Wave X CLOSED; summary evidence: phoenix-closeout-evidence-2026-04-27.md
graph-work-vs-workspace-unification-dry-plan-2026-04-28.md _archive/graph-work-vs-workspace-unification-dry-plan-2026-04-28.md — Phases 0–5 DONE
graph-navigation-hash-router-remediation-plan-2026-04-28.md _archive/graph-navigation-hash-router-remediation-plan-2026-04-28.mdDONE
workspace-graph-methods-citations-root-cause-2026-04-27.md _archive/workspace-graph-methods-citations-root-cause-2026-04-27.md — P0 DONE
orchestration-stabilization-plan-2026-05-07.md _archive/orchestration-stabilization-plan-2026-05-07.mdCLOSED program; closeout orchestration-stabilization-closeout-2026-05-08.md
orchestration-stabilization-baseline-2026-05-08.md _archive/orchestration-stabilization-baseline-2026-05-08.md — pre-WS snapshot
wave-a-residual-structural-hardening-2026-05-08.md _archive/wave-a-residual-structural-hardening-2026-05-08.md — Wave A checklist DONE
agent-runtime-train-t1-acceptance-2026-05-06.md _archive/agent-runtime-train-t1-acceptance-2026-05-06.md — Train T1 milestone DONE
agent-note-cost-eval-2026-05-06.md _archive/agent-note-cost-eval-2026-05-06.mdagent_note cost methodology archived; live 50-turn token pilot open (see stub header + R2 spec §agent_note)

Reader authorship contract (implemented Phases 0–3): work-graph-authorship-reader-contract-2026-04-28.md — closed as a delivery plan; keep for contract text.

Frontend verification checklist (content archived): agent-chat-frontend-verification-gaps-next-wave.md_archive/agent-chat-frontend-verification-gaps-next-wave-2026-04-26.md.


Publication / Habr (measurement spine — does not replace engineering roadmaps)

habr-article-narrative-and-measurement-plan-2026-07.md — pinned eval/results/habr-window-*, links to ../report/habr-article-2026-04-29.md (stub) and claims benchmark contract.


Archive index

_archive/ — completed waves (ingest async U–W, Wave 4–6 write-ups, full chat roadmap, gold phase log, historical UX), orchestration stabilization program artifacts, Train T1 / agent_note milestone notes, plus full copies of closed plans listed under Closed / superseded.


Root markdown inventory (catalog)

Sorted alphabetically. See sections above for roles; stubs point into _archive/.

File Bucket
README.md This index
agent-chat-frontend-ui-plan-2026-04-26.md Agent UI
agent-chat-frontend-verification-gaps-next-wave.md Stub → archived frontend verification checklist
agent-chat-prod-rollout-2026-04-27.md Prod rollout
agent-chat-tools-and-trace-audit-master-2026-04-28.md Eval / trace audit
agent-engine-and-benchmarks-next-waves-2026-05-09.md Stub → archived detailed wave log
agent-engine-feature-status-2026-05-13.md Agent R0: feature flags matrix (companion)
agent-engine-next-horizon-2026-05-13.md Agent next horizon: architecture / chat / ingestion / refactor
r2-chat-contract-closeout-2026-05-13.md R2 chat SSE product contract closeout (degraded_mode, product layers, doc sync)
r3-long-thread-live-baseline-2026-05-13.md R3 operator checklist: live long-thread trace-review + compare (Wave H gates)
agent-graph-subprocess-isolation-spike-2026-04-27.md Spike
agent-note-cost-eval-2026-05-06.md Stub → archived agent_note cost methodology; live token pilot open (see R2 spec)
agent-runtime-tools-context-roadmap-2026-05-04.md Agent · tools · context roadmap
agent-runtime-train-t1-acceptance-2026-05-06.md Stub → archived Train T1 acceptance
agent-tools-admin-settings-proposal-2026-05-07.md Design proposal (agent_tools admin)
agent-tools-constants-inventory-2026-05-07.md Reference inventory (tool knob classes)
agent-unified-plan-doing-and-benchmarks-2026-05-08.md Agent master-plan
agent-v3-quality-benchmark-implementation-plan-2026-05-08.md Agent v3 quality benchmark implementation plan
agent-v3-quality-llm-judge-benchmark-plan-2026-05-08.md Agent v3 quality benchmark spec
benchmark-panel-research-redesign-plan-2026-04-27.md Benchmark UI
chat-agent-od-workspace-restoration-and-eval-plan-2026-04-27.md OD eval
completed-work-snapshot.md Shipped / closed summary
contradicts-ontology-and-evidence-gap-2026-04-27.md CONTRADICTS gap
corpus-gold-pack-v1-2026-04-25.md Gold reference
dedup-ingest-parity-matrix-2026-04-26.md Dedup matrix
graph-communities-and-gds-roadmap-2026-04-27.md Graph structural UX
graph-force-simulation-performance-analysis-2026-04-29.md Perf analysis
graph-navigation-hash-router-remediation-plan-2026-04-28.md Stub → archived hash router plan
graph-readability-followup-2026-04-25.md Graph UX
graph-work-vs-workspace-unification-dry-plan-2026-04-28.md Stub → archived DRY plan
habr-article-narrative-and-measurement-plan-2026-07.md Publication spine
ingest-entity-extraction-and-dedup-complexity-analysis-2026-04-27.md Ingest analysis
ingestion-llm-architecture-and-instructor-standardization-2026-04-27.md Ingest LLM
instructor-adoption-dual-validate-2026-04-25.md dual_validate
langgraph-migration-plan-2026-04-25.md Y5/Y6
light-theme-roadmap-2026-04-27.md Light theme
llm-concurrency-semaphore-and-timeout-hardening-plan-2026-04-27.md LLM pools
llm-distributed-quota-phase5b-advanced-scope.md Quota 5B
logging-system-deep-dive-and-improvement-plan-2026-04-28.md Logging
master-roadmap-and-refactor-plan-2026-04-25.md Master tracks
method-ontology-rich-description-and-dedup-roadmap-2026-04-27.md Method ontology
minio-integration-and-artifact-storage-roadmap-2026-04-27.md Artifacts
od-corpus-claims-methods-post-restore-closeout-2026-04-27.md OD closeout
od-corpus-claims-methods-trust-audit-2026-04-27.md OD audit
ontology-benchmarks-roadmap-2026-04-24.md Reference inventory (Wave M–T tables)
ontology-benchmarks-trust-audit-2026-04-25.md Live BT / trust queue
ontology-extraction-benchmarks-plan.md Entry point ontology / extraction / benchmarks
orchestration-stabilization-baseline-2026-05-08.md Stub → archived pre-program baseline
orchestration-stabilization-closeout-2026-05-08.md Closeout (orchestration stabilization program)
orchestration-stabilization-plan-2026-05-07.md Stub → archived orchestration stabilization plan
p0-graph-canvas-perf-baseline-2026-05.md Frontend graph canvas perf baseline
phoenix-closeout-evidence-2026-04-27.md Phoenix evidence
phoenix-tracing-coverage-2026-04-25.md Stub → archived Phoenix plan
reader-ux-and-translation-roadmap-2026-04-25.md Reader / LX
smolagents-prompt-patterns-for-agent-runtime-2026-05-17.md External research · prompt/loop/final_answer discipline (Phase 0 note)
wave-a-residual-structural-hardening-2026-05-08.md Stub → archived Wave A checklist
work-graph-authorship-reader-contract-2026-04-28.md Authorship contract
workspace-graph-methods-citations-root-cause-2026-04-27.md Stub → archived workspace graph RCA
workspace-ux-redesign-2026-04-25.md Workspace UX

Non-markdown snippets: _snippets/phoenix-trace-multistep-excerpt.json (see _snippets/README.md).

Backlog (structural debt): ../backlog/refactor-backend.md, ../backlog/refactor-frontend.md.