- Add real TigerGraph tiger logo (tg-logo.webp) to public/
- Add icon.svg to src/app/ for browser tab favicon (Next.js auto-picks it)
- Navbar: replace SVG circle with actual TigerGraph tiger logo
- Footer: same logo swap; remove Anthropic Claude link; fix copyright line
- HeroSection: badge 'TigerGraph x Claude' -> 'Powered by TigerGraph';
code window model string 'claude-sonnet-4' -> 'gemini-2.5-flash'
- layout.tsx: remove Claude from meta description; drop favicon.ico reference
Explorer live query: better error, science chips, fallback answer
- Replace alarming 'Demo mode may be active' with helpful corpus-scope message
- Add 5 example science question chips (click to fill input)
- Show GraphRAG text answer even when 0 entities found (off-topic queries)
- Error message now explains corpus covers physics/biology/chemistry/astronomy
Explorer: 4 science scenarios, BFS hop filter, live query section
- Replace placeholder HotpotQA nodes with 4 real science scenarios:
General Relativity, DNA Central Dogma, Photosynthesis, Quantum Mechanics
- BFS computeReachability() drives hop slider — visible nodes/edges react
- Hop depth badge on each node; reasoning steps dim beyond hop count
- Stats panel shows Visible Nodes X/Y and Visible Edges A/B
- Live Query section: calls /api/compare, renders graphrag.entities as star graph
- PlaygroundContent: add LLM-Only panel (Pipeline 1) so all 3 pipelines
display side-by-side; update chart to 3 bars (gray/blue/orange);
add token reduction badge; replace HotpotQA examples with science questions;
rename button to 'Run All 3 Pipelines'
- playground/page.tsx: update subtitle from 2-pipeline to 3-pipeline framing
- ArchitectureContent: fix tech stack card HotpotQA → Wikipedia Science
- DocsContent: fix env vars (TG_HOST/TG_TOKEN/HF_TOKEN), update /api/compare
example to show 3-pipeline response, update /api/benchmark to show
correct Wikipedia corpus numbers, fix Dual Pipeline → 3-Pipeline section,
fix Docker env vars
- compare/route: LLM-only + embedding fetch run in parallel (phase 1);
basicRag + graphRAG run in parallel after retrieval (phase 2).
Reduces compare wall time ~50% vs sequential.
- benchmark/route: all 10 samples run via Promise.allSettled (parallel);
within each sample same 2-phase parallel structure.
Reduces benchmark wall time from N*LLM_time to ~1*LLM_time.
- retrieval.ts: add 256-entry in-process embedding cache keyed by
normalized query text — eliminates repeated HF API round trips.
- llm-providers.ts: cache OpenAI client instances per (baseURL, apiKey)
pair — eliminates re-instantiation and dynamic import overhead on
every callLLM() invocation.
Add .gitignore, dataset metadata, retrieval layer, and latest web/graphrag updates