Reademe

gitignore

gitignore

Deploy.md edit

DEPLOY EDIT

README Update

Fix blank-screen crash when clearing the conversation mid-stream

Polished README: live demo link, recorded GIF, deployment-era features + lessons

Remove accidentally committed API-response scratch file

Merge deploy-poc: live demo deployment (model switcher, rate limiting, HF Space)

Code-review fixes: guard empty SSE choices, unstick UI on truncated stream

Record verified deployment + production operational notes

Fix Z.ai model id, surface LLM errors as SSE frames, cache-friendly Docker layers

Pin transformers<5 — FlagEmbedding reranker needs prepare_for_model

Add HF Space frontmatter to README (Piero7/ProbabilityRAG, docker, port 7860)

Dockerfile, static serving, and DEPLOY.md handoff (step 4 of deploy plan)

decompose() falls back to the default hosted model when qwen isn't enabled

Model switcher, public hardening, and CI (steps 1-3+5 of deploy plan)

Add deployment plan for the deploy-poc branch

Apply ponytail audit: delete dead endpoints, scripts, params, and assets

Add unit test suite (40 tests) + fix requirements for fresh installs

Merge pull request #1 from pieroevcc/ui-design-system

Added changed bot locatio nand edited some buttons

Fix "I don't know" on derivation questions (gate + over-strict refusal)

Add interactive mesh background, glass chrome, and reactive mascot

Add design-token system + reactive Rive mascot

Humanize README personal sections (behuman)

Rewrite README in portfolio format (write-readme)

Consolidate docs into eval/IMPROVEMENTS.md; gitignore .claude

Fix dark-mode readability: deterministic dark theme for the app chrome

Add query decomposition to fix comparison/multi-hop retrieval misses

Add ELI5 "explain simply" button + scrape jargon from all answer prompts

Fix LaTeX rendering: normalize \(…\)/\[…\] delimiters before markdown

Merge branch 'claude/elastic-zhukovsky-9e4580'

Add /rag-start session command + running-app screenshot; ignore .playwright-mcp

Cut time-to-sources ~5.5s -> ~0.8s: fix reranker double-pass + shrink pool

Stream answers over SSE (/ask/stream, /why/stream) + stop button

Add session-2 HANDOFF.md + commit launch.json server configs

Add PDF page citations (side-tab viewer) + "Why?" deeper-dive

Add multi-hop eval pass (facts coverage metric) + surface comparison-query failures

Expand retrieval eval: breadth pass (39 questions) + covariance gap

Retrieval + generation reliability for specific computed quantities

Fix chapter-metadata bleed in chunker; canonical chapter labels

V2: local web UI (FastAPI + React/Tailwind) + 8GB GPU fixes

Stage 7: retrieval eval loop + data-driven threshold

Full-corpus ingest (Grinstead & Snell, 518pp) + retrieval/generation fixes

Stage 6: grounded local generation (DeepSeek-R1-Distill-7B via Ollama)

Stage 5: cross-encoder reranking (bge-reranker-v2-m3)

Stage 4: embed (BGE-M3 dense+sparse) + store (Qdrant local), retrieval verified

Stage 3: structure-aware chunking (problem+solution atomic)