Spaces:
Running
Running
Commit History
9 days ago
-
perf: load months from a pre-built HF bundle and page the monthly grid Running
elfsong +1: Claude Fable 5.1The Monthly view fetched ~25 daily parquet files on every cold load and then sent 600+ cards (about 2 MB) over the websocket, which took around two minutes on the Space. - Add src/monthly_papers_bundle.py, which merges a month of daily splits into one `month_YYYY_MM` split in Elfsong/hf_paper_monthly_papers (--month, --all, --force, --no-push/--out) - The app now loads a month from that bundle with a single download and falls back to the per-day loader when a month has no bundle yet - Render 60 cards at a time with a "Show more" button; the websocket payload for Sep 2026 drops from 2.1 MB to 0.4 MB Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>81ae9f3 -
fix: equal-height cards, white background, and a Lifecycle view that opens
elfsong +1: Claude Fable 5.1- Stretch the column > layout wrapper > card chain so every card in a row has the same height, with the footer pinned to the bottom - Restore a white page background (CSS token and theme config) - Lifecycle looked stuck on the Space because Streamlit lazy-loads ~2 MB of chart JavaScript on the first chart and the Space serves static files slowly. Pre-fetch that bundle with a hidden 1px chart, and render the per-phase topic tables as plain HTML instead of st.dataframe - Bring back HF first-page thumbnails on paper cards (lazy-loaded) - Pin streamlit==1.55.0 for the Space build so it matches the version the UI was verified against (the unpinned build picked up 1.66, which has a different widget DOM and left the nav and date input unstyled) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>e7988f9 -
feat: redesign Streamlit frontend for clearer navigation and denser cards
elfsong +1: Claude Fable 5.1- Replace the mode dropdown with a segmented Daily / Monthly / Lifecycle control and sync mode, date, month and snapshot to the URL - Add prev/next period navigation that only steps through dates with data, plus an empty state with jump buttons instead of a toast - Show the top 12 topic pills with counts and move the rest into a searchable "All topics" expander; add a results line with Clear filters - Rebuild paper cards: thumbnail, rank, upvotes, date, title, authors, TL;DR, topic chips, HF/arXiv/PDF links and a Summary button - Restructure the summary dialog (meta row, TL;DR callout, analysis, pros/cons rendered as markdown so links work) - Improve monthly metrics, bar chart and heatmap legend; add lifecycle phase tables and readable trend axes - Warm espresso theme in config.toml and tailwind.css; responsive layout Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>8d46e8d
26 days ago
-
fix: update example date range in usage instructions for daily paper retrieval
elfsongfa8d7d1
-
fix: restore compact lifecycle layout
elfsonge218ebb -
fix: correct responsive header spacing and language wrapping
elfsonga6af448 -
fix: prevent transparent toolbar from blocking header controls
elfsong17610d2 -
fix: correct formatting in usage instructions for daily paper retrieval
elfsong16ab7b7 -
feat: refresh UI with Tailwind and expand lifecycle topics
elfsong288cc4b
-
feat: enhance Codex integration with reasoning effort configuration and improve error handling in lifecycle retrieval
elfsongd9e8acf
-
feat: integrate truststore for enhanced SSL certificate handling across scripts
elfsongef650bb
-
fix: pin Codex reasoning effort and surface worker crashes
Mingzhe +1: Claude Fable 5The pipeline forces model gpt-5.5 but inherited the global model_reasoning_effort setting, which rejects "max" (only none/minimal/low/medium/high/xhigh are valid), causing every fresh summary call to fail with HTTP 400. Add a CODEX_REASONING_EFFORT env var (default "high") passed explicitly via -c, matching the existing CODEX_BIN/CODEX_MODEL/CODEX_TIMEOUT_SECONDS pattern. Also log worker exceptions in the multi-worker path instead of swallowing them with `except Exception: pass`, which had hidden a full run of failures behind a clean exit code. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Eop1XMyhdQScCJdLRxEsQM76e7a3d
-
feat: enhance error handling in daily paper retrieval and add tests for HTTP errors and future date skips
elfsong25a7be2
-
Refactor code structure for improved readability and maintainability
elfsongbe7c1da -
feat: update README and implement Codex integration for daily paper retrieval and summarization
elfsongacd82f7
-
feat: add functionality to upload and manage author cache files to Hugging Face
elfsongeacbaad -
Refactor code structure for improved readability and maintainability
elfsong2cba952
-
feat: enforce HF_TOKEN validation and improve error handling in retriever scripts
elfsong9cac6e0
-
feat: add weekday-based color encoding to the paper publication bar chart
elfsong55a18c8 -
refactor: integrate ProgressDisplay into single-date processing and remove unused summarization and API utilities
elfsong4ad33f5
-
chore: initialize virtual environment dependencies and project structure
dd2e7f2
-
refactor: replace matplotlib with altair for interactive topic trend visualizations
elfsong5b1fa09 -
feat: increase top-n topic display limit from 20 to 40 in streamlit app
elfsong3ee5365 -
refactor: replace matplotlib heatmap with interactive altair chart for topic co-occurrence visualization
elfsong64b6361 -
refactor: replace Altair heatmap with custom Matplotlib visualization showing co-occurrence counts and Jaccard similarity
elfsong365bb5c -
feat: replace streamlit line charts with smoothed matplotlib visualizations for topic trends
elfsong1d3a61a -
feat: apply Gaussian smoothing to topic trend charts in streamlit_app.py
elfsong0078e6b -
feat: add --force flag to allow re-computing and overwriting existing snapshots
elfsong5e842ff -
feat: add time-series topic analysis charts and integrate monthly topic distribution data into lifecycle processing
elfsong99527b8 -
feat: update CJK font selection logic and install Noto CJK fonts in Dockerfile
elfsongdf81b28 -
feat: add Chinese localization for Hype Cycle phase labels and topic counts in Streamlit app
elfsongd71e920 -
feat: add automatic CJK font detection and configuration for matplotlib plots
elfsongceab9c4 -
feat: add lifecycle_retrieve script to compute and upload bimonthly topic hype cycle snapshots to HuggingFace
elfsonga603af9 -
feat: add topic lifecycle analysis and Gartner-style hype cycle visualization
elfsongbcec4fb -
feat: increase top_n parameter for topic co-occurrence analysis from 20 to 30
elfsong423ded7 -
feat: update tooltip labels and set fixed step dimensions for topic co-occurrence chart
elfsongdbb9f2a -
feat: disable label overlap in co-occurrence heatmap axes for improved readability
elfsong40ef3a3 -
feat: add interactive topic co-occurrence heatmap to streamlit dashboard
elfsong216b3e9
-
chore: initialize virtual environment dependencies and project site-packages
79bb5a9 -
chore: initialize virtual environment dependencies
24f531e -
feat: rename app title to Paper Espresso
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>8bad644
-
chore: initialize virtual environment dependencies and project site-packages
91dde67 -
chore: initialize virtual environment with project dependencies
c661390 -
chore: initialize virtual environment and install project dependencies
c6e1987 -
chore: initialize virtual environment and install project dependencies
c4bba08 -
chore: initialize virtual environment dependencies and project libraries
1384bc9 -
chore: initialize virtual environment dependencies and project structure
8e4b732 -
chore: initialize virtual environment with project dependencies
8259c07 -
chore: initialize virtual environment dependencies and site-packages
2adb949 -
chore: initialize virtual environment and install project dependencies
d4f259b