Core justification library expanding cross-collection sidecars, figure recognition, and explication readings across PDAI collections

The library powers evidence extraction, sentence sidecars, figure citations, and the explication column. Recent work extended figure referencing and register binding across collections like karman-context, eulerOO, and enestroem under a standing propagation rule. Heavy development continues with 179 files changed in the past 24 hours.

• Re-indexing 17 collections where model text entered the search corpus requires owner releaseS2
• Full propagation of sidecars, Schemakarte, and figure extraction across all collections like eulerOO and enestroem is still runningM3S4
• Long-running OCR and figure transcription on karman-context documents still has over 5,600 pages pendingS2
• Famulus Normal Form NLI bakeoff has not been run against the frozen environmentW1

• YouApprove re-indexing of the 17 collections affected by model text in the search corpusS2
• AgentRun the needle suite against the frozen environment for Famulus Normal FormW1
• AgentWrite the fnf-spec and freeze config/strength.jsonW1
• AgentRun local, cloud, and AFM bakeoff for FNF DuckDB integrationW1

Famulus News newsletter generator, GitHub Pages deployment, and Gmail subscriber mailer setup

The Famulus News site is published to GitHub Pages, and the subscription box and issue generator have been built. The automated mailer cannot send newsletters until the Gmail App Password is configured. Issue 2026-10 has been rebuilt covering October 12–20.

• Gmail App Password must be configured before the Subscribe box works and emails can be dispatchedS1M1
• Work item milestones remain pending completion across mailer and activity gathering tasksW1W2

• YouStore Gmail App Password to enable subscription storage and newsletter mailingsS1M1
• YouRun famulus-news publish dry run and dispatch issue 2026-10 to subscribersS1M1

Collation of Babylonian Diaries OCR variants and preparation for re-OCR across 594 pages

A collation of surviving OCR readings across 63 multi-engine pages yielded 2,687 disputed positions and prompt instruction v8. A new OCR run over 594 pages awaits an engine choice, while tablet edition assembly and DuckLake export remain partial. The Babylon landing page has unpublished local changes.

• A decision is required on which engine should reprocess the 594 remaining pages.S1
• ADART tablet edition DTD validation, volume audit, remarks capture, and DuckLake export are uncompleted.W1S2
• Landing page 'Babylonische Astronomie' has local build changes that are not published to GitHub.L

• YouChoose the OCR engine for reprocessing the 594 pending diary pages.S1
• AgentValidate tablet-edition.dtd and assemble_editions.py against a generated sample page using xmllint.W1
• AgentRun the round-trip audit across all ADRT volumes and review failures.W1
• AgentPublish the local build of the Babylonische Astronomie landing page to GitHub.L

Publishing multilingual development news for Famulus with an integrated subscriber mailer

Static pages for the site, guide, and October 2026 issue have been exported. A subscriber management system and Gmail mailer CLI were implemented. The mailer cannot send until the owner configures a Gmail App Password.

• Gmail App Password has not yet been set by the ownerM1

• YouSet Gmail App Password to enable mailer deliveryM1

Static landing pages for scholarly collections and Famulus platform documentation

The Famulus Elucidation pages were generated locally, and the new Famulus landing page gathers twelve platform papers. Fact collection for the publications page currently fails on three folders, while six landing pages show outdated data. The live hub links and card assets remain incomplete pending a manual launchd script run and file sync.

• Fact collector for the publications landing failed on folders astronomical-expert, calculation-tools, and famulus-normalisationL
• Live hub docs links remain dead until repoint_hub_notebook_launchd.sh is executed outside the sandboxS2
• Six landing pages have outdated fact counts and the Famulus landing inputs changed after buildL
• Local builds for six landings are not yet published to their GitHub repositoriesL
• Seven paper covers or PDFs must sync from mac-studio to complete the Famulus landing cardsS2

• YouRun platform/scripts/repoint_hub_notebook_launchd.sh to restore live hub document routingS2
• AgentFix the arXiv landing fact collector to handle uncaptured paper foldersL
• AgentRebuild the Famulus landing and the six landings with changed factsL
• AgentPublish updated static builds to their GitHub Pages repositoriesL

FastAPI reverse proxy routing landing pages and dashboard traffic across dual backend machines

The proxy handles failover routing between openscience as primary and mac-studio as fallback for landing backends. Short URLs and newsletter paths were implemented locally on openscience but fail publicly because mac-studio runs older code. mac-studio could not be reached via SSH to restart its proxy.

• Public requests to /babylon/ and /famulus-news/ return 404 or inventory pages because mac-studio proxy is not restarted with synced codeS1
• SSH connection to mac-studio failed during the last deployment attemptS1

• AgentSync proxy code to mac-studio and restart the proxy service on port 5001S1
• AgentVerify public resolution of https://lettre.ngrok.dev/babylon/ and https://lettre.ngrok.dev/famulus-news/S1

Writing arXiv paper on astronomical-expert in Typst and Quarto format

The project is actively drafting the astronomical-expert paper across Typst and Quarto documents. Per owner correction, all Diaries sky engine calculations in astronomical-expert must strictly use topocentric apparent ecliptic coordinates. Work is progressing without any blockers.

• Astronomical-expert sky situation engine must use topocentric apparent ecliptic coordinates rather than equatorial readingsM2

• AgentEnsure astronomical-expert /sky_situation calculates topocentric apparent ecliptic coordinates of dateM2

Secondary literature collection paused with new ingest batch pending processing

Recent Springer articles were processed and audited with figure crops fixed. A new batch of seven PDFs in ~/pdai-ingest was sorted, but ingestion into kepler-sec and kepler-context stopped before execution.

• Uncompleted ingestion tasks: create-context, ingest-sec, ingest-context, and register for seven PDFs in ~/pdai-ingestS1

• AgentRun ingest-sec and register the secondary literature PDFs from ~/pdai-ingestS1
• AgentCreate kepler-context folder and ingest the context PDFsS1

Famulus reference DuckLake migration partially implemented and currently stale

The Quarto paper notebook has seen updates to chapters and migration documentation, with 76 file changes over the past month. The Famulus reference DuckLake migration is marked stale with partial implementation and completion pending. No files have changed in the project over the last five days.

• Famulus reference DuckLake migration remains stale with partial implementation and completion pendingW1

• AgentFinish remaining implementation for the Famulus reference DuckLake migrationW1
• AgentClose out pending done milestone for the DuckLake migration work itemW1

Building Obsidian ingest pipeline, Konferenz Partitur processing, and Operator Mac app backend

The backend watchdog is being reworked alongside a new ingest folder for incoming notes and Konferenz planning. Implementation of the ingest layout, Partitur processing, and the Operator Mac app backend stopped with unfinished tasks four days ago.

• Watchdog fix, ingest layout, and ingest router tasks remain uncompletedS1
• Konferenz ops and Konferenz Partitur tasks are not completedS1
• Lease management and backend server integration for the Operator app remain pendingS1

• AgentFix the watchdog runner and complete the ingest folder routing layoutS1
• AgentImplement the Konferenz-Planung Partitur and related operationsS1
• AgentComplete the lease and serve implementations for the backend APIS1

Partitur-driven routing and conference digest agents for Obsidian vault notes

The project uses ai-agent-orchestrator to route notes from 0-Inbox and run conference planning workflows. Recent edits focused on conference planning and area report tests, but activity has stopped for 4 days.

• Work item to review how to configure Obsidian data is stale with implementation partialW1

• AgentComplete implementation and finalize review of Obsidian data configurationW1

Pipeline setup and table segmentation for Tycho Brahe Opera Omnia

The page census of 14 tomi is complete and configuration files for table segments and Famulus are in place. The remaining pipeline steps, Famulus database build, and DuckLake export are stalled because mac-studio is unreachable and the data volume is unmounted. The public Brahe landing page remains outdated and lacks volume links.

• mac-studio times out on SSH and the PDAI-1 volume is unmounted, blocking OCR re-reads and DuckLake exportS1
• Brahe landing page is outdated with mismatched page counts and no linked volumes or tabulaeS1L
• Small fractions in Opera omnia XII were misread in roughly 4 out of 10 rowsS2

• YouReview config/segments.yaml against the Hartmann templateW1
• AgentRestore mac-studio Tailscale reachability and mount the PDAI-1 volumeS1
• AgentRun the full build, Famulus database generation, and DuckLake exportS1W1
• AgentRebuild and publish the Brahe landing page with updated statistics and volume linksLW1

ArXiv paper comparing Grok chat claims against evidence on Kepler's star catalogue

The project folder, verbatim Grok transcript, claim ledger, and first evidence-funnel filter run across all 11 subgoals are set up. Core analytical tasks—close reading, recomputations, ledger verdicts, and writing paper sections—remain unstarted. Work has been inactive for 4 days with tasks W7 to W12 open.

• Model version and chat date are missing from the verbatim chat transcriptD1M1
• Close reading per subgoal, quote verification, and Latin-English alignments are not doneD1
• Recomputations for epoch, precession, and Tycho-Kepler collation have not startedD1
• Paper sections, ledger verdicts, and full bibliography verification remain unwrittenD1

• YouProvide model version and chat date for source/grok-chat.mdD1M1
• AgentImprove OCR partitur for Opera omnia XII fraction recognitionS1F
• AgentPerform close reading and Latin-English quote verification for subgoals Q1–Q11D1
• AgentRun epoch and precession recomputations and populate the claim ledger verdictsD1

Public website for Markus Wäfler's Tall al-Hamidiya archive and the Mitanni empire

The project builds a public website featuring a 3D card cabinet, excavation slides, and Hamidiya volumes 1-5 linked to pdai-viewer. Progress stands at 76% across tasks, with no file activity recorded in the past 7 days. Several backlog tasks remain open, including rights clarification, OCR, and domain activation.

• Clarification of rights for slides and estate photographs with the estate remains pendingT
• SwissTransfer DNG series needs to be reviewedT
• OCR for Hamidiya volumes 4 and 5 (Waefler2003, Waefler2011) is missingT
• Checklist of uncertain cross-references requires reviewT
• Captions for excavation slides need writing, currently using machine tags onlyT

• YouClarify rights for slides and estate photos with the estateT
• AgentReview SwissTransfer DNG seriesT
• AgentRun OCR for Hamidiya volumes 4 and 5T
• AgentActivate public domain mitanni.lettre.aiT

Schema-first OCR processing for Bischoff's ninth-century Continental manuscript catalogue

Band I OCR is complete with 518 pages, while Band II has XML on 330 of 465 pages. No file activity has occurred in 16 days, leaving Band II unfinished and Band III unprocessed. In addition, the related landing page has uncommitted build changes waiting to publish.

• Band II OCR is incomplete at 330 of 465 pages, with Band III pendingD1
• Landing page 'Mittelalterliche Handschriften' has 5 uncommitted changes and differs from published GitHub PagesL

• AgentResume schema-first OCR processing for the remaining pages of Band IID1F
• AgentCommit and publish the updated 'Mittelalterliche Handschriften' landing page build to GitHub PagesL

Archival research and website materials for Kepler Kommission facsimile scans

The project documents and publishes materials related to rediscovered Munich facsimile scans from March 2024. Over 900 files including site pages and source documentation were updated within the last month, but file activity has ceased for 19 days. Extensive archival inquiries and literature reviews remain pending in the project todo list.

• Landing page 'Kepleriana 2026' has changes since last publish (pdai:kepleriana-munich-2024)L
• Extensive archival requests across publishing houses, Bavarian academies, and national archives remain unaddressedD1
• Secondary literature evaluations for Dyck and Kepler edition history remain openD1

• AgentRebuild and publish landing page 'Kepleriana 2026' with recent changesL
• YouBegin archival queries starting with C. H. Beck and Bavarian Academy archivesD1

Hevelius collection gaps resolved; landing page rebuild and publish remaining

Collection gaps across volumes I, II, and III have been closed down to zero. The Hevelius landing page is outdated because published figures do not reflect these updated facts. Redundant backends have been configured on openscience and mac-studio to serve landing page data.

• The Hevelius landing page is outdated with eight facts changed, including gap counts dropping to zero.L

• AgentRebuild the Hevelius landing page with the updated facts and publish it.L

Museum landing page and narratives for Theodore von Kármán Nachlass awaiting repository sync and deployment

Interactive linked filters driving sidebar selections were updated across the museum site and welcome page. The local build is not yet committed to the GitHub repository, causing GitHub Pages to serve an out-of-date index.html. Additional charts and period links remain unverified on the museum site.

• Repository karman-2026 has 6 uncommitted changes and GitHub Pages serves an out-of-date index.htmlL
• Museum site other chart types and the new period link were not clicked through or testedS1

• AgentCommit and push local changes in karman-2026 to update GitHub PagesL
• AgentClick through the remaining museum site chart types and period links to verify filteringS1

macOS mail hub adapted for mac-studio data split with verification pending

MailHub runs as a native macOS app for notes, dictation, and mail preparation. Recent updates split unanswered mail fetching between Gmail and the mac-studio service while routing calendar reminders to Google Calendar. No changes have occurred in the last four days.

• korrespondenz.duckdb resides on mac-studio and its index was locked when last queriedS1
• Verification milestone remains pending for the completed implementationW1

• AgentVerify mail retrieval and calendar entry creation against the mac-studio serviceW1S1

Diagnostics on viewer loading performance and outdated landing page metrics

Viewer loading latency for Reuleaux 1875 was diagnosed as tunnel startup overhead rather than image serving size. The landing page numbers have drifted across 12 facts and the local build differs from the published copy. No repository files have changed in the past 18 days.

• Landing page numbers are outdated across 12 facts including blank and figure counts for Reuleaux 1875L
• Local build of landing page differs from the published copy by 1 fileL

• AgentRebuild the landing page with updated collection facts and publish the buildL

OCR package for Johannes Kepler Gesammelte Werke awaiting landing page rebuild and publish

The collection provides schema-first OCR resources for Johannes Kepler's Gesammelte Werke. Recent updates configured audit rules, OCR instructions, and partitur settings across volumes, expanding the collection to 25 documents and 6,049 pages. The Kepler landing page is outdated and requires a rebuild to reflect the current data.

• The Kepler landing page is outdated with document and page counts diverging from the dataL

• AgentRebuild and publish the Kepler landing page with current metricsL

Collection scripts and database build updated while landing page awaits republish

Thirty-three files including entity typing, OCR scripts, and database configs were updated 16 days ago. The landing page 'Kepleriana 2026' has not been published since those changes occurred.

• Landing page 'Kepleriana 2026' has not been republished since project changes were made.L

• AgentPublish landing page 'Kepleriana 2026' reflecting recent project changes.L

Landing page and animated narratives on regularity theory needing rebuild and publication update

The site contains eight animated narratives and landing pages in German and English. Three paper page counts have changed since the last build, leaving the published landing page outdated. Local files also differ from the published copy by five files.

• Landing page is outdated because bell_pages, ca_pages, and pages_causal changed since the buildL
• Local build differs from the published copy across five filesL

• AgentRebuild site data and pages with build_facts.py and build_site.py to update page countsD1L
• AgentPublish the updated site folder to the grasshoff/kausalitaet-2026 repositoryD1L

Static GitHub Pages site for ancient sundials awaiting publication of local builds

The project serves the static site for ancient sundials measured on 3D models. No files were modified in the past 19 days. Local builds for both English and German landing pages differ from the published copies and have not yet been pushed to GitHub.

• Landing page 'Ancient Sundials' local build is not in GitHub and differs from published copy by 3 files.L
• Landing page 'Antike Sonnenuhren' local build is not in GitHub and differs from published copy by 1 file.L

• AgentPublish the local build of 'Ancient Sundials' to the GitHub repository.L
• AgentPublish the local build of 'Antike Sonnenuhren' to the GitHub repository.L

Static site for Euler works and correspondence awaiting publish of local build

The project hosts a static GitHub Pages site for Leonhard Euler's works, correspondence, and bibliography. File activity has been quiet for 19 days. The local landing page build differs from the published copy following changes in upstream datasets.

• Local landing page build differs from published copy and is not yet in the GitHub repositoryL
• Datasets pdai:enestroem, pdai:eulerOO, and pdai:eulerOO-IVA changed since the last publishL

• AgentPublish the local Euler landing page build to the GitHub repositoryL
• AgentRun build.py euler and export_pages.py euler in ../landings2026/buildD1

Schema-first OCR package for Copernicus primary sources with an unpublished landing page

The Copernicus primary-source OCR schemas, rules, templates, and quality fault dossiers are in place. No files have changed in the past 21 days. The local build of the Kopernikus landing page differs from the published copy and has not been pushed to GitHub.

• Landing page 'Kopernikus' local build is not yet in the GitHub repository and differs from published copyL

• AgentPublish the Kopernikus landing page build to the GitHub repositoryL

Orchestrator flows, PDAI agent tools, and monitor Partituren are in active development

The engine has heavy file activity across monitor flows, PDAI agents, and schema definitions. Integration of the person dossier model remains partially completed. A requested background agent for PDAI table extraction on Tycho and Kepler-prim was sidelined during recent setup.

• Table extraction background agent for Tycho and Kepler-prim collections remains unbuilt.S1
• Person dossier model with source references is partially implemented and pending completion.W1
• Copilot tool cap exceeded 128 tools on LLM Gateway, potentially dropping MCP tools during agent tasks.S1

• AgentImplement the background agent for table extraction in Tycho and Kepler-prim collections.S1
• AgentComplete the implementation and verification for the person dossier reference model.W1

Adding the Bedeutung passage-reading column and expanding viewer components across PDAI collections

The viewer recently gained the right-edge Bedeutung badge and third-column reading interface for passage explication. Active development continues across 22 files in the past 24 hours covering server routing, dossier views, and styling. Features established on Kármán still need propagation to Euler and Eneström collections.

• Viewer features built for Kármán have not yet reached eulerOO, eulerOO-IVA, or enestroem.M3
• Viewer URL references such as Wittgenstein-Mechanik Hertz1894 failed over the ngrok endpoint.S1
• Karman-nachlass architecture and sidecars work item remains partially implemented.W1
• Babylonian diaries OCR quality review remains incomplete with verification pending.W2

• AgentPropagate the Bedeutung column and viewer surfaces to eulerOO, eulerOO-IVA, and enestroem.M3
• AgentFix broken ngrok routing for Wittgenstein-Mechanik Hertz1894 viewer references.S1
• AgentComplete implementation and verification for Karman-nachlass architecture and sidecars.W1
• AgentVerify Babylonian diaries OCR quality review.W2

Active development on central web hub components including monitor page and application routes

The central web hub for the ai-mac-studio platform is in active development. Three files were updated in the past 24 hours, with the most recent changes made one hour ago. Recent work touches monitor_page.py, app.py, and famulus_news.py.

Unifying nightly project monitoring and implementing dual-backend proxy failover across machines

The platform unified the nightly project monitor onto gemini-3.8-flash running eight parallel workers ahead of the scheduled 02:30 run. Proxy failover with fallback hosts was implemented and verified between openscience and mac-studio to handle upstream drops. The Work Monitor hub page is live, though collector scripts remain uninstalled on secondary machines.

• Worklog collector scripts are not yet installed on mac-studio, mac-mini, and mac-air to provide daily snapshots.S6W2
• The first scheduled 02:30 Project Monitor run on gemini-3.8-flash requires verification of its report output.S1
• The edit access token has not been entered into /admin/work.W2

• AgentCheck report output and dashboard status after the 02:30 Project Monitor run.S1M2
• YouRun scripts/install-worklog.sh on mac-studio, mac-mini, and mac-air.W2
• YouPaste the edit token into Edit access on /admin/work.W2

Building VS Code extension editor and updating goal enfolding for famulus-proof

The project develops a multi-pane VS Code Insiders extension featuring an editor, artifacts pane, inspector, debugger, and literate AI composer. Goal enfolding was recently overhauled to align with the updated famulus-proof pipeline, adding fundlage bridging and validation guardrails. Development continues with active file changes across the composer engine, guardrails, and Quarto documentation.

• Implementation of the interactive VS Code Insiders notebook editor remains only partially finishedW1
• VS Code Insiders distribution is not installed or synchronized onto mac-airS1

• AgentComplete the remaining implementation tasks for the VS Code Insiders notebook editor extensionW1
• YouSet up the VS Code Insiders distribution or syncthing sync on mac-airS1

Configuration updates and ongoing famulus-proof propagation for Euler series IVA

Segments configuration was updated within the past hour following recent register and OCR instruction work. Architecture, sidecars, and DuckDB content must adopt the famulus-proof standards established on karman-nachlass. The Euler landing page contains local changes that have not yet been published to GitHub.

• Local build of the Euler landing page differs from the published copy and is not in GitHub.L
• Famulus-proof developments from karman-nachlass must be propagated to sidecars, DuckDB content, and viewer surfaces.M1

• AgentPublish the local build of the Euler landing page to GitHub.L
• AgentPropagate famulus-proof sidecars, DuckDB schema, satz, and register bindings to eulerOO-IVA.M1

Adding Bedeutung side badge column and interactive filters to the Kármán Nachlass viewer

The Kármán viewer UI is being updated with a Bedeutung side badge column beside the XML transcript. Linked filtering across box, year, and Sachgebiet counts has been completed and verified against the running viewer on port 5725. Implementation of the side badge column and landing backend health checks remain partially finished.

• Implementation of the Bedeutung side badge column remains partialW1S1
• Landing page backend route audit and nightly health check list are incompleteS1

• AgentFinish implementing and verifying the Bedeutung side badge column in the viewerW1S1
• AgentCollect deep backend URLs across landing pages to verify route healthS1

Updating OCR pipelines, collection configs, and piloting figure and sidecar integrations

Scripts and configurations across OCR, harvest, and page segments were updated within the last four hours. The collection is actively used as a pilot for figure reference indexing and sidecar dossier integration. Overall implementation under the primary work item remains partial.

• Work item for Argard dossier history video and karman-context collection is partially implemented and pending completionW1

• AgentComplete the remaining implementation tasks for the Argard dossier history and collectionW1

Gated scholarly editorial pipeline updating model routing and viewer components

The project develops an eight-stage gated editorial pipeline for arXiv publication. Recent changes updated model routing standards and viewer goals to adhere to updated model instructions. Active development continues across standards and viewer modules.

• AgentVerify all lektorat roles and Partitur runtimes default to gemini-3.8-flash.M1F

Developing query resolution pipelines, classification subagents, and evaluation scripts

The ai-query agent coordinates natural-language query resolution across scholarly collections via Partitur and FastAPI. Active development continues across assessment storage, query extraction, classification subagents, and test loops. Gemini 3.8-flash is now the mandatory default runtime across all agent stages.

• AgentConfigure gemini-3.8-flash runtime across all new Partitur stages and monitor workflows.M1

Workspace language hub updated with Sprachformen configuration and sidecar integration support

LWgrammar provides central reference and rule-checking services across the workspace based on Wittgenstein's philosophy and OED entry models. Recent work updated language forms and expression normalization guides, alongside integration into the famulus-proof sidecar engine.

Building and evaluating fast NLI classifiers and Famulus Normal Form on Karman Nachlass sentences

The project benchmarks logical relation classifiers against sentences from PDAI collections, focusing on fast DuckDB inference and local MLX or mDeBERTa execution. mDeBERTa achieved 0.836 accuracy on XNLI test pairs and scored 114 Karman triples, but close-miss identification remains weak. Recent work touches cascade architectures, cloud judges, and Famulus Normal Form implementations.

• Close-miss trap rate sits at roughly 0.13, with 56 of 114 Karman close misses misclassified.S2W1
• Classification accuracy drops on non-English pairs, reaching 0.72 in French and 0.75 in Hungarian.S2W1
• Famulus Normal Form NLI plan has uncompleted to-dos for baseline-freeze, gold-needle, rules, compilers, and duckdb-core.S1
• Gemini 3.8 Flash hung on structured output during earlier drafting runs.S2

• AgentReduce close-miss errors and non-English failures on Karman relation evaluations.W1
• YouHuman-verify the gold set benchmark (gold-v2).W1S2
• AgentImprove Latin person coverage in the sentence extraction pipeline.W1
• AgentComplete remaining Famulus Normal Form to-dos including fnf-spec, dictionary, and duckdb-core integration.S1

Multi-engine OCR runner and table extraction pipeline for schema-first collections

The project provides shared OCR client tooling and multi-engine table extraction for PDAI collections. Recent work updated page census and specialized table-reading modules. Table extraction pairs Kimi and Gemini readings with Vision line geometry and an arbiter.

Inverted lemma index builder with active register and sidecar development

The package builds Lemma-to-Page inverted indices using DuckDB and Parquet. Active development continues across register modules and tests, with recent file activity within the last 24 hours. A prior dictionary DuckDB lock issue on SMB shares was resolved by pointing lwgrammar to a local database.

• AgentKeep lwgrammar databases off SMB for read-write useW1
• AgentHarden Glider startup so SMB health probes cannot time outW1

Hermeneutic expert packages under active development with pending normaliser learning tasks

Recent edits updated astronomical-expert, calendar-expert, and expert-core packages within the last 24 hours. An agent session restarted the Nachlass viewer on port 5725 and scheduled learning workflows. Multiple tasks around relevanz code, probe context, and learning rounds remain open.

• Incomplete todos for audit-remove, relevanz-yaml, relevanz-code, probe-context, heldout, guard-gate, learn-rounds, and partitur-skillS1
• Unexecuted plans for general normaliser learning, Famulus core and Erläuterung, and probe lab focusS1

• AgentImplement open todos for relevanz-yaml, relevanz-code, and probe-contextS1
• AgentProceed with general normaliser learning and learn-rounds workflowsS1

Viewer Bedeutung column and sidecar integration built, requiring collection propagation and landing page updates

The Kármán viewer received a new Bedeutung column for tiered passage reading alongside sidecar integration for person dossiers. A standing owner rule requires all developments made on karman-nachlass to propagate to eulerOO and enestroem. Both the Kármán-Nachlass and Famulus Elucidation landing pages are currently out of sync with recent changes.

• Landing page 'Kármán-Nachlass' has six uncommitted changes in repository karman-2026 and serves an outdated index.html.L
• Landing page 'Famulus Elucidation' is outdated after sixteen metrics changed since the last build.L
• Developments built on karman-nachlass must be propagated to architecture, sidecars, and DuckDB of eulerOO and enestroem.M2

• AgentCommit changes in karman-2026 and publish the updated Kármán-Nachlass landing page.L
• AgentRebuild the Famulus Elucidation landing page to reflect updated fact counts.L
• AgentPropagate new sidecar, DuckDB schema, and viewer developments to eulerOO and enestroem.M2

Building DuckLake schemas and extensions for reference collections and dossier integration

The package establishes a unified DuckLake and MCP server for PDAI collections with stable fr: addresses. Recent commits added dossier and adart extensions and schemas. PDAI-1 mount drops on openscience previously interrupted workflow and require manual credential handling.

• SMB mounts to PDAI-1, 2026project, and 2025Berlin dropped on openscience and require Apple ID credentials to remountS1

• YouRemount PDAI-1 share on openscience using Apple ID credentialsS1
• AgentRun test suite to verify new dossier and adart extension tablesF

Harvesting and profiling AGARD and context documents for the PDAI karman-context collection

The new PDAI collection was established and populated with historical material, completing an IA/DTIC harvest of 560 PDFs (7.7 GB) alongside Wayback imports. A background profiling run across approximately 1,000 PDFs was initiated to measure page counts and text layers before OCR planning. Work item W1 is partially implemented with verification pending.

• Background profiling run of approximately 1,000 PDFs is pending completionS1
• OCR processing plan with Kimi Ollama engine and learning Partitur not yet finalizedS1
• Wayback harvest reached 269 of 401 target documents and awaits completion and manifest checkS1

• AgentFix document groups based on profiling results and write the OCR planS1
• AgentStart OCR pilots with Kimi engine on Ollama across document groupsS1
• AgentFinish Wayback harvest and build the final import manifestS1

AI-powered OCR and document structuring pipeline using the fastapple backbone

ai-scholar version 3.3.0 turns source page images into structured PDAI XML and figure manifests via the fastapple pipeline. Development is active with recent updates to the pipeline code, CLI, and letter schemas. Work progressed within the last 13 hours.

Research focus on post-war German fashion and pattern magazines (1948–1959)

The research repository was structured with findings on dating, library holdings, copyright limits, and feasibility. The companion digital edition was moved to scholarly/historyFashionDesign and linked from this research folder. The magazines themselves are not yet digitized into a full research corpus.

• Magazines are not yet digitized as a full research corpus beyond the initial six cover scansD1
• Open feasibility questions tracked in machbarkeit/OFFEN.md remain to be addressedF

• AgentReview open points listed in machbarkeit/OFFEN.mdF
• YouEvaluate library holdings and sources to expand digitization beyond the initial six coversD1F

Digital edition of post-war German fashion and pattern periodicals

The project develops a trilingual digital edition of post-war German fashion magazines. The exhibition portal index.html and individual pages in pages/ were updated yesterday. No blockers or requests are pending.

Autopsy and mathematical fitting of sundial drawings in Georg Hartmann's Collectio figurarum

The measurement pipeline analyzes sheet scans (057, 062, 133) from BSB Rar. 434 and renders Typst PDF reports. Recent work verified the pivot and calendar ring on scan 57 and fit horizontal dial hour lines to 49° latitude on scan 133. Several planned tasks remain open from the recent analysis session.

• Incomplete plan items from sheet analysis session: pilot-58, pairs, and work-pdfS1

• AgentImplement remaining autopsy plan tasks for pilot-58, pairs, and work-pdfS1
• AgentComplete autopsy reporting for sheet 062 where pole height inscription is absentS1

Research project on Georg Hartmann's instrument collection and figure reference integration

The project investigates Georg Hartmann's sundial and instrument collection from BSB Munich (Rar. 434) across 324 scans. Recent commits established the project structure, documentation, import manifests, and DTD review. Hartmann is also serving as a pilot collection for citable figure references in famulus-proof.

Intake, OCR configuration, and figure extraction for Georg Hartmann's Collectio figurarum

Hartmann collection intake for BSB Rar. 434 is underway, superseding the Hartmann1500 project. Recent work updated OCR execution scripts, model prompts, and cure configurations, alongside piloting figure reference extraction. Page DTD and segmentation plans remain in draft status.

• config/page.dtd is still in draft v0.1 as a wrapper around page.dtd v4.0D1
• config/segments.yaml segment plan covering sheets, prints, instruments, and inscriptions remains a draftD1

• AgentFinalize draft specifications in config/page.dtd and config/segments.yamlD1
• AgentRun finish_collection.py and hartmann_ocr_run.py to advance intake processingF

Drafting and Typst template configuration for the Famulus normalisation arXiv paper

The repository holds the paper draft and arXiv Typst extension templates. Thirteen files were updated over the past seven days, with the most recent changes made two days ago. No sessions, open issues, or pending decisions are recorded.

Famulus expression normalisation model for mentions, dossiers, time, and keys in DuckDB

The package establishes a statement model for Famulus expression normalisation in a DuckDB database per collection. Fifty-one files across documentation, SQL schemas, and source code were updated within the last seven days. Activity was last recorded two days ago with no pending blockers.

Developing the calculation-tools arXiv paper with Quarto and Typst extension templates

Twenty files changed over the past week, with the latest edits made two days ago. Updates centered on the calculation-tools source documents and the arXiv Typst template extension.

arXiv paper preparation and Typst template setup for famulus-probe-lab

Thirteen files were updated three days ago, including the paper manuscript in Typst and Quarto formats alongside arXiv template extensions. No file activity occurred in the past 24 hours. No pending tasks or open issues are recorded in the dossier.

Corpus letter files and PDF manifest updated six days ago with no current activity

Ninety-three files, including corpus letters and the PDF manifest, were updated six days ago. No file modifications have occurred in the last 24 hours. The dossier records no active blockers or open session tasks.

Measurement pipeline, object database, and 3D animator for ancient conic sundials

The project is at version 0.2.0 with tools for sundial measurement, 3D viewing, and object database building. No file modifications have occurred in the last seven days following 3,301 changes over the past month.

Regularity-theoretic causal reasoning engine and hybrid publication module

The engine implements Graßhoff/May regularity theory and minimal theory search for the Erasmus-Science package. The repository has been inactive for 16 days following 82 file changes over the last month.

Kirch collection scripts and volume document profiles inactive for 19 days

The Kirch collection has had no file modifications in the last seven days. The last activity was 19 days ago across volume profiles and execution scripts. No active runs or pending tasks are documented.

Kepler Commission history paper with Quarto and Typst outputs

The project contains Quarto sources, Typst files, and rendered HTML output for the Kepler Commission history paper and facsimile documents. There has been no file activity over the past 22 days.