Add cross-repo links and data source documentation

- Add Data Sources section with canonical R2 URLs
- Add Related Repositories table
- Part of MVP cleanup strategy (issue #49)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Author: rdhyee

README.md

@@ -51,3 +51,23 @@ The generated docs are placed under `models/generated/vocabularies`
 After editing, push the sources to GitHub. The rendered pages are generated using the `Render using Quarto and push to GH-pages` GitHub action that is currently manually triggered.
 
 Updating dependencies using `pip -U <<package name>>` and regenerate `requirements.txt` with `pip freeze > requirements.txt`.
+
+## Data Sources
+
+All tutorials query parquet files hosted on Cloudflare R2:
+
+```javascript
+// Wide format (recommended) - 280 MB, 20M rows
+const WIDE_URL = "https://pub-a18234d962364c22a50c787b7ca09fa5.r2.dev/isamples_202601_wide.parquet";
+
+// Narrow format (advanced) - 850 MB, 106M rows
+const NARROW_URL = "https://pub-a18234d962364c22a50c787b7ca09fa5.r2.dev/isamples_202512_narrow.parquet";
+```
+
+## Related Repositories
+
+| Repo | Purpose | Start Here |
+|------|---------|------------|
+| [isamplesorg-metadata](https://github.com/isamplesorg/metadata) | Schema definition (8 types, 14 predicates) | `src/schemas/isamples_core.yaml` |
+| [isamples-python](https://github.com/isamplesorg/examples) | Jupyter examples (DuckDB + Lonboard) | `examples/basic/isamples_explorer.ipynb` |
+| [vocabularies](https://github.com/isamplesorg/vocabularies) | SKOS vocabulary terms | Material types, context categories |

SESSION_NOTES.md

@@ -0,0 +1,116 @@
+# Session Continuity Notes - Oct 3, 2025
+
+## Current Status: Phase 2 Complete ✅
+
+We're in the middle of integrating Eric Kansa's 4 query functions from `oc_parquet_analysis_enhanced.ipynb` into `/tutorials/parquet_cesium.qmd`.
+
+### Completed Work
+
+**Phase 1: Documentation** (Commit d5d6690)
+- ✅ Added comprehensive Path 1/Path 2 explanation with diagrams
+- ✅ Added full relationship map (Agent, IdentifiedConcept paths)
+- ✅ Added Eric's 4 query function analysis with summary table
+
+**Phase 2: First Query Implementation** (Commit 3224eb1)
+- ✅ Implemented `get_samples_at_geo_cord_location_via_sample_event()`
+- ✅ Combines Path 1 + Path 2 with UNION
+- ✅ Returns rich metadata: thumbnail_url, description, alternate_identifiers, site info
+- ✅ Added reactive cell `selectedSamplesCombined` with loading state
+- ✅ Added display section "Combined Samples at Location"
+
+### What's Next: Phase 3
+
+**Remaining 2 queries from Eric's notebook** (cell 59):
+
+1. **Agent Query** - `get_sample_data_agents_sample_pid(sample_pid)`
+   - Shows who collected/registered samples
+   - Path: MaterialSampleRecord → produced_by → SamplingEvent → {responsibility, registrant} → Agent
+   - Returns: sample metadata + agent info (agent_pid, agent_name, predicate)
+   - **Note**: Independent of Path 1/Path 2 (no geographic data needed)
+
+2. **Keywords/Concepts Query** - `get_sample_types_and_keywords_via_sample_pid(sample_pid)`
+   - Shows material types and classifications
+   - Path: MaterialSampleRecord → {keywords, has_sample_object_type, has_material_category} → IdentifiedConcept
+   - Returns: sample metadata + concept info (keyword_pid, keyword, predicate)
+   - **Note**: Direct edges to concepts, bypasses SamplingEvent entirely!
+
+### Key Technical Details
+
+**Table alias difference**:
+- Eric's Python notebook uses: `FROM pqg AS ...`
+- Our JavaScript implementation uses: `FROM nodes ...`
+- DuckDB view is created as: `CREATE VIEW nodes AS SELECT * FROM read_parquet('${parquet_path}')`
+
+**Loading pattern to follow**:
+```javascript
+async function get_FUNCTION_NAME(pid) {
+    if (pid === null || pid ==="" || pid == "unset") return [];
+    const q = `SQL QUERY HERE`;
+    const result = await loadData(q, [pid], "loading_ID", "key");
+    return result ?? [];
+}
+
+mutable FUNCTIONLoading = false;
+
+selectedFUNCTION = {
+    mutable FUNCTIONLoading = true;
+    try {
+        return await get_FUNCTION_NAME(clickedPointId);
+    } finally {
+        mutable FUNCTIONLoading = false;
+    }
+}
+```
+
+**Display pattern**:
+```markdown
+## Section Title
+
+<div id="loading_ID" hidden>Loading message…</div>
+
+Explanation text...
+
+\`\`\`{ojs}
+//| echo: false
+variable = selectedFUNCTION
+FUNCTIONLoading ? md`(loading…)` : md`\`\`\`
+${JSON.stringify(variable, null, 2)}
+\`\`\`
+`
+\`\`\`
+```
+
+### File Locations
+
+**Main working file**: `/Users/raymondyee/C/src/iSamples/isamplesorg.github.io/tutorials/parquet_cesium.qmd`
+
+**Reference notebook**: `/Users/raymondyee/C/src/iSamples/isamples-python/examples/basic/oc_parquet_analysis_enhanced.ipynb` (cell 59)
+
+**Local parquet**: `http://localhost:4979/assets/oc_isamples_pqg.parquet` (691MB, in `docs/assets/`)
+
+**Branch**: `issue-13-parquet-duckdb`
+
+### Quick Start Commands
+
+```bash
+# Navigate to project
+cd /Users/raymondyee/C/src/iSamples/isamplesorg.github.io
+
+# Check current branch
+git status
+
+# Start Quarto preview (if needed)
+quarto preview
+
+# View notebook for reference
+code /Users/raymondyee/C/src/iSamples/isamples-python/examples/basic/oc_parquet_analysis_enhanced.ipynb
+```
+
+### Session Pickup Prompt
+
+"Let's continue integrating Eric's queries into parquet_cesium.qmd. We completed Phase 2 (combined samples query). Next we need to add the agent query and keywords/concepts query. Should we start with the agent query?"
+
+### Notes
+- All existing queries (`get_samples_1`, `get_samples_2`, `get_samples_at_geo_cord_location_via_sample_event`) are working and preserved
+- Pattern is established - just need to adapt Eric's remaining 2 SQL queries to JavaScript
+- Consider UI improvements after Phase 3 complete (tables, clickable links, thumbnails)