docs: regen card with normal hyphens

README.md

@@ -37,7 +37,7 @@ canonical schema across four entities: `cell_metadata` + `test_metadata`
 (JSON), `timeseries` + `cycle_summary` (Parquet). Consumers read one format
 instead of writing per-source loaders.
 
-**Scope: harmonization only.** celljar focuses on measurements — unit
+**Scope: harmonization only.** celljar focuses on measurements - unit
 conversion and schema normalization. It deliberately leaves fitting and
 modeling to downstream tools that specialize in those steps.
 
@@ -58,10 +58,10 @@ cycle_summary.parquet     # per-cycle aggregates (aging studies); join on (test_
 
 Four entities; field list generated from the authoritative [JSON Schemas](https://github.com/mihnathul/celljar/tree/main/schemas):
 
-- **`cell_metadata`** (JSON, one file per cell) — `cell_id`*, `source`*, `source_cell_id`, `manufacturer`, `model_number`, `chemistry`*, `cathode`, `anode`, `electrolyte`, `form_factor`*, `nominal_capacity_Ah`, `nominal_voltage_V`, `max_voltage_V`, `min_voltage_V`
-- **`test_metadata`** (JSON, one file per test) — `test_id`*, `cell_id`*, `test_type`*, `temperature_C_min`, `temperature_C_max`, `soc_range_min`, `soc_range_max`, `soc_step`, `c_rate_charge`, `c_rate_discharge`, `protocol_description`, `num_cycles`, `soh_pct`, `soh_method`, `cycle_count_at_test`, `test_year`, `source_doi`, `source_url`, `source_citation`, `source_license`, `source_license_url`, `n_samples`, `duration_s`, `voltage_observed_min_V`, `voltage_observed_max_V`, `current_observed_min_A`, `current_observed_max_A`, `temperature_observed_min_C`, `temperature_observed_max_C`, `sample_dt_min_s`, `sample_dt_median_s`, `sample_dt_max_s`
-- **`timeseries`** (Parquet, one row per measurement sample) — `test_id`*, `cycle_number`*, `step_number`, `step_type`, `timestamp_s`*, `voltage_V`, `current_A`, `temperature_C`, `coulomb_count_Ah`, `energy_Wh`, `displacement_um`
-- **`cycle_summary`** (Parquet, one row per cycle / aging checkpoint) — `test_id`*, `cell_id`*, `cycle_number`, `equivalent_full_cycles`, `elapsed_time_s`, `capacity_Ah`, `capacity_retention_pct`, `resistance_dc_ohm`, `resistance_dc_pulse_duration_s`, `resistance_dc_soc_pct`, `energy_Wh`, `coulombic_efficiency`, `temperature_C_mean`
+- **`cell_metadata`** (JSON, one file per cell) - `cell_id`*, `source`*, `source_cell_id`, `manufacturer`, `model_number`, `chemistry`*, `cathode`, `anode`, `electrolyte`, `form_factor`*, `nominal_capacity_Ah`, `nominal_voltage_V`, `max_voltage_V`, `min_voltage_V`
+- **`test_metadata`** (JSON, one file per test) - `test_id`*, `cell_id`*, `test_type`*, `temperature_C_min`, `temperature_C_max`, `soc_range_min`, `soc_range_max`, `soc_step`, `c_rate_charge`, `c_rate_discharge`, `protocol_description`, `num_cycles`, `soh_pct`, `soh_method`, `cycle_count_at_test`, `test_year`, `source_doi`, `source_url`, `source_citation`, `source_license`, `source_license_url`, `n_samples`, `duration_s`, `voltage_observed_min_V`, `voltage_observed_max_V`, `current_observed_min_A`, `current_observed_max_A`, `temperature_observed_min_C`, `temperature_observed_max_C`, `sample_dt_min_s`, `sample_dt_median_s`, `sample_dt_max_s`
+- **`timeseries`** (Parquet, one row per measurement sample) - `test_id`*, `cycle_number`*, `step_number`, `step_type`, `timestamp_s`*, `voltage_V`, `current_A`, `temperature_C`, `coulomb_count_Ah`, `energy_Wh`, `displacement_um`
+- **`cycle_summary`** (Parquet, one row per cycle / aging checkpoint) - `test_id`*, `cell_id`*, `cycle_number`, `equivalent_full_cycles`, `elapsed_time_s`, `capacity_Ah`, `capacity_retention_pct`, `resistance_dc_ohm`, `resistance_dc_pulse_duration_s`, `resistance_dc_soc_pct`, `energy_Wh`, `coulombic_efficiency`, `temperature_C_mean`
 
 `*` = required field (others nullable). See [JSON Schemas](https://github.com/mihnathul/celljar/tree/main/schemas) for full type info, enum values, and constraints.
 
@@ -75,7 +75,7 @@ Join keys: `cells.cell_id = tests.cell_id`, `tests.test_id = timeseries.test_id`
 ## Download the whole bundle
 
 ```bash
-# CLI — pulls everything (cells/*.json, tests/*.json, timeseries.parquet, cycle_summary.parquet)
+# CLI - pulls everything (cells/*.json, tests/*.json, timeseries.parquet, cycle_summary.parquet)
 pip install huggingface_hub
 huggingface-cli download mihnathul/celljar --repo-type dataset --local-dir ./celljar-bundle
 
@@ -91,9 +91,9 @@ local = snapshot_download(repo_id="mihnathul/celljar", repo_type="dataset", revi
 print(local)  # local path containing cells/, tests/, timeseries.parquet, cycle_summary.parquet
 ```
 
-## Query in place — no download needed
+## Query in place - no download needed
 
-### DuckDB — full SQL across all entities over HTTPS
+### DuckDB - full SQL across all entities over HTTPS
 
 ```sql
 INSTALL httpfs; LOAD httpfs;
@@ -109,7 +109,7 @@ GROUP BY 1,2,3,4,5
 ORDER BY t.test_id;
 ```
 
-### pandas / Polars — predicate-pushdown read of one test
+### pandas / Polars - predicate-pushdown read of one test
 
 ```python
 import pandas as pd
@@ -119,7 +119,7 @@ df = pd.read_parquet(
 )
 ```
 
-### `datasets` library — streaming
+### `datasets` library - streaming
 
 ```python
 from datasets import load_dataset
@@ -141,7 +141,7 @@ the data, and, if it's helpful, celljar alongside.
 
 - **This harmonized bundle** (packaging, schema, derived test-metadata fields):
   [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/).
-- **Upstream raw data** retains each publisher's original license — listed
+- **Upstream raw data** retains each publisher's original license - listed
   per-source below. Each source's license terms apply when you use its tests.
 
 To make attribution easy, every `tests/*.json` row carries its own