diff --git a/TODO.md b/TODO.md
new file mode 100644
index 0000000..c602830
--- /dev/null
+++ b/TODO.md
@@ -0,0 +1,113 @@
+# HAGFISH Project todo
+
+`GET statbank.hagstova.fo/api/v1/fo/H2/VV/VV01/fisknv_md.px — metadata`
+POST same URL — JSON-stat2 query
+8 dimensions, "-" → NULL
+
+## Ingestion
+
+Monthly run (cron or systemd timer, user's choice)
+Incremental with full backfill option
+GET metadata → build lookup maps → POST query → parse → insert into DuckDB
+Upsert semantics for revised months
+
+## Storage 
+DuckDB on disk
+
+## Fact table
+landings(species_code, gear_code, zone_code, processing_code, preservation_code, shipsize_code, month, mass_kg, value_kr)
+Lookup tables: species, gear, zone, processing, preservation, shipsize
+Parquet export per run
+
+## API (Axum)
+
+GET /api/species — list species codes + Faroese names
+GET /api/landings — filtered query, JSON response
+GET /api/summary — aggregates
+GET /api/export.parquet — download Parquet
+Serve embedded static frontend
+
+## Frontend 
+JS + ECharts
+
+Line chart, stacked bar, donut, dropdown filters
+Plain HTML/CSS/JS, no build step
+
+## Deployment
+
+Statically linked Rust binary
+config.json for settings (DuckDB path, bind addr, data source URL)
+Systemd timer for monthly ingestion (bare metal, no containers)
+
+
+## Project TODO
+
+hagfish/
+├── Cargo.toml
+├── Taskfile.yml
+├── config.json
+├── static/
+│   ├── index.html
+│   ├── app.js
+│   └── style.css
+└── src/
+    ├── main.rs
+    ├── ingest.rs
+    ├── db.rs
+    ├── api.rs
+    └── types.rs
+
+### Phase 1: Types & Ingestion
+
+- [ ] 1.1 Define types in types.rs: MetadataResponse, VariableMeta, DataResponse, DataRow, Query, Selection, QueryItem, Config — all with serde derives
+- [ ] 1.2 Implement ingest.rs::fetch_metadata(url) — GET request, parse JSON, return HashMap<(variable_code, value_code), faroese_label>
+- [ ] 1.3 Implement ingest.rs::build_query(months: &[String]) — construct POST body with all species/gear/zones set to "*", processing/preservation/shipsize set to TOTAL, measure set to both MASS and VALUE
+- [ ] 1.4 Implement ingest.rs::fetch_data(url, query) — POST request, parse JSON-stat2 response, return Vec<DataRow>
+- [ ] 1.5 Implement ingest.rs::parse_row(row, lookup_maps) — decode key[] positions into labeled Landing struct. Handle "-" → None.
+- [ ] 1.6 Write unit tests: mock JSON-stat2 response, verify key-to-label mapping, verify "-" handling, verify Faroese Unicode characters in species names (ð, á, í, ý, ø, ó)
+
+### Phase 2: DuckDB Storage
+
+- [ ] 2.1 Add duckdb crate dependency (bundled feature)
+- [ ] 2.2 Implement db.rs::init(path) — create tables: landings fact table + 6 lookup tables. Add indexes on month, species_code.
+- [ ] 2.3 Implement db.rs::upsert_landings(rows) — batch insert with delete+insert per month or INSERT OR REPLACE
+- [ ] 2.4 Implement db.rs::update_lookups(metadata) — populate lookup tables from metadata response
+- [ ] 2.5 Implement db.rs::get_last_month() — query max month from landings table for incremental ingestion
+- [ ] 2.6 Implement db.rs::export_parquet(path) — COPY landings TO 'path' (FORMAT PARQUET) partitioned by month
+- [ ] 2.7 Write integration tests: init in-memory DB, insert sample rows, query back, verify NULL handling
+
+### Phase 3: API (Axum)
+
+- [ ] 3.1 Set up Axum router in main.rs with Tokio runtime. AppState holds DuckDB connection wrapped in Mutex.
+- [ ] 3.2 Implement GET /api/species — query species lookup table, return JSON array
+- [ ] 3.3 Implement GET /api/landings — parse query params, build DuckDB SQL with WHERE clauses. Support: months, species, gear, zone, measure filters.
+- [ ] 3.4 Implement GET /api/summary — aggregate query: total mass + value by month, top 10 species by value, price/kg trend
+- [ ] 3.5 Implement GET /api/export.parquet — generate and stream Parquet via DuckDB COPY
+- [ ] 3.6 Implement GET /healthz
+- [ ] 3.7 Serve static files via rust-embed
+- [ ] 3.8 Write API tests
+
+### Phase 4: Frontend
+
+- [ ] 4.1 index.html — dropdown filters (species, zone, gear, month range) and 3 chart containers
+- [ ] 4.2 app.js — fetch species list on load, populate dropdowns, fetch /api/landings, render charts
+- [ ] 4.3 ECharts line chart: x=month, y=mass/value toggle
+- [ ] 4.4 ECharts stacked bar: x=month, y=value by species (top 10 + "other")
+- [ ] 4.5 ECharts donut: species distribution for selected month
+- [ ] 4.6 Loading states, error handling, empty state
+- [ ] 4.7 Responsive layout, plain CSS
+
+### Phase 5: CLI & Scheduling
+
+- [ ] 5.1 Add clap derive subcommands: hagfish ingest [--full] and hagfish serve
+- [ ] 5.2 Implement incremental logic: read get_last_month(), compute remaining months from metadata, fetch in batches if >12 months
+- [ ] 5.3 Log ingestion runs with slog (rows inserted, duration, errors)
+- [ ] 5.4 Add hagfish export --out /path/to/parquet subcommand
+- [ ] 5.5 Load config.json on startup (DuckDB path, bind address, data source URL, log file path)
+
+### Phase 6: Bare Metal Deployment
+
+- [ ] 6.1 Write systemd service unit file (hagfish.service) — ExecStart=/usr/local/bin/hagfish serve, restart policy
+- [ ] 6.2 Write systemd timer (hagfish-ingest.timer + hagfish-ingest.service) — monthly, runs hagfish ingest
+- [ ] 6.3 Taskfile: build (release, static), deploy (rsync binary + config + units, ssh reload)
+- [ ] 6.4 README with ELI5 Technology Choices section (why DuckDB, why Rust, why embedded static assets)
diff --git a/src/ingest.rs b/src/ingest.rs
new file mode 100644
index 0000000..5541eee
--- /dev/null
+++ b/src/ingest.rs
@@ -0,0 +1,509 @@
+//! Data ingestion from PX-Web API.
+//!
+//! Handles metadata fetching, query construction, data retrieval, and
+//! parsing of JSON-stat2 responses into structured rows.
+
+use crate::types::*;
+use reqwest::Client;
+use tracing::{debug, info, warn};
+use std::collections::HashMap;
+
+const PX_WEB_LANGUAGE: &str = "fo"; // Faroese labels
+
+/// Fetches metadata from PX-Web API endpoint.
+///
+/// Returns a HashMap mapping variable_id → code → label mappings.
+/// This lookup is used to decode opaque codes in data responses.
+pub async fn fetch_metadata(client: &Client, url: &str) -> Result<LookupMap> {
+    info!("Fetching metadata from {}", url);
+    
+    let resp = client
+        .get(url)
+        .header("Accept", "application/json")
+        .send()
+        .await?;
+    
+    if !resp.status().is_success() {
+        return Err(IngestError::HttpError(reqwest::Error::from(
+            std::io::Error::new(
+                std::io::ErrorKind::Other,
+                format!("API returned status: {}", resp.status()),
+            ),
+        )));
+    }
+    
+    let meta: MetadataResponse = resp.json().await?;
+    debug!("Received {} variables from metadata", meta.variables.len());
+    
+    // Build lookup map from metadata
+    let mut lookup_map: LookupMap = HashMap::new();
+    
+    for var in &meta.variables {
+        let codes = meta.values.get(&var.id);
+        if let Some(codes) = codes {
+            let labels: HashMap<String, String> = codes
+                .iter()
+                .map(|vm| (vm.code.clone(), vm.text.clone()))
+                .collect();
+            lookup_map.insert(var.id.clone(), labels);
+            info!("Loaded {} codes for variable '{}'", labels.len(), var.id);
+        } else {
+            warn!("No codes found for variable '{}'", var.id);
+        }
+    }
+    
+    Ok(lookup_map)
+}
+
+/// Constructs a query body for fetching all landing data.
+///
+/// Sets all categorical dimensions to "*" (all values) and measures to MASS+VALUE.
+/// Used for full backfill ingestion.
+pub fn build_query(all_months: &[String]) -> Query {
+    Query {
+        query: vec![
+            QueryItem {
+                id: "Tid".to_string(),      // Month
+                values: all_months.to_vec(),
+            },
+            QueryItem {
+                id: "Art".to_string(),     // Species
+                values: vec!["*".to_string()],
+            },
+            QueryItem {
+                id: "Redskab".to_string(), // Gear
+                values: vec!["*".to_string()],
+            },
+            QueryItem {
+                id: "Økonomisk zone".to_string(),
+                values: vec!["*".to_string()],
+            },
+            QueryItem {
+                id: "Tilstand".to_string(), // Processing
+                values: vec!["TOTAL".to_string()],
+            },
+            QueryItem {
+                id: "Konservering".to_string(), // Preservation
+                values: vec!["TOTAL".to_string()],
+            },
+            QueryItem {
+                id: "Skibsstørrelse".to_string(), // Vessel size
+                values: vec!["TOTAL".to_string()],
+            },
+            QueryItem {
+                id: "Måleenhed".to_string(), // Measure
+                values: vec!["MASS".to_string(), "VALUE".to_string()],
+            },
+        ],
+        language: PX_WEB_LANGUAGE.to_string(),
+    }
+}
+
+/// Fetches data from PX-Web API using the provided query.
+pub async fn fetch_data(client: &Client, url: &str, query: &Query) -> Result<DataResponse> {
+    info!("Sending data query for {} month(s)", query.query[0].values.len());
+    
+    let resp = client
+        .post(url)
+        .json(query)
+        .send()
+        .await?;
+    
+    if !resp.status().is_success() {
+        let body = resp.text().await.unwrap_or_default();
+        warn!("API error response: {}", body);
+        return Err(IngestError::HttpError(reqwest::Error::from(
+            std::io::Error::new(
+                std::io::ErrorKind::Other,
+                format!("API returned status: {}", resp.status()),
+            ),
+        )));
+    }
+    
+    let data: DataResponse = resp.json().await?;
+    
+    if data.dataset.value.is_empty() {
+        return Err(IngestError::EmptyDataset);
+    }
+    
+    info!(
+        "Retrieved {} data points from API",
+        data.dataset.value.len()
+    );
+    
+    Ok(data)
+}
+
+/// Extracts dimension codes from a flat row index in JSON-stat2 format.
+///
+/// JSON-stat2 uses a flattened array where each element corresponds to a unique
+/// combination of dimension values. The dimension.keys array tells us how many
+/// values exist per dimension. We decode the flat index into per-dimension indices.
+fn decode_key_indices(flat_index: usize, key_counts: &[usize]) -> Vec<usize> {
+    let mut indices = Vec::with_capacity(key_counts.len());
+    let mut remaining = flat_index;
+    
+    // Process dimensions in reverse (last dimension varies fastest)
+    for count in key_counts.iter().rev() {
+        indices.push((remaining % count) as usize);
+        remaining /= count;
+    }
+    
+    indices.reverse(); // Restore original order
+    indices
+}
+
+/// Parses a single row from JSON-stat2 format into a DataRow.
+///
+/// # Arguments
+/// * `row_index` - Index into the dataset's value array
+/// * `dataset` - The complete dataset response
+/// * `lookup_maps` - Code → label mappings for each dimension
+///
+/// # Errors
+/// Returns an error if the row_index exceeds bounds or required lookups are missing.
+pub fn parse_row(
+    row_index: usize,
+    dataset: &DataResponse,
+    lookup_maps: &LookupMap,
+) -> Result<DataRow> {
+    let dim_info = &dataset.dataset.dimension;
+    let categories = &dim_info.category;
+    let keys = &dim_info.keys;
+    let value = dataset.dataset.value.get(row_index).copied().flatten();
+    
+    // Validate bounds
+    if row_index >= dataset.dataset.value.len() {
+        return Err(IngestError::InvalidValueCode(format!(
+            "Row index {} out of bounds (max {})",
+            row_index,
+            dataset.dataset.value.len()
+        )));
+    }
+    
+    // Get category label lists for each dimension (in key order)
+    let category_lists: Vec<Vec<&str>> = keys
+        .iter()
+        .filter_map(|k| {
+            categories.label.get(k).map(|labels| {
+                labels.iter().map(|s| s.as_str()).collect::<Vec<_>>()
+            })
+        })
+        .collect();
+    
+    if category_lists.len() != keys.len() {
+        return Err(IngestError::MissingDimension(
+            format!(
+                "Category count mismatch: {} keys vs {} category lists",
+                keys.len(),
+                category_lists.len()
+            )
+        ));
+    }
+    
+    // Decode flat index into per-dimension indices
+    let key_counts: Vec<usize> = category_lists.iter().map(|c| c.len()).collect();
+    let indices = decode_key_indices(row_index, &key_counts);
+    
+    // Map indices back to actual dimension values
+    let dimension_values: Vec<&str> = indices
+        .into_iter()
+        .zip(category_lists.iter())
+        .map(|(idx, list)| list[idx])
+        .collect();
+    
+    // Expected 8 dimensions: [month, species, gear, zone, processing, preservation, shipsize, measure]
+    const EXPECTED_DIMS: usize = 8;
+    if dimension_values.len() != EXPECTED_DIMS {
+        return Err(IngestError::MissingDimension(format!(
+            "Expected {} dimensions, got {}",
+            EXPECTED_DIMS,
+            dimension_values.len()
+        )));
+    }
+    
+    let [month_code, species_code, gear_code, zone_code, processing_code, preservation_code, shipsize_code, measure_code]: [&str; 8] = 
+        dimension_values.try_into().unwrap();
+    
+    // Helper to safely get label from lookup map
+    fn get_label<'a>(maps: &'a LookupMap, var_name: &str, code: &str) -> &'a str {
+        maps.get(var_name)
+            .and_then(|m| m.get(code))
+            .map(|s| s.as_str())
+            .unwrap_or(code)
+    }
+    
+    let data_row = DataRow {
+        month: month_code.to_string(),
+        species_code: species_code.to_string(),
+        species_label: get_label(lookup_maps, "Art", species_code).to_string(),
+        gear_code: gear_code.to_string(),
+        gear_label: get_label(lookup_maps, "Redskab", gear_code).to_string(),
+        zone_code: zone_code.to_string(),
+        zone_label: get_label(lookup_maps, "Økonomisk zone", zone_code).to_string(),
+        processing_code: processing_code.to_string(),
+        processing_label: get_label(lookup_maps, "Tilstand", processing_code).to_string(),
+        preservation_code: preservation_code.to_string(),
+        preservation_label: get_label(lookup_maps, "Konservering", preservation_code).to_string(),
+        shipsize_code: shipsize_code.to_string(),
+        shipsize_label: get_label(lookup_maps, "Skibsstørrelse", shipsize_code).to_string(),
+        measure_code: measure_code.to_string(),
+        measure_label: get_label(lookup_maps, "Måleenhed", measure_code).to_string(),
+        value, // Already handled None case above
+    };
+    
+    Ok(data_row)
+}
+
+/// Converts DataRow to Landing struct for database insertion.
+/// Removes redundant fields that aren't needed in the fact table.
+pub fn data_row_to_landing(row: &DataRow) -> Landing {
+    Landing {
+        month: row.month.clone(),
+        species_code: row.species_code.clone(),
+        species_label: row.species_label.clone(),
+        gear_code: row.gear_code.clone(),
+        zone_code: row.zone_code.clone(),
+        processing_code: row.processing_code.clone(),
+        preservation_code: row.preservation_code.clone(),
+        shipsize_code: row.shipsize_code.clone(),
+        measure_code: row.measure_code.clone(),
+        value: row.value,
+    }
+}
+
+/// Collects all months available in the metadata response.
+/// Used for building incremental ingestion queries.
+pub fn extract_available_months(meta: &MetadataResponse) -> Vec<String> {
+    meta.values
+        .get("Tid")
+        .map(|codes| codes.iter().map(|c| c.code.clone()).collect())
+        .unwrap_or_default()
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::sync::OnceLock;
+
+    static CLIENT: OnceLock<Client> = OnceLock::new();
+
+    fn client() -> &'static Client {
+        CLIENT.get_or_init(Client::new)
+    }
+
+    /// Test fixture: mock JSON-stat2 response with Faroese Unicode characters.
+    /// Simulates a response with 2 months × 2 species × 2 measures = 8 values.
+    fn mock_dataset_response() -> DataResponse {
+        DataResponse {
+            dataset: Dataset {
+                // Keys specify cardinality per dimension (2 months, 2 species, ..., 2 measures)
+                keys: vec![
+                    "2".to_string(), // Time (2 months)
+                    "2".to_string(), // Art (2 species)
+                    "1".to_string(), // Redskab (1 gear - TOTAL)
+                    "1".to_string(), // Zone (1 zone - TOTAL)
+                    "1".to_string(), // Tilstand (1 - TOTAL)
+                    "1".to_string(), // Konservering (1 - TOTAL)
+                    "1".to_string(), // Skibsstørrelse (1 - TOTAL)
+                    "2".to_string(), // Måleenhed (2 - MASS, VALUE)
+                ],
+                category: CategoryInfo {
+                    label: HashMap::from([
+                        (
+                            "0".to_string(),
+                            vec!["2015M01".to_string(), "2015M02".to_string()],
+                        ),
+                        (
+                            "1".to_string(),
+                            vec!["Sild".to_string(), "Þorskur".to_string()],
+                        ),
+                        ("2".to_string(), vec!["TOTAL".to_string()]),
+                        ("3".to_string(), vec!["TOTAL".to_string()]),
+                        ("4".to_string(), vec!["TOTAL".to_string()]),
+                        ("5".to_string(), vec!["TOTAL".to_string()]),
+                        ("6".to_string(), vec!["TOTAL".to_string()]),
+                        (
+                            "7".to_string(),
+                            vec!["MASS".to_string(), "VALUE".to_string()],
+                        ),
+                    ]),
+                    index: None,
+                },
+                value: vec![
+                    Some(1234.5), // 2015M01, Sild, TOTAL..., MASS
+                    Some(2345.6), // 2015M01, Sild, TOTAL..., VALUE
+                    Some(-1.0),   // 2015M01, Þorskur, TOTAL..., MASS (marker)
+                    Some(3456.7), // 2015M01, Þorskur, TOTAL..., VALUE
+                    Some(4567.8), // 2015M02, Sild, TOTAL..., MASS
+                    Some(5678.9), // 2015M02, Sild, TOTAL..., VALUE
+                    None,         // 2015M02, Þorskur, TOTAL..., MASS (missing)
+                    Some(6789.0), // 2015M02, Þorskur, TOTAL..., VALUE
+                ],
+            },
+        }
+    }
+
+    /// Test fixture: mock lookup maps with Faroese labels.
+    fn mock_lookup_maps() -> LookupMap {
+        let mut maps = LookupMap::new();
+        maps.insert(
+            "0".to_string(),
+            HashMap::from([
+                ("2015M01".to_string(), "Januar 2015".to_string()),
+                ("2015M02".to_string(), "Februar 2015".to_string()),
+            ]),
+        );
+        maps.insert(
+            "1".to_string(),
+            HashMap::from([
+                ("Sild".to_string(), "Sild".to_string()),
+                ("Þorskur".to_string(), "Þorskur".to_string()),
+            ]),
+        );
+        maps.insert(
+            "7".to_string(),
+            HashMap::from([
+                ("MASS".to_string(), "Kilo".to_string()),
+                ("VALUE".to_string(), "Krónur".to_string()),
+            ]),
+        );
+        maps
+    }
+
+    #[tokio::test]
+    async fn test_build_query_all_wildcards() {
+        let months = vec!["2015M01".to_string(), "2015M02".to_string()];
+        let query = build_query(&months);
+        
+        assert_eq!(query.language, PX_WEB_LANGUAGE);
+        assert_eq!(query.query.len(), 8); // All 8 dimensions
+        
+        // Check that species is wildcard
+        let species_query = query.query.iter().find(|q| q.id == "Art").unwrap();
+        assert_eq!(species_query.values, vec!["*".to_string()]);
+        
+        // Check that measures includes both MASS and VALUE
+        let measure_query = query.query.iter().find(|q| q.id == "Måleenhed").unwrap();
+        assert!(measure_query.values.contains(&"MASS".to_string()));
+        assert!(measure_query.values.contains(&"VALUE".to_string()));
+    }
+
+    #[test]
+    fn test_decode_key_indices_basic() {
+        // 2 × 2 × 2 = 8 combinations
+        let key_counts = vec![2, 2, 2];
+        
+        // Index 0 should give [0, 0, 0]
+        let indices = decode_key_indices(0, &key_counts);
+        assert_eq!(indices, vec![0, 0, 0]);
+        
+        // Index 7 should give [1, 1, 1]
+        let indices = decode_key_indices(7, &key_counts);
+        assert_eq!(indices, vec![1, 1, 1]);
+        
+        // Index 3 should give [0, 1, 1] (middle combination)
+        let indices = decode_key_indices(3, &key_counts);
+        assert_eq!(indices, vec![0, 1, 1]);
+    }
+
+    #[test]
+    fn test_parse_row_faroese_unicode() {
+        let dataset = mock_dataset_response();
+        let lookup_maps = mock_lookup_maps();
+        
+        // Parse first row (index 0)
+        let row = parse_row(0, &dataset, &lookup_maps).expect("Failed to parse row");
+        
+        // Verify Faroese characters survive intact
+        assert_eq!(row.month, "2015M01");
+        assert_eq!(row.species_code, "Sild");
+        assert!(row.species_label.contains("Sild"));
+        assert_eq!(row.measure_code, "MASS");
+        
+        // Verify value is preserved
+        assert!(row.value.is_some());
+        assert_eq!(row.value.unwrap(), 1234.5);
+    }
+
+    #[test]
+    fn test_parse_row_second_species() {
+        let dataset = mock_dataset_response();
+        let lookup_maps = mock_lookup_maps();
+        
+        // Row index 2 should be Þorskur (second species, first time, MASS)
+        let row = parse_row(2, &dataset, &lookup_maps).expect("Failed to parse row");
+        
+        // Verify the special character survives
+        assert_eq!(row.species_code, "Þorskur");
+        assert!(row.species_label.contains("Þ"));
+        assert!(row.species_label.contains("orskur"));
+    }
+
+    #[test]
+    fn test_parse_row_missing_value_handling() {
+        let dataset = mock_dataset_response();
+        let lookup_maps = mock_lookup_maps();
+        
+        // Row index 6 has None value (missing data for 2015M02, Þorskur, MASS)
+        let row = parse_row(6, &dataset, &lookup_maps).expect("Failed to parse row");
+        
+        // Missing values should become None, not panic
+        assert!(row.value.is_none());
+    }
+
+    #[test]
+    fn test_parse_row_out_of_bounds() {
+        let dataset = mock_dataset_response();
+        let lookup_maps = mock_lookup_maps();
+        
+        // Requesting out-of-bounds index should return error
+        let result = parse_row(100, &dataset, &lookup_maps);
+        assert!(result.is_err());
+        
+        if let Err(IngestError::InvalidValueCode(msg)) = result {
+            assert!(msg.contains("out of bounds"));
+        } else {
+            panic!("Expected InvalidValueCode error");
+        }
+    }
+
+    #[test]
+    fn test_data_row_to_landing() {
+        let dataset = mock_dataset_response();
+        let lookup_maps = mock_lookup_maps();
+        
+        let row = parse_row(0, &dataset, &lookup_maps).expect("Failed to parse row");
+        let landing = data_row_to_landing(&row);
+        
+        // Verify core fields are copied
+        assert_eq!(landing.month, row.month);
+        assert_eq!(landing.species_code, row.species_code);
+        assert_eq!(landing.measure_code, row.measure_code);
+        assert_eq!(landing.value, row.value);
+        
+        // Verify lookup fields are preserved
+        assert_eq!(landing.species_label, row.species_label);
+        assert_eq!(landing.gear_code, row.gear_code);
+        assert_eq!(landing.zone_code, row.zone_code);
+    }
+
+    #[test]
+    fn test_extract_available_months() {
+        let meta = MetadataResponse {
+            variables: vec![],
+            values: HashMap::from([(
+                "Tid".to_string(),
+                vec![
+                    ValueMeta { code: "2015M01".to_string(), text: "Jan 2015".to_string() },
+                    ValueMeta { code: "2015M02".to_string(), text: "Feb 2015".to_string() },
+                ],
+            )]),
+        };
+        
+        let months = extract_available_months(&meta);
+        assert_eq!(months.len(), 2);
+        assert!(months.contains(&"2015M01".to_string()));
+        assert!(months.contains(&"2015M02".to_string()));
+    }
+}
diff --git a/src/types.rs b/src/types.rs
new file mode 100644
index 0000000..f42c1bb
--- /dev/null
+++ b/src/types.rs
@@ -0,0 +1,171 @@
+//! Type definitions for hagfish data structures.
+//!
+//! These types represent the PX-Web API contract and internal data models.
+//! All public types derive Serialize/Deserialize for JSON (de)serialization.
+
+use serde::{Deserialize, Serialize};
+use std::collections::HashMap;
+
+/// Configuration loaded from config.json on startup.
+#[derive(Debug, Clone, Deserialize, Serialize)]
+pub struct Config {
+    pub duckdb_path: String,
+    pub bind_address: String,
+    pub data_source_url: String,
+    pub log_file_path: Option<String>,
+}
+
+impl Default for Config {
+    fn default() -> Self {
+        Self {
+            duckdb_path: "hagfish.db".to_string(),
+            bind_address: "127.0.0.1:8090".to_string(),
+            data_source_url: "https://statbank.hagstova.fo/data/api/table/fisknv_md".to_string(),
+            log_file_path: Some("hagfish.log".to_string()),
+        }
+    }
+}
+
+/// Metadata response from PX-Web API GET request.
+/// Contains variable definitions and value codes with labels.
+#[derive(Debug, Clone, Deserialize, Serialize)]
+pub struct MetadataResponse {
+    pub variables: Vec<VariableMeta>,
+    pub values: HashMap<String, Vec<ValueMeta>>,
+}
+
+/// Variable metadata describing a dimension in the PX-Web table.
+#[derive(Debug, Clone, Deserialize, Serialize)]
+pub struct VariableMeta {
+    pub id: String,
+    pub text: String,
+    pub role: String,
+    /// Position in the key array (0-indexed)
+    #[serde(rename = "keyPosition", skip_serializing_if = "Option::is_none")]
+    pub key_position: Option<usize>,
+}
+
+/// Value metadata: code → display label mapping for a variable.
+#[derive(Debug, Clone, Deserialize, Serialize)]
+pub struct ValueMeta {
+    pub code: String,
+    pub text: String,
+}
+
+/// Query body sent to PX-Web API POST endpoint.
+#[derive(Debug, Clone, Serialize)]
+pub struct Query {
+    pub query: Vec<QueryItem>,
+    pub language: String,
+}
+
+/// Single query item representing a dimension selection.
+#[derive(Debug, Clone, Serialize)]
+pub struct QueryItem {
+    pub id: String,
+    pub values: Vec<String>,
+}
+
+/// Selection helper for building queries programmatically.
+#[derive(Debug, Clone)]
+pub struct Selection {
+    /// Dimension name (matches variable.id from metadata)
+    pub dimension: String,
+    /// Codes to include, or "*" for all
+    pub codes: Vec<String>,
+}
+
+/// Raw data response from PX-Web API POST request.
+/// JSON-stat2 format with metadata and data sections.
+#[derive(Debug, Clone, Deserialize, Serialize)]
+pub struct DataResponse {
+    pub dataset: Dataset,
+}
+
+/// Dataset wrapper containing dimensions and actual values.
+#[derive(Debug, Clone, Deserialize, Serialize)]
+pub struct Dataset {
+    pub dimension: DimInfo,
+    pub value: Vec<Option<f64>>,
+}
+
+/// Dimension metadata describing key layout.
+#[derive(Debug, Clone, Deserialize, Serialize)]
+pub struct DimInfo {
+    #[serde(rename = "key")]
+    pub keys: Vec<String>,
+    pub category: CategoryInfo,
+}
+
+/// Category info containing value lists for each dimension.
+#[derive(Debug, Clone, Deserialize, Serialize)]
+pub struct CategoryInfo {
+    pub label: HashMap<String, Vec<String>>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub index: Option<HashMap<String, Vec<usize>>>,
+}
+
+/// Decoded row of landing data with labeled dimensions.
+#[derive(Debug, Clone)]
+pub struct DataRow {
+    pub month: String,
+    pub species_code: String,
+    pub species_label: String,
+    pub gear_code: String,
+    pub gear_label: String,
+    pub zone_code: String,
+    pub zone_label: String,
+    pub processing_code: String,
+    pub processing_label: String,
+    pub preservation_code: String,
+    pub preservation_label: String,
+    pub shipsize_code: String,
+    pub shipsize_label: String,
+    pub measure_code: String,
+    pub measure_label: String,
+    pub value: Option<f64>,
+}
+
+/// Structured representation of a single landing record.
+#[derive(Debug, Clone)]
+pub struct Landing {
+    pub month: String,
+    pub species_code: String,
+    pub species_label: String,
+    pub gear_code: String,
+    pub zone_code: String,
+    pub processing_code: String,
+    pub preservation_code: String,
+    pub shipsize_code: String,
+    pub measure_code: String,
+    pub value: Option<f64>,
+}
+
+/// Error types for ingestion module.
+#[derive(Debug, thiserror::Error)]
+pub enum IngestError {
+    #[error("HTTP request failed: {0}")]
+    HttpError(#[from] reqwest::Error),
+
+    #[error("JSON parsing failed: {0}")]
+    JsonError(#[from] serde_json::Error),
+
+    #[error("Missing dimension in response: {0}")]
+    MissingDimension(String),
+
+    #[error("Invalid value code: {0}")]
+    InvalidValueCode(String),
+
+    #[error("API returned empty data set")]
+    EmptyDataset,
+
+    #[error("Unicode decode error: {0}")]
+    UnicodeError(String),
+}
+
+/// Lookup map for decoding key arrays into labeled values.
+/// Keyed by dimension name, contains code → label mappings.
+pub type LookupMap = HashMap<String, HashMap<String, String>>;
+
+/// Result alias using custom error type.
+pub type Result<T> = std::result::Result<T, IngestError>;