diff --git a/CHANGELOG.md b/CHANGELOG.md
index 95e18093..d337804e 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -6,6 +6,26 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 
+## [1.10.1]
+
+### Added
+
+- **Meta-ontology `1.1.0`** (`graflo/rdf/ontology/graflo.ttl`) — `gf:assigned` (datatype, boolean), `gf:hasHashIdentity` (object property reusing `gf:Identity`), `gf:SecondaryIdentity` (class) and `gf:hasSecondaryIdentity`. `gf:hasIdentity`'s domain widens to include `gf:SecondaryIdentity`, so a secondary identity's field list reuses the existing identity-chain vocabulary rather than adding a parallel one. Mirrored in `graflo-context.jsonld` and `rdf/namespace.py`; `docs/assets/graflo-ontology-viz/` regenerated.
+- **Resource sampling** — a connector-driven sampling primitive, the shared input stage for schema inference (algorithmic or agentic). Previously `infer_manifest` performed this privately for PostgreSQL and nothing else could reach it.
+- **`architecture.onto_sample`** (new, L2 leaf models) — `ResourceSample` holds documents **verbatim as JSON**: flat rows for tables, arbitrarily nested for APIs, nothing flattened at the boundary. `connector` records where the documents came from, so the relation that later becomes a resource plus its `resource_connector` binding survives the round trip. `SourceSample.samples_by_resource` returns `dict[str, list[dict]]` — the shape cross-resource identity inference consumes, so no adapter is needed. `ForeignKeyHint` carries *declared* references, which are ground truth for edge inference rather than a name-suffix guess.
+- **Sampling and profiling are separate operations.** `ResourceSampler` (`hq/sampler.py`) pulls documents; `profile_sample` derives the description. `ResourceProfile` is **path-keyed** (`address.city`, `items[].sku`) with a `FieldType`, null rate, cardinality and nesting depth, so hierarchical and tabular sources take one code path — a flat column list cannot represent a nested response at all. `ResourceProfile.flat_docs` projects nested documents into flat records, which is what makes `IdentityInferencer` (flat-records only) usable on an API source.
+- **`GraphEngine.sample_resources(...)`** — accepts a `PostgresConfig`, a `Bindings` block, or a file/directory path. Bindings-driven sampling reuses `RegistryBuilder.discover_files` and the `resource_connector` mapping, so provenance comes for free instead of being reconstructed. PostgreSQL sampling carries `primary_key` and `foreign_keys` through from introspection. Documents are normalized to JSON-safe values (`datetime`, `Decimal`, `memoryview`, numpy scalars), capped by `max_docs` and `max_cell_chars`, with `truncated` set when anything was dropped or clipped.
+- **`PostgresConfig.from_dsn(dsn)`** — a bare `uri=` derives only host and port, leaving `username`/`password`/`database` unset, so `to_sqlalchemy_connection_string()` raised. Parses credentials, database and a `search_path` from either `?schema=` or libpq `?options=-csearch_path=`.
+- Concept page: [Sampling and profiling](docs/concepts/schema/sampling_and_profiling.md), cross-linked from the identity inference guide.
+
+### Fixed
+
+- **RDF round-trip no longer silently degrades a vertex's identity mode** (`CORE-RDF-001`). The serializer wrote only `gf:blank` and `gf:identityName`, so `assigned`, `hash_identity_properties` and `secondary_identities` were dropped and every vertex read back as `natural` — a wrong-but-valid schema, which is the worst failure shape for a documented round-trip format. All four identity modes (`natural`, `hash`, `blank`, `assigned`) now survive, and `examples/16-secondary-identities` round-trips to canonical equality.
+- **Identity field order survives.** Identity nodes are `BNode`s and RDF triples are unordered, so a multi-field `identity` could come back permuted. They now carry `gf:artifactIndex` and are read through `_ordered_nodes`; graphs written by the previous serializer still parse (a missing index degrades to arbitrary order rather than failing).
+- **`strict_references=True` now rejects a pipeline `vertex:` step naming an undeclared vertex.** `filter_vertex_config_for_resource` intersects a resource's vertex names with the schema's and silently drops unknowns, so a resource that ingested nothing validated clean — a name mismatch between the vertex definition and the step was invisible. **Behaviour change:** manifests that previously passed under `strict_references=True` may now fail, which is the point; lenient validation is unchanged.
+- `ChunkerFactory._guess_chunker_type` raises the documented `ValueError` for a file with no extension instead of `IndexError`, so callers scanning a directory can skip it like any other unknown type.
+- The CSV/TSV chunker no longer raises `RuntimeError: generator raised StopIteration` (PEP 479) on an empty file; it yields no rows.
+
 ## [1.10.0]
 
 
diff --git a/docs/assets/graflo-ontology-viz/embed.html b/docs/assets/graflo-ontology-viz/embed.html
index da83e1b4..7c27f656 100644
--- a/docs/assets/graflo-ontology-viz/embed.html
+++ b/docs/assets/graflo-ontology-viz/embed.html
@@ -2,7 +2,7 @@
 <html lang="en">
 <head>
   <meta charset="utf-8" />
-  <title>GraFlo Ontology (v1.0.0)</title>
+  <title>GraFlo Ontology (v1.1.0)</title>
   <link rel="stylesheet" href="graph-view.css" />
 </head>
 <body class="embed">
@@ -213,6 +213,13 @@
       "source": "https://ontology.growgraph.dev/graflo/Schema",
       "target": "https://ontology.growgraph.dev/graflo/GrafloArtifact"
     },
+    {
+      "id": "sub:https://ontology.growgraph.dev/graflo/SecondaryIdentity->https://ontology.growgraph.dev/graflo/GrafloArtifact",
+      "kind": "subClassOf",
+      "label": "subClassOf",
+      "source": "https://ontology.growgraph.dev/graflo/SecondaryIdentity",
+      "target": "https://ontology.growgraph.dev/graflo/GrafloArtifact"
+    },
     {
       "id": "sub:https://ontology.growgraph.dev/graflo/SparqlConnector->https://ontology.growgraph.dev/graflo/BoundConnector",
       "kind": "subClassOf",
@@ -367,6 +374,20 @@
       "source": "https://ontology.growgraph.dev/graflo/Vertex",
       "target": "https://ontology.growgraph.dev/graflo/Identity"
     },
+    {
+      "id": "prop:https://ontology.growgraph.dev/graflo/hasHashIdentity",
+      "kind": "objectProperty",
+      "label": "hasHashIdentity",
+      "source": "https://ontology.growgraph.dev/graflo/Vertex",
+      "target": "https://ontology.growgraph.dev/graflo/Identity"
+    },
+    {
+      "id": "prop:https://ontology.growgraph.dev/graflo/hasSecondaryIdentity",
+      "kind": "objectProperty",
+      "label": "hasSecondaryIdentity",
+      "source": "https://ontology.growgraph.dev/graflo/Vertex",
+      "target": "https://ontology.growgraph.dev/graflo/SecondaryIdentity"
+    },
     {
       "id": "prop:https://ontology.growgraph.dev/graflo/edgeSource",
       "kind": "objectProperty",
@@ -791,6 +812,13 @@
       "label": "Schema",
       "local": "Schema"
     },
+    {
+      "comment": "Alternate named field-set that identifies a vertex for lookup only. Edge endpoints may match on it; upserts always use the primary identity.",
+      "id": "https://ontology.growgraph.dev/graflo/SecondaryIdentity",
+      "kind": "gf",
+      "label": "SecondaryIdentity",
+      "local": "SecondaryIdentity"
+    },
     {
       "comment": null,
       "id": "https://ontology.growgraph.dev/graflo/SparqlConnector",
@@ -877,7 +905,7 @@
     }
   ],
   "ontology": "https://ontology.growgraph.dev/graflo",
-  "version": "1.0.0"
+  "version": "1.1.0"
 };</script>
   <script src="graph-view.js"></script>
 </body>
diff --git a/docs/assets/graflo-ontology-viz/graph-data.json b/docs/assets/graflo-ontology-viz/graph-data.json
index e75addef..9c8ad710 100644
--- a/docs/assets/graflo-ontology-viz/graph-data.json
+++ b/docs/assets/graflo-ontology-viz/graph-data.json
@@ -189,6 +189,13 @@
             "source": "https://ontology.growgraph.dev/graflo/Schema",
             "target": "https://ontology.growgraph.dev/graflo/GrafloArtifact"
         },
+        {
+            "id": "sub:https://ontology.growgraph.dev/graflo/SecondaryIdentity->https://ontology.growgraph.dev/graflo/GrafloArtifact",
+            "kind": "subClassOf",
+            "label": "subClassOf",
+            "source": "https://ontology.growgraph.dev/graflo/SecondaryIdentity",
+            "target": "https://ontology.growgraph.dev/graflo/GrafloArtifact"
+        },
         {
             "id": "sub:https://ontology.growgraph.dev/graflo/SparqlConnector->https://ontology.growgraph.dev/graflo/BoundConnector",
             "kind": "subClassOf",
@@ -343,6 +350,20 @@
             "source": "https://ontology.growgraph.dev/graflo/Vertex",
             "target": "https://ontology.growgraph.dev/graflo/Identity"
         },
+        {
+            "id": "prop:https://ontology.growgraph.dev/graflo/hasHashIdentity",
+            "kind": "objectProperty",
+            "label": "hasHashIdentity",
+            "source": "https://ontology.growgraph.dev/graflo/Vertex",
+            "target": "https://ontology.growgraph.dev/graflo/Identity"
+        },
+        {
+            "id": "prop:https://ontology.growgraph.dev/graflo/hasSecondaryIdentity",
+            "kind": "objectProperty",
+            "label": "hasSecondaryIdentity",
+            "source": "https://ontology.growgraph.dev/graflo/Vertex",
+            "target": "https://ontology.growgraph.dev/graflo/SecondaryIdentity"
+        },
         {
             "id": "prop:https://ontology.growgraph.dev/graflo/edgeSource",
             "kind": "objectProperty",
@@ -767,6 +788,13 @@
             "label": "Schema",
             "local": "Schema"
         },
+        {
+            "comment": "Alternate named field-set that identifies a vertex for lookup only. Edge endpoints may match on it; upserts always use the primary identity.",
+            "id": "https://ontology.growgraph.dev/graflo/SecondaryIdentity",
+            "kind": "gf",
+            "label": "SecondaryIdentity",
+            "local": "SecondaryIdentity"
+        },
         {
             "comment": null,
             "id": "https://ontology.growgraph.dev/graflo/SparqlConnector",
@@ -853,5 +881,5 @@
         }
     ],
     "ontology": "https://ontology.growgraph.dev/graflo",
-    "version": "1.0.0"
+    "version": "1.1.0"
 }
diff --git a/docs/assets/graflo-ontology-viz/index.html b/docs/assets/graflo-ontology-viz/index.html
index 3b1cc55d..1b8a911b 100644
--- a/docs/assets/graflo-ontology-viz/index.html
+++ b/docs/assets/graflo-ontology-viz/index.html
@@ -2,13 +2,13 @@
 <html lang="en">
 <head>
   <meta charset="utf-8" />
-  <title>GraFlo Ontology (v1.0.0)</title>
+  <title>GraFlo Ontology (v1.1.0)</title>
   <link rel="stylesheet" href="graph-view.css" />
 </head>
 <body>
   <div class="layout">
     <aside class="sidebar">
-      <h1>GraFlo Ontology (v1.0.0)</h1>
+      <h1>GraFlo Ontology (v1.1.0)</h1>
       <p>Class graph from <code>graflo.ttl</code>: hierarchy laid out by <code>subClassOf</code> (force-assisted, stable), standalone classes in the top-left. All property edges are shown; thick arrows are <code>subClassOf</code>.</p>
       <div class="control">
         <label for="search">Search classes</label>
@@ -236,6 +236,13 @@ <h1>GraFlo Ontology (v1.0.0)</h1>
       "source": "https://ontology.growgraph.dev/graflo/Schema",
       "target": "https://ontology.growgraph.dev/graflo/GrafloArtifact"
     },
+    {
+      "id": "sub:https://ontology.growgraph.dev/graflo/SecondaryIdentity->https://ontology.growgraph.dev/graflo/GrafloArtifact",
+      "kind": "subClassOf",
+      "label": "subClassOf",
+      "source": "https://ontology.growgraph.dev/graflo/SecondaryIdentity",
+      "target": "https://ontology.growgraph.dev/graflo/GrafloArtifact"
+    },
     {
       "id": "sub:https://ontology.growgraph.dev/graflo/SparqlConnector->https://ontology.growgraph.dev/graflo/BoundConnector",
       "kind": "subClassOf",
@@ -390,6 +397,20 @@ <h1>GraFlo Ontology (v1.0.0)</h1>
       "source": "https://ontology.growgraph.dev/graflo/Vertex",
       "target": "https://ontology.growgraph.dev/graflo/Identity"
     },
+    {
+      "id": "prop:https://ontology.growgraph.dev/graflo/hasHashIdentity",
+      "kind": "objectProperty",
+      "label": "hasHashIdentity",
+      "source": "https://ontology.growgraph.dev/graflo/Vertex",
+      "target": "https://ontology.growgraph.dev/graflo/Identity"
+    },
+    {
+      "id": "prop:https://ontology.growgraph.dev/graflo/hasSecondaryIdentity",
+      "kind": "objectProperty",
+      "label": "hasSecondaryIdentity",
+      "source": "https://ontology.growgraph.dev/graflo/Vertex",
+      "target": "https://ontology.growgraph.dev/graflo/SecondaryIdentity"
+    },
     {
       "id": "prop:https://ontology.growgraph.dev/graflo/edgeSource",
       "kind": "objectProperty",
@@ -814,6 +835,13 @@ <h1>GraFlo Ontology (v1.0.0)</h1>
       "label": "Schema",
       "local": "Schema"
     },
+    {
+      "comment": "Alternate named field-set that identifies a vertex for lookup only. Edge endpoints may match on it; upserts always use the primary identity.",
+      "id": "https://ontology.growgraph.dev/graflo/SecondaryIdentity",
+      "kind": "gf",
+      "label": "SecondaryIdentity",
+      "local": "SecondaryIdentity"
+    },
     {
       "comment": null,
       "id": "https://ontology.growgraph.dev/graflo/SparqlConnector",
@@ -900,7 +928,7 @@ <h1>GraFlo Ontology (v1.0.0)</h1>
     }
   ],
   "ontology": "https://ontology.growgraph.dev/graflo",
-  "version": "1.0.0"
+  "version": "1.1.0"
 };</script>
   <script src="graph-view.js"></script>
 </body>
diff --git a/docs/concepts/index.md b/docs/concepts/index.md
index 5f665846..d1d18b1b 100644
--- a/docs/concepts/index.md
+++ b/docs/concepts/index.md
@@ -67,6 +67,10 @@ flowchart LR
 
 Entry points: **`GraphEngine.migrate_graph()`** (schema + data), **`infer_schema_from_graph()`** (schema only), **`export_graph()`** (in-memory `GraFloOutput`). See [Graph export and migration](operations/graph_export_migration.md) and the [Graph DB migration guide](../guides/graph_db_migration.md).
 
+- **SourceSample** — a bounded, verbatim-JSON sample of one or more resources, retaining the
+  connector each document came from. The shared input stage for schema inference, algorithmic or
+  agentic; produced by `GraphEngine.sample_resources()`. See
+  [Sampling and profiling](schema/sampling_and_profiling.md).
 - **Source Instance** — a concrete data artifact (a file, a table, a SPARQL endpoint, an API, a Kafka topic), wrapped by an `AbstractDataSource` with a `DataSourceType` (`FILE`, `SQL`, `SPARQL`, `API`, `KAFKA`, `IN_MEMORY`).
 - **Resource** — a reusable transformation pipeline (actor steps: descend, transform, vertex, edge) that maps raw records to graph elements. Data sources bind to Resources by name via the `DataSourceRegistry`.
 - **GraphManifest** — the canonical top-level contract that composes `schema`, `ingestion_model`, and `bindings`. High-level **contract evolution** (remove/merge vertex types and keep ingestion aligned) is described in [Manifest evolution](schema/manifest_evolution.md).
diff --git a/docs/concepts/schema/sampling_and_profiling.md b/docs/concepts/schema/sampling_and_profiling.md
new file mode 100644
index 00000000..feb89ba2
--- /dev/null
+++ b/docs/concepts/schema/sampling_and_profiling.md
@@ -0,0 +1,163 @@
+# Sampling and profiling
+
+Before a schema can be inferred, something has to look at the data. GraFlo splits that into two
+operations that are deliberately kept apart:
+
+| | What it does | Produced by | Model |
+|---|---|---|---|
+| **Sampling** | Pulls a bounded set of documents from a connector, **verbatim** | `ResourceSampler` | `SourceSample` / `ResourceSample` |
+| **Profiling** | Derives paths, types, null rates and cardinality from those documents | `profile_sample` | `ResourceProfile` / `FieldProfile` |
+
+The split is what lets one code path serve both a CSV table and a paginated JSON API. A sample is
+**pure JSON** — a list of flat rows for a table, an arbitrarily nested object for an API — and
+nothing is flattened at the boundary. The flat, typed view is a *derived projection*, computed on
+demand. Collapsing the two into a single "here are the columns and their types" model cannot
+represent a hierarchical response at all.
+
+`GraphEngine.infer_manifest()` performed this privately for PostgreSQL and nothing else could reach
+it. Sampling is the same input stage, exposed, so that **any** inferencer — the algorithmic
+identity inferencers, an LLM agent, a studio preview — consumes the same substrate.
+
+## The sample
+
+```python
+from graflo.hq.graph_engine import GraphEngine
+
+source = GraphEngine().sample_resources("data/", max_docs=100)
+```
+
+`sample_resources` dispatches on what it is given: a `PostgresConfig`, a `Bindings` block, or a
+file/directory path (or list of paths). The result:
+
+```text
+source_name: sample-source
+  resource='api_orders' connector='api_orders' docs=2 truncated=False
+  resource='customers'  connector='customers'  docs=3 truncated=False
+  resource='orders'     connector='orders'     docs=3 truncated=False
+```
+
+Three fields on `ResourceSample` carry more weight than the documents themselves:
+
+- **`connector`** — the connector the documents came from. This is the relation that later becomes
+  a [`resource_connector` binding](../index.md), so provenance survives the round trip
+  instead of being reconstructed downstream. `sample_bindings` treats the existing
+  `resource_connector` mapping as the authority, so what is sampled is exactly what will be
+  ingested.
+- **`primary_key`** / **`foreign_keys`** — what the source *declared*. A `ForeignKeyHint` is ground
+  truth for edge inference; a `*_id` name-suffix guess is not, and must never be recorded here.
+  PostgreSQL sampling fills both from introspection; file sampling leaves them empty.
+- **`truncated`** — set when documents were dropped or string values clipped. Sampling reads one
+  document past `max_docs` precisely so that a source holding exactly `max_docs` documents is
+  distinguishable from one that was cut short.
+
+`SourceSample.samples_by_resource` returns `dict[str, list[dict]]` — the input shape cross-resource
+identity inference consumes, so no adapter sits between sampling and inference.
+
+### Guards
+
+Sampled documents leave the trust boundary: they land in prompts, previews and logs. Three caps
+apply, all on `ResourceSampler`:
+
+- `max_docs` (default 100) — documents per resource
+- `max_cell_chars` — length of any single string value
+- JSON normalization — `datetime`, `Decimal`, `memoryview`, `UUID` and numpy scalars are coerced to
+  JSON-safe values, because `dict[str, Any]` accepts them but only best-effort serializes them
+
+Files that yield no documents (an empty CSV, a `NOTES.txt`) are skipped with a warning rather than
+producing an empty resource.
+
+## The profile
+
+```python
+from graflo.architecture.onto_sample import profile_sample
+
+profile = profile_sample(source.get("api_orders"))
+```
+
+Profiles are **path-keyed**. Nested objects extend the path with `.`; lists of objects extend it
+with `[]`:
+
+```text
+max_depth: 1   nested: True
+
+order_id        STRING  depth=0  null_ratio=0.00
+customer.id     STRING  depth=1  null_ratio=0.00
+customer.city   STRING  depth=1  null_ratio=0.50
+items[].sku     STRING  depth=1  null_ratio=0.00
+items[].qty     INT     depth=1  null_ratio=0.00
+tags            LIST    depth=0  null_ratio=0.00
+```
+
+A list of *scalars* (`tags`) is typed whole as `LIST` with an `item_type`; a list of *objects*
+(`items`) is descended into. `max_depth > 0` is the signal that ingestion needs
+[`descend` steps](../ingestion/transforms.md) — a flat table is simply the `depth=0` case of the
+same code path.
+
+Type inference checks `bool` before `int` deliberately: `bool` is an `int` subclass in Python, so
+the naive order mistypes every boolean column as `INT`.
+
+!!! note "Types come from values, not from a declaration"
+    A CSV reader yields strings, so `orders.csv` profiles `total` and `paid` as `STRING`. A
+    PostgreSQL source carries real column types through introspection. Profiling describes what was
+    *observed*; it does not invent a declaration the source never made.
+
+### `flat_docs` — the bridge to identity inference
+
+`IdentityInferencer` operates on flat records. `ResourceProfile.flat_docs` projects nested documents
+onto the profile's paths, which is how an API source becomes eligible for it at all:
+
+```python
+profile.flat_docs(sample.docs)[0]
+# {'order_id': 'o1', 'customer.id': 'c1', 'customer.city': 'Berlin',
+#  'items[].sku': 'A-1', 'items[].qty': 2, 'tags': ['priority', 'gift']}
+```
+
+!!! warning "`unique` is a property of the sample, not of the source"
+    `FieldProfile.unique` means every non-null value observed was distinct — over as few as two
+    documents. Treat it as a candidate signal to be confirmed against a larger sample or a declared
+    `primary_key`, never as a uniqueness constraint. `min_sample_size` in identity inference exists
+    for this reason.
+
+## Where it fits
+
+```mermaid
+flowchart LR
+    C["Connectors<br/>File · Table · SPARQL · API"]
+    S["ResourceSampler<br/>bounded, verbatim JSON"]
+    SS["SourceSample<br/>docs + connector + declared keys"]
+    P["profile_sample<br/>paths · types · null rate"]
+    II["IdentityInferencer"]
+    AG["Agentic inference<br/>(ScheWea)"]
+    M["GraphManifest"]
+
+    C --> S --> SS
+    SS --> P --> II --> M
+    SS -- samples_by_resource --> II
+    SS --> AG --> M
+```
+
+Sampling deliberately stops short of proposing anything. What consumes it:
+
+- **[Identity inference](../../guides/identity_inference.md)** — vertex `identity` and
+  `hash_identity_properties` from flat samples.
+- **Cross-resource vertex discovery** — aligns fields across resources to find a shared key;
+  consumes `samples_by_resource` directly.
+- **Agentic inference** — an external service receives a serialized `SourceSample` over the wire.
+  Because the model is defined once here, the producer and the consumer cannot drift into
+  disagreement.
+
+Note the asymmetry a `SourceSample` deliberately preserves: it names connectors but does not carry
+their definitions, which hold paths, DSNs and credentials. A consumer can therefore propose
+resources but cannot, on its own, emit a `bindings` block — the caller that did the sampling holds
+the connectors and assembles it. This is the secret-free manifest doctrine falling out of the type
+system rather than being enforced by convention.
+
+## API
+
+| Symbol | Module |
+|---|---|
+| `SourceSample`, `ResourceSample`, `ForeignKeyHint` | `graflo.architecture.onto_sample` |
+| `ResourceProfile`, `FieldProfile` | `graflo.architecture.onto_sample` |
+| `profile_sample`, `profile_source`, `iter_paths`, `infer_field_type` | `graflo.architecture.onto_sample` |
+| `ResourceSampler` (`sample_file`, `sample_files`, `sample_postgres`, `sample_connector`, `sample_bindings`) | `graflo.hq.sampler` |
+| `GraphEngine.sample_resources` | `graflo.hq.graph_engine` |
diff --git a/docs/guides/identity_inference.md b/docs/guides/identity_inference.md
index 61366821..ac103371 100644
--- a/docs/guides/identity_inference.md
+++ b/docs/guides/identity_inference.md
@@ -43,6 +43,28 @@ uv run python ingest.py
 
 Writes a chunked GraFlo file backend under `artifacts/csv-backend/`.
 
+## Samples from a non-CSV source
+
+`IdentityInferencer.infer()` takes **flat records**, which is why this guide starts from CSV. To
+reach a PostgreSQL table, a directory of mixed files, or a nested API response, sample the source
+first and flatten through its profile:
+
+```python
+from graflo.architecture.onto_sample import profile_sample
+from graflo.hq.graph_engine import GraphEngine
+
+source = GraphEngine().sample_resources(pg_config, schema_name="public", max_docs=500)
+sample = source.get("customers")
+profile = profile_sample(sample)
+
+records = profile.flat_docs(sample.docs)   # nested paths become 'customer.id', 'items[].sku'
+```
+
+A sampled source also carries the **declared** `primary_key` and `foreign_keys` when it has them —
+prefer those over an inferred identity, and see
+[Sampling and profiling](../concepts/schema/sampling_and_profiling.md) for the caps and the caveat
+on `unique`.
+
 ## Identity modes
 
 After inference, each vertex has a derived **`identity_mode`**:
@@ -63,6 +85,7 @@ For attaching edges by a business key that is not the upsert identity, see
 ## Related documentation
 
 - [Vertex identity modes](../concepts/schema/vertex_identity.md)
+- [Sampling and profiling](../concepts/schema/sampling_and_profiling.md) — where samples come from
 - [Example 16 — Secondary identities](../examples/example-16.md)
 - [Core components — Vertex](../concepts/architecture/core_components.md)
 - [Graph export and replay](graph_export_and_replay.md) — file backend ingest pattern
diff --git a/docs/guides/index.md b/docs/guides/index.md
index be128fac..df3dce5e 100644
--- a/docs/guides/index.md
+++ b/docs/guides/index.md
@@ -10,7 +10,8 @@ Task-oriented walkthroughs for common GraFlo workflows. Each guide links to the
 | Export a graph to disk and replay it | [Graph export and replay](graph_export_and_replay.md) | [Example 13](../examples/example-13.md) | [Graph export and migration](../concepts/operations/graph_export_migration.md) |
 | Wire REST API credentials from environment variables | [API env wiring](api_env_wiring.md) | [Example 14](../examples/example-14.md) | [API connector](../concepts/connectors/api_connector.md) |
 | Ingest JSON messages from Kafka topics | — | — | [Kafka connector](../concepts/connectors/kafka_connector.md) |
-| Infer vertex identities from CSV samples | [Identity inference](identity_inference.md) | [Example 15](../examples/example-15.md) | [Vertex identity](../concepts/schema/vertex_identity.md) |
+| Look at a source before writing a manifest | — | — | [Sampling and profiling](../concepts/schema/sampling_and_profiling.md) |
+| Infer vertex identities from CSV samples | [Identity inference](identity_inference.md) | [Example 15](../examples/example-15.md) | [Vertex identity](../concepts/schema/vertex_identity.md), [Sampling and profiling](../concepts/schema/sampling_and_profiling.md) |
 | Attach edges by a business key (ISIN, LEI, …) without owning the vertex | — | [Example 16](../examples/example-16.md) | [Secondary identities](../concepts/schema/vertex_identity.md#secondary-identities-edge-endpoint-lookup) |
 | Bulk-load TigerGraph via CSV and S3 staging | [TigerGraph bulk load](tigergraph_bulk_load.md) | [Example 10](../examples/example-10.md) | [Object storage](../concepts/operations/object_storage.md) |
 | Use a pre-provisioned graph namespace (least privilege) | [Graph namespace and schema](graph_namespace_and_schema.md) | — | [Capabilities](../concepts/architecture/capabilities.md) |
diff --git a/docs/reference/architecture/contract/runtime.md b/docs/reference/architecture/contract/runtime.md
index 9ccd5e93..e5014ccd 100644
--- a/docs/reference/architecture/contract/runtime.md
+++ b/docs/reference/architecture/contract/runtime.md
@@ -1,5 +1,8 @@
-# `graflo.architecture.contract.runtime`
+# Runtime (moved)
 
-::: graflo.architecture.contract.runtime
-    options:
-      show_submodules: false
+`graflo.architecture.contract.runtime` was removed in the contract ↔ pipeline
+inversion. Schema-bound execution now lives under
+[`graflo.architecture.pipeline.runtime`](../pipeline/runtime.md).
+
+- [`ResourceRuntime`](runtime/resource.md) → `graflo.architecture.pipeline.runtime.resource`
+- [`EdgeDerivation`](runtime/edge_derivation.md) → `graflo.architecture.graph_types.edge_derivation`
diff --git a/docs/reference/architecture/contract/runtime/edge_derivation.md b/docs/reference/architecture/contract/runtime/edge_derivation.md
index 12ea09ee..bcbf7d94 100644
--- a/docs/reference/architecture/contract/runtime/edge_derivation.md
+++ b/docs/reference/architecture/contract/runtime/edge_derivation.md
@@ -1,3 +1,4 @@
-# `graflo.architecture.graph_types.edge_derivation`
+# `graflo.architecture.contract.runtime.edge_derivation` (moved)
 
-::: graflo.architecture.graph_types.edge_derivation
+`graflo.architecture.contract.runtime.edge_derivation` moved to [`graflo.architecture.graph_types.edge_derivation`](../../../architecture/graph_types/edge_derivation.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../../../../guides/importing.md).
diff --git a/docs/reference/architecture/contract/runtime/resource.md b/docs/reference/architecture/contract/runtime/resource.md
index 52252238..fb8cf0b4 100644
--- a/docs/reference/architecture/contract/runtime/resource.md
+++ b/docs/reference/architecture/contract/runtime/resource.md
@@ -1,3 +1,4 @@
-# `graflo.architecture.pipeline.runtime.resource`
+# `graflo.architecture.contract.runtime.resource` (moved)
 
-::: graflo.architecture.pipeline.runtime.resource
+`graflo.architecture.contract.runtime.resource` moved to [`graflo.architecture.pipeline.runtime.resource`](../../../architecture/pipeline/runtime/resource.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../../../../guides/importing.md).
diff --git a/docs/reference/architecture/database_features.md b/docs/reference/architecture/database_features.md
index 38f48703..978a469b 100644
--- a/docs/reference/architecture/database_features.md
+++ b/docs/reference/architecture/database_features.md
@@ -1,3 +1,4 @@
-# `graflo.architecture.schema.database_features`
+# `graflo.architecture.database_features` (moved)
 
-::: graflo.architecture.schema.database_features
+`graflo.architecture.database_features` moved to [`graflo.architecture.schema.database_features`](../architecture/schema/database_features.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../../guides/importing.md).
diff --git a/docs/reference/architecture/edge_derivation.md b/docs/reference/architecture/edge_derivation.md
index 12ea09ee..5236aec1 100644
--- a/docs/reference/architecture/edge_derivation.md
+++ b/docs/reference/architecture/edge_derivation.md
@@ -1,3 +1,4 @@
-# `graflo.architecture.graph_types.edge_derivation`
+# `graflo.architecture.edge_derivation` (moved)
 
-::: graflo.architecture.graph_types.edge_derivation
+`graflo.architecture.edge_derivation` moved to [`graflo.architecture.graph_types.edge_derivation`](../architecture/graph_types/edge_derivation.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../../guides/importing.md).
diff --git a/docs/reference/architecture/pipeline/runtime/actor/config.md b/docs/reference/architecture/pipeline/runtime/actor/config.md
index f4370391..1c84007a 100644
--- a/docs/reference/architecture/pipeline/runtime/actor/config.md
+++ b/docs/reference/architecture/pipeline/runtime/actor/config.md
@@ -1,5 +1,4 @@
-# `graflo.architecture.contract.ingestion.steps`
+# `graflo.architecture.pipeline.runtime.actor.config` (moved)
 
-::: graflo.architecture.contract.ingestion.steps
-    options:
-      show_submodules: false
+`graflo.architecture.pipeline.runtime.actor.config` moved to [`graflo.architecture.contract.ingestion.steps`](../../../../architecture/contract/ingestion/steps.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../../../../../guides/importing.md).
diff --git a/docs/reference/architecture/pipeline/runtime/actor/config/models.md b/docs/reference/architecture/pipeline/runtime/actor/config/models.md
index 4bb2afc2..d0ec861d 100644
--- a/docs/reference/architecture/pipeline/runtime/actor/config/models.md
+++ b/docs/reference/architecture/pipeline/runtime/actor/config/models.md
@@ -1,3 +1,4 @@
-# `graflo.architecture.contract.ingestion.steps.models`
+# `graflo.architecture.pipeline.runtime.actor.config.models` (moved)
 
-::: graflo.architecture.contract.ingestion.steps.models
+`graflo.architecture.pipeline.runtime.actor.config.models` moved to [`graflo.architecture.contract.ingestion.steps.models`](../../../../../architecture/contract/ingestion/steps/models.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../../../../../../guides/importing.md).
diff --git a/docs/reference/architecture/pipeline/runtime/actor/config/normalize.md b/docs/reference/architecture/pipeline/runtime/actor/config/normalize.md
index af6d7494..26d48151 100644
--- a/docs/reference/architecture/pipeline/runtime/actor/config/normalize.md
+++ b/docs/reference/architecture/pipeline/runtime/actor/config/normalize.md
@@ -1,3 +1,4 @@
-# `graflo.architecture.contract.ingestion.steps.normalize`
+# `graflo.architecture.pipeline.runtime.actor.config.normalize` (moved)
 
-::: graflo.architecture.contract.ingestion.steps.normalize
+`graflo.architecture.pipeline.runtime.actor.config.normalize` moved to [`graflo.architecture.contract.ingestion.steps.normalize`](../../../../../architecture/contract/ingestion/steps/normalize.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../../../../../../guides/importing.md).
diff --git a/docs/reference/architecture/pipeline/runtime/actor/config/parse.md b/docs/reference/architecture/pipeline/runtime/actor/config/parse.md
index b32d63af..adbcbada 100644
--- a/docs/reference/architecture/pipeline/runtime/actor/config/parse.md
+++ b/docs/reference/architecture/pipeline/runtime/actor/config/parse.md
@@ -1,3 +1,4 @@
-# `graflo.architecture.contract.ingestion.steps.parse`
+# `graflo.architecture.pipeline.runtime.actor.config.parse` (moved)
 
-::: graflo.architecture.contract.ingestion.steps.parse
+`graflo.architecture.pipeline.runtime.actor.config.parse` moved to [`graflo.architecture.contract.ingestion.steps.parse`](../../../../../architecture/contract/ingestion/steps/parse.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../../../../../../guides/importing.md).
diff --git a/docs/reference/connection_models.md b/docs/reference/connection_models.md
index 008a5413..5fd11021 100644
--- a/docs/reference/connection_models.md
+++ b/docs/reference/connection_models.md
@@ -1,3 +1,4 @@
-# `graflo.connections.sources`
+# `graflo.connection_models` (moved)
 
-::: graflo.connections.sources
+`graflo.connection_models` moved to [`graflo.connections.sources`](connections/sources.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../guides/importing.md).
diff --git a/docs/reference/db/connection.md b/docs/reference/db/connection.md
index c1880bcb..23d9f77b 100644
--- a/docs/reference/db/connection.md
+++ b/docs/reference/db/connection.md
@@ -1,5 +1,4 @@
-# `graflo.connections.onto`
+# `graflo.db.connection` (moved)
 
-::: graflo.connections.onto
-    options:
-      show_submodules: false
+`graflo.db.connection` moved to [`graflo.connections.onto`](../connections/onto.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../../guides/importing.md).
diff --git a/docs/reference/db/connection/config_mapping.md b/docs/reference/db/connection/config_mapping.md
index d94a75e1..a9c2edb5 100644
--- a/docs/reference/db/connection/config_mapping.md
+++ b/docs/reference/db/connection/config_mapping.md
@@ -1,3 +1,4 @@
-# `graflo.connections.mapping`
+# `graflo.db.connection.config_mapping` (moved)
 
-::: graflo.connections.mapping
+`graflo.db.connection.config_mapping` moved to [`graflo.connections.mapping`](../../connections/mapping.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../../../guides/importing.md).
diff --git a/docs/reference/db/connection/onto.md b/docs/reference/db/connection/onto.md
index 9127bfe7..a901b5ab 100644
--- a/docs/reference/db/connection/onto.md
+++ b/docs/reference/db/connection/onto.md
@@ -1,3 +1,4 @@
-# `graflo.connections.onto`
+# `graflo.db.connection.onto` (moved)
 
-::: graflo.connections.onto
+`graflo.db.connection.onto` moved to [`graflo.connections.onto`](../../connections/onto.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../../../guides/importing.md).
diff --git a/docs/reference/db/connection/wsgi.md b/docs/reference/db/connection/wsgi.md
index bd146fd3..51dd0579 100644
--- a/docs/reference/db/connection/wsgi.md
+++ b/docs/reference/db/connection/wsgi.md
@@ -1,3 +1,4 @@
-# `graflo.connections.onto.wsgi`
+# WSGI connection config (removed)
 
-::: graflo.connections.onto.wsgi
+`WSGIConfig` / `graflo.connections.onto.wsgi` was deleted. Use the connection
+configs under [`graflo.connections.onto`](onto.md).
diff --git a/docs/reference/db/graflo_backend/config.md b/docs/reference/db/graflo_backend/config.md
index 9e4daf30..cdc4bb23 100644
--- a/docs/reference/db/graflo_backend/config.md
+++ b/docs/reference/db/graflo_backend/config.md
@@ -1,3 +1,4 @@
-# `graflo.db.graflo_backend.config`
+# `graflo.db.graflo_backend.config` (moved)
 
-::: graflo.db.graflo_backend.config
+Backend connection config now lives under
+[`graflo.connections.graflo_backend`](../../connections/graflo_backend.md).
diff --git a/docs/reference/hq/connection_provider.md b/docs/reference/hq/connection_provider.md
index de3c4e1f..2eafe0e3 100644
--- a/docs/reference/hq/connection_provider.md
+++ b/docs/reference/hq/connection_provider.md
@@ -1,3 +1,4 @@
-# `graflo.connections.provider`
+# `graflo.hq.connection_provider` (moved)
 
-::: graflo.connections.provider
+`graflo.hq.connection_provider` moved to [`graflo.connections.provider`](../connections/provider.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../../guides/importing.md).
diff --git a/docs/reference/hq/fuzzy_matcher.md b/docs/reference/hq/fuzzy_matcher.md
index 5be21787..d4886ecd 100644
--- a/docs/reference/hq/fuzzy_matcher.md
+++ b/docs/reference/hq/fuzzy_matcher.md
@@ -1,3 +1,4 @@
-# `graflo.util.fuzzy_matcher`
+# `graflo.hq.fuzzy_matcher` (moved)
 
-::: graflo.util.fuzzy_matcher
+`graflo.hq.fuzzy_matcher` moved to [`graflo.util.fuzzy_matcher`](../util/fuzzy_matcher.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../../guides/importing.md).
diff --git a/docs/reference/util/chunker.md b/docs/reference/util/chunker.md
index 9add37c3..aef5caf0 100644
--- a/docs/reference/util/chunker.md
+++ b/docs/reference/util/chunker.md
@@ -1,3 +1,4 @@
-# `graflo.data_source.chunker`
+# `graflo.util.chunker` (moved)
 
-::: graflo.data_source.chunker
+`graflo.util.chunker` moved to [`graflo.data_source.chunker`](../data_source/chunker.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../../guides/importing.md).
diff --git a/docs/reference/util/merge.md b/docs/reference/util/merge.md
index 256bfca1..f4d28928 100644
--- a/docs/reference/util/merge.md
+++ b/docs/reference/util/merge.md
@@ -1,3 +1,4 @@
-# `graflo.architecture.graph_types.merge`
+# `graflo.util.merge` (moved)
 
-::: graflo.architecture.graph_types.merge
+`graflo.util.merge` moved to [`graflo.architecture.graph_types.merge`](../architecture/graph_types/merge.md) in 1.10.0.
+Full old→new path table: [Importing and layering](../../guides/importing.md).
diff --git a/docs/reference/util/onto.md b/docs/reference/util/onto.md
index a28895be..e2c84a65 100644
--- a/docs/reference/util/onto.md
+++ b/docs/reference/util/onto.md
@@ -1,3 +1,4 @@
-# `graflo.util.onto`
+# `graflo.util.onto` (moved)
 
-::: graflo.util.onto
+`graflo.util.onto` was removed. Shared ontology constants now live under
+[`graflo.onto`](../onto.md).
diff --git a/graflo/architecture/graph_types/container.py b/graflo/architecture/graph_types/container.py
index d316cdb1..9fa94664 100644
--- a/graflo/architecture/graph_types/container.py
+++ b/graflo/architecture/graph_types/container.py
@@ -3,8 +3,6 @@
 from __future__ import annotations
 
 from collections import defaultdict
-from datetime import date, datetime, time
-from decimal import Decimal
 from typing import Any
 
 from pydantic import Field, field_serializer, field_validator
@@ -18,30 +16,7 @@
     serialize_edge_key,
     serialize_entity_key,
 )
-
-
-def _pick_unique_dict(docs: list) -> list:
-    """Deduplicate dicts by structure; preserves original objects (local copy of merge logic)."""
-
-    def make_hashable(obj: object) -> Any:
-        if isinstance(obj, dict):
-            return tuple(sorted((k, make_hashable(v)) for k, v in obj.items()))
-        if isinstance(obj, (list, tuple)):
-            return tuple(make_hashable(item) for item in obj)
-        if isinstance(obj, (datetime, date, time)):
-            return ("__datetime__", obj.isoformat())
-        if isinstance(obj, Decimal):
-            return ("__decimal__", str(obj))
-        if isinstance(obj, set):
-            return tuple(sorted(make_hashable(item) for item in obj))
-        return obj
-
-    seen: dict[Any, object] = {}
-    for doc in docs:
-        key = make_hashable(doc)
-        if key not in seen:
-            seen[key] = doc
-    return list(seen.values())
+from graflo.util.transform import pick_unique_dict
 
 
 def _serialize_linear_item(
@@ -133,9 +108,9 @@ def items(self):
     def pick_unique(self):
         """Remove duplicate entries from vertices and edges."""
         for k, v in self.vertices.items():
-            self.vertices[k] = _pick_unique_dict(v)
+            self.vertices[k] = pick_unique_dict(v)
         for k, v in self.edges.items():
-            self.edges[k] = _pick_unique_dict(v)
+            self.edges[k] = pick_unique_dict(v)
 
     @classmethod
     def from_docs_list(
diff --git a/graflo/architecture/onto_sample.py b/graflo/architecture/onto_sample.py
new file mode 100644
index 00000000..507eddf7
--- /dev/null
+++ b/graflo/architecture/onto_sample.py
@@ -0,0 +1,331 @@
+"""Resource sampling contract — pure-JSON samples and their derived profiles.
+
+Samples are the raw material every schema inferencer consumes, whether it reasons
+algorithmically (:mod:`graflo.db.identity_inference`) or with a language model
+(ScheWea). Two ideas, deliberately kept apart:
+
+* **Sampling** pulls documents from a connector. :class:`ResourceSample` holds
+  them **verbatim as JSON** — tabular sources yield flat ``list[dict]`` rows,
+  API sources yield arbitrarily nested documents. Nothing is flattened at this
+  boundary, so a hierarchical response survives intact.
+* **Profiling** describes those documents. :func:`profile_sample` derives the
+  flat, path-keyed, typed view (:class:`ResourceProfile`) used for prompting,
+  studio previews and identity inference.
+
+:attr:`ResourceSample.connector` records *where* the documents came from. That
+relation is what later becomes a resource plus its ``resource_connector``
+binding, so it must survive the round trip — a sample that has lost its
+provenance cannot be turned back into an ingestion model.
+"""
+
+from __future__ import annotations
+
+import re
+from collections.abc import Iterator
+from datetime import date, datetime, time
+from typing import Any
+
+from pydantic import Field as PydanticField
+
+from graflo.architecture.base import ConfigBaseModel
+from graflo.architecture.schema.identity_uuid import UUID_PATTERN
+from graflo.architecture.schema.vertex import FieldType
+
+#: Path segment appended when descending into a list of objects, e.g. ``items[].sku``.
+LIST_MARKER = "[]"
+
+_ISO_DATETIME_PATTERN = re.compile(
+    r"^\d{4}-\d{2}-\d{2}(?:[T ]\d{2}:\d{2}:\d{2}(?:\.\d+)?(?:Z|[+-]\d{2}:?\d{2})?)?$"
+)
+_LONG_TEXT_THRESHOLD = 256
+
+DEFAULT_MAX_DOCS = 100
+DEFAULT_MAX_PATHS = 200
+DEFAULT_MAX_EXAMPLES = 3
+
+
+class ForeignKeyHint(ConfigBaseModel):
+    """A declared reference from one resource to another.
+
+    Populated only when the source *declares* it (a SQL foreign key, an RDF
+    range). This is ground truth for edge inference and must not be confused
+    with the name-suffix guessing an inferencer falls back to.
+    """
+
+    field: str
+    references_resource: str
+    references_field: str | None = None
+
+
+class ResourceSample(ConfigBaseModel):
+    """Documents sampled from one resource, plus what the source declared about it."""
+
+    resource_name: str
+    """Logical resource name; becomes ``ResourceConfig.name``."""
+
+    connector: str | None = None
+    """Name of the connector the documents came from; becomes the
+    ``resource_connector`` binding. ``None`` when sampled without bindings."""
+
+    docs: list[dict[str, Any]] = PydanticField(default_factory=list)
+    """Sampled documents, verbatim JSON. Flat rows for tables, nested for APIs."""
+
+    description: str | None = None
+    primary_key: list[str] = PydanticField(default_factory=list)
+    """Declared primary key, when the source has one."""
+
+    foreign_keys: list[ForeignKeyHint] = PydanticField(default_factory=list)
+    """Declared outbound references, when the source has them."""
+
+    truncated: bool = False
+    """True when documents were dropped or values clipped to respect caps."""
+
+    total_estimate: int | None = None
+    """Approximate total document count at the source, when cheaply available."""
+
+
+class SourceSample(ConfigBaseModel):
+    """A set of resource samples drawn from one logical source."""
+
+    source_name: str
+    description: str | None = None
+    samples: list[ResourceSample] = PydanticField(min_length=1)
+
+    @property
+    def samples_by_resource(self) -> dict[str, list[dict[str, Any]]]:
+        """Documents keyed by resource name.
+
+        This is the input shape consumed by cross-resource identity inference,
+        so no adapter is needed between sampling and inference.
+        """
+        return {sample.resource_name: sample.docs for sample in self.samples}
+
+    def get(self, resource_name: str) -> ResourceSample | None:
+        """Return the sample for *resource_name*, or ``None``."""
+        for sample in self.samples:
+            if sample.resource_name == resource_name:
+                return sample
+        return None
+
+
+class FieldProfile(ConfigBaseModel):
+    """Derived description of one field path within a resource sample."""
+
+    path: str
+    """Dotted path to the value, e.g. ``address.city`` or ``items[].sku``."""
+
+    type: FieldType = FieldType.STRING
+    item_type: FieldType | None = None
+    """Element type; set only when ``type`` is ``LIST``."""
+
+    depth: int = 0
+    """Nesting depth of the path. ``0`` for a top-level scalar."""
+
+    present: int = 0
+    """Documents in which the path occurred."""
+
+    null_count: int = 0
+    distinct: int = 0
+    examples: list[str] = PydanticField(default_factory=list)
+
+    @property
+    def null_ratio(self) -> float:
+        """Fraction of occurrences whose value was null."""
+        return self.null_count / self.present if self.present else 0.0
+
+    @property
+    def unique(self) -> bool:
+        """True when every non-null occurrence was distinct."""
+        non_null = self.present - self.null_count
+        return non_null > 0 and self.distinct == non_null
+
+
+class ResourceProfile(ConfigBaseModel):
+    """Derived, flat, typed view of a :class:`ResourceSample`."""
+
+    resource_name: str
+    connector: str | None = None
+    doc_count: int = 0
+    max_depth: int = 0
+    """Deepest nesting observed. ``> 0`` means ingestion needs ``descend`` steps."""
+
+    fields: list[FieldProfile] = PydanticField(default_factory=list)
+    primary_key: list[str] = PydanticField(default_factory=list)
+    foreign_keys: list[ForeignKeyHint] = PydanticField(default_factory=list)
+    truncated: bool = False
+
+    @property
+    def field_paths(self) -> list[str]:
+        return [field.path for field in self.fields]
+
+    @property
+    def nested(self) -> bool:
+        return self.max_depth > 0
+
+    def flat_docs(self, docs: list[dict[str, Any]]) -> list[dict[str, Any]]:
+        """Project *docs* onto this profile's paths as flat records.
+
+        Identity inference operates on flat records; this is how a nested source
+        becomes eligible for it.
+        """
+        paths = set(self.field_paths)
+        flattened: list[dict[str, Any]] = []
+        for doc in docs:
+            record: dict[str, Any] = {}
+            for path, _depth, value in iter_paths(doc):
+                if path in paths and path not in record:
+                    record[path] = value
+            flattened.append(record)
+        return flattened
+
+
+def iter_paths(
+    doc: dict[str, Any], prefix: str = "", depth: int = 0
+) -> Iterator[tuple[str, int, Any]]:
+    """Yield ``(path, depth, value)`` for every leaf in a JSON document.
+
+    Nested objects extend the path with ``.``; lists of objects extend it with
+    ``[]``. Lists of scalars are yielded whole so they can be typed as ``LIST``.
+    """
+    for key, value in doc.items():
+        path = f"{prefix}.{key}" if prefix else key
+        if isinstance(value, dict):
+            yield from iter_paths(value, path, depth + 1)
+        elif isinstance(value, list) and any(isinstance(item, dict) for item in value):
+            nested_path = f"{path}{LIST_MARKER}"
+            for item in value:
+                if isinstance(item, dict):
+                    yield from iter_paths(item, nested_path, depth + 1)
+        else:
+            yield path, depth, value
+
+
+def infer_field_type(values: list[Any]) -> tuple[FieldType, FieldType | None]:
+    """Infer a ``FieldType`` (and ``item_type`` for lists) from observed values.
+
+    Checks ``bool`` before ``int`` deliberately — ``bool`` is an ``int`` subclass
+    in Python, so the naive order mistypes every boolean column as ``INT``.
+    """
+    non_null = [value for value in values if value is not None]
+    if not non_null:
+        return FieldType.STRING, None
+
+    if all(isinstance(value, list) for value in non_null):
+        items = [item for value in non_null for item in value]
+        item_type, _ = infer_field_type(items) if items else (FieldType.STRING, None)
+        return FieldType.LIST, item_type
+
+    if all(isinstance(value, bool) for value in non_null):
+        return FieldType.BOOL, None
+    if all(
+        isinstance(value, int) and not isinstance(value, bool) for value in non_null
+    ):
+        return FieldType.INT, None
+    if all(
+        isinstance(value, (int, float)) and not isinstance(value, bool)
+        for value in non_null
+    ):
+        return FieldType.FLOAT, None
+    if all(isinstance(value, (datetime, date, time)) for value in non_null):
+        return FieldType.DATETIME, None
+
+    if all(isinstance(value, str) for value in non_null):
+        if all(UUID_PATTERN.match(value) for value in non_null):
+            return FieldType.UUID, None
+        if all(_ISO_DATETIME_PATTERN.match(value) for value in non_null):
+            return FieldType.DATETIME, None
+
+    return FieldType.STRING, None
+
+
+def _example(value: Any) -> str:
+    text = str(value)
+    if len(text) > _LONG_TEXT_THRESHOLD:
+        return text[:_LONG_TEXT_THRESHOLD] + "…"
+    return text
+
+
+def profile_sample(
+    sample: ResourceSample,
+    *,
+    max_paths: int = DEFAULT_MAX_PATHS,
+    max_examples: int = DEFAULT_MAX_EXAMPLES,
+) -> ResourceProfile:
+    """Derive a :class:`ResourceProfile` from a sample's documents.
+
+    Handles tabular and hierarchical documents through one code path: a flat row
+    is simply the depth-0 case.
+    """
+    observed: dict[str, list[Any]] = {}
+    depths: dict[str, int] = {}
+
+    for doc in sample.docs:
+        for path, depth, value in iter_paths(doc):
+            if path not in observed and len(observed) >= max_paths:
+                continue
+            observed.setdefault(path, []).append(value)
+            depths[path] = max(depths.get(path, depth), depth)
+
+    truncated = sample.truncated or len(observed) >= max_paths
+
+    fields: list[FieldProfile] = []
+    for path, values in observed.items():
+        field_type, item_type = infer_field_type(values)
+        non_null = [value for value in values if value is not None]
+        hashable = {
+            value if isinstance(value, (str, int, float, bool)) else repr(value)
+            for value in non_null
+        }
+        fields.append(
+            FieldProfile(
+                path=path,
+                type=field_type,
+                item_type=item_type,
+                depth=depths.get(path, 0),
+                present=len(values),
+                null_count=len(values) - len(non_null),
+                distinct=len(hashable),
+                examples=[_example(value) for value in non_null[:max_examples]],
+            )
+        )
+
+    return ResourceProfile(
+        resource_name=sample.resource_name,
+        connector=sample.connector,
+        doc_count=len(sample.docs),
+        max_depth=max(depths.values(), default=0),
+        fields=fields,
+        primary_key=list(sample.primary_key),
+        foreign_keys=list(sample.foreign_keys),
+        truncated=truncated,
+    )
+
+
+def profile_source(
+    source_sample: SourceSample,
+    *,
+    max_paths: int = DEFAULT_MAX_PATHS,
+    max_examples: int = DEFAULT_MAX_EXAMPLES,
+) -> list[ResourceProfile]:
+    """Profile every resource in a :class:`SourceSample`."""
+    return [
+        profile_sample(sample, max_paths=max_paths, max_examples=max_examples)
+        for sample in source_sample.samples
+    ]
+
+
+__all__ = [
+    "DEFAULT_MAX_DOCS",
+    "DEFAULT_MAX_EXAMPLES",
+    "DEFAULT_MAX_PATHS",
+    "LIST_MARKER",
+    "FieldProfile",
+    "ForeignKeyHint",
+    "ResourceProfile",
+    "ResourceSample",
+    "SourceSample",
+    "infer_field_type",
+    "iter_paths",
+    "profile_sample",
+    "profile_source",
+]
diff --git a/graflo/architecture/pipeline/runtime/resource.py b/graflo/architecture/pipeline/runtime/resource.py
index 672b06a1..c63e1fa9 100644
--- a/graflo/architecture/pipeline/runtime/resource.py
+++ b/graflo/architecture/pipeline/runtime/resource.py
@@ -104,7 +104,9 @@ def __init__(
         self._vertex_config = runtime_vertex_config
         self._edge_config = local_edge_config
 
-        self._validate_vertex_references(vertex_config)
+        self._validate_vertex_references(
+            vertex_config, strict_references=strict_references
+        )
         self._validate_infer_edge_spec_targets(self._edge_config)
 
         edge_derivation_registry = EdgeDerivationRegistry()
@@ -190,8 +192,25 @@ def _filter_vertex_edge_configs(
         )
         return runtime_vertex_config, local_edge_config
 
-    def _validate_vertex_references(self, vertex_config: VertexConfig) -> None:
+    def _validate_vertex_references(
+        self, vertex_config: VertexConfig, *, strict_references: bool = False
+    ) -> None:
         known_vertices = set(vertex_config.vertex_set)
+
+        if strict_references:
+            # Vertex *steps* naming an undeclared vertex are otherwise silently
+            # dropped by filter_vertex_config_for_resource, so a pipeline that
+            # writes nothing validates clean. Under strict references that is an
+            # error: the resource claims to produce a vertex the schema lacks.
+            undeclared = sorted(self.collect_vertex_names() - known_vertices)
+            if undeclared:
+                raise ValueError(
+                    f"Resource '{self.config.name}' pipeline references undefined "
+                    f"vertices: {undeclared}. Declare them in vertex_config, or fix "
+                    "the step names (a common cause is a naming mismatch between "
+                    "the vertex definition and the pipeline step)."
+                )
+
         referenced_vertices: set[str] = set()
 
         for spec in self.config.infer_edge_only:
diff --git a/graflo/connections/onto.py b/graflo/connections/onto.py
index 4fc120d0..032163e2 100644
--- a/graflo/connections/onto.py
+++ b/graflo/connections/onto.py
@@ -1,6 +1,7 @@
 import abc
 import logging
 import os
+import re
 import warnings
 from pathlib import Path
 from typing import TYPE_CHECKING, Any, Literal, Self, TypeVar, cast
@@ -1275,6 +1276,54 @@ def to_sqlalchemy_connection_string(self) -> str:
         else:
             return f"postgresql://{user_encoded}@{host}:{port}/{database_encoded}"
 
+    @classmethod
+    def from_dsn(cls, dsn: str, **overrides: Any) -> "PostgresConfig":
+        """Build a config from a libpq/SQLAlchemy DSN.
+
+        ``uri`` alone is not enough: ``hostname`` and ``port`` are derived from
+        it, but ``username`` / ``password`` / ``database`` are independent fields,
+        so a config built from a bare URI has no database name and
+        :meth:`to_sqlalchemy_connection_string` raises. This parses the DSN and
+        populates those fields.
+
+        Args:
+            dsn: e.g. ``postgresql://user:pass@host:5432/dbname?options=-csearch_path=sales``
+            **overrides: Explicit field values that win over the parsed DSN.
+
+        Returns:
+            PostgresConfig: Config with credentials and database populated.
+        """
+        from urllib.parse import parse_qs, unquote, urlsplit
+
+        parts = urlsplit(dsn)
+        if not parts.hostname:
+            raise ValueError(f"DSN has no host: {dsn!r}")
+
+        fields: dict[str, Any] = {
+            "uri": f"{parts.scheme or 'postgresql'}://{parts.netloc}"
+        }
+        if parts.username:
+            fields["username"] = unquote(parts.username)
+        if parts.password:
+            fields["password"] = unquote(parts.password)
+        database = unquote(parts.path).lstrip("/")
+        if database:
+            fields["database"] = database
+
+        # Honour a search_path passed either as ?schema= or in libpq options.
+        query = parse_qs(parts.query)
+        schema = next(iter(query.get("schema", [])), None)
+        if schema is None:
+            options = next(iter(query.get("options", [])), "")
+            match = re.search(r"-c\s*search_path=([^\s,]+)", options)
+            if match:
+                schema = match.group(1)
+        if schema:
+            fields["schema_name"] = schema
+
+        fields.update(overrides)
+        return cls(**fields)
+
     @classmethod
     def from_docker_env(cls, docker_dir: str | Path | None = None) -> "PostgresConfig":
         """Load PostgreSQL config from docker/postgres/.env file."""
diff --git a/graflo/data_source/chunker.py b/graflo/data_source/chunker.py
index ddc09a72..500a987d 100644
--- a/graflo/data_source/chunker.py
+++ b/graflo/data_source/chunker.py
@@ -266,7 +266,15 @@ def _prepare_iteration(self):
         # After super()._prepare_iteration(), file_obj is guaranteed to be open
         if self.file_obj is None:
             raise RuntimeError("File should be opened by parent _prepare_iteration()")
-        header_line = next(self.file_obj)
+        try:
+            header_line = next(self.file_obj)
+        except StopIteration:
+            # An empty file has no header row. Leave the header empty and let
+            # iteration yield nothing; raising StopIteration from here would
+            # surface as `RuntimeError: generator raised StopIteration` (PEP 479)
+            # to any caller iterating batches.
+            self.header = []
+            return
         if isinstance(header_line, bytes):
             header_line = header_line.decode(self.encoding or "utf-8")
         self.header = header_line.rstrip("\n").split(self.sep)
@@ -486,10 +494,15 @@ def _guess_chunker_type(cls, filename: Path) -> ChunkerType:
             ValueError: If file extension is not recognized
         """
         # Get all suffixes and remove compression extensions
-        suffixes = filename.suffixes
-        base_suffix = [y for y in suffixes if y.lower() not in (".gz", ".zip")][
-            -1
-        ].lower()
+        suffixes = [y for y in filename.suffixes if y.lower() not in (".gz", ".zip")]
+        if not suffixes:
+            # A file with no extension (or only .gz/.zip) has nothing to go on.
+            # Raise the documented ValueError rather than an IndexError, so
+            # callers scanning a directory can skip it like any other unknown type.
+            raise ValueError(
+                f"Could not guess chunker type: '{filename.name}' has no file extension"
+            )
+        base_suffix = suffixes[-1].lower()
 
         if base_suffix == ".json":
             return ChunkerType.JSON
diff --git a/graflo/db/arango/conn.py b/graflo/db/arango/conn.py
index 27fdf113..55302a59 100644
--- a/graflo/db/arango/conn.py
+++ b/graflo/db/arango/conn.py
@@ -899,7 +899,7 @@ def insert_edges_batch(
             docs_edges = docs_edges[:head]
         if filter_uniques:
             docs_edges = pick_unique_dict(docs_edges)
-        docs_edges_str = json.dumps(docs_edges)
+        docs_edges_str = json.dumps(docs_edges, default=json_serializer)
 
         result_from, source_filter = _arango_edge_endpoint_aql(
             source_class, match_keys_source, 0
diff --git a/graflo/hq/graph_engine.py b/graflo/hq/graph_engine.py
index 4260a21e..46f8119e 100644
--- a/graflo/hq/graph_engine.py
+++ b/graflo/hq/graph_engine.py
@@ -14,6 +14,7 @@
 from graflo.architecture.contract.ingestion import IngestionModel
 from graflo.architecture.contract.manifest import GraphManifest
 from graflo.architecture.graph_types import GraphContainer
+from graflo.architecture.onto_sample import DEFAULT_MAX_DOCS, SourceSample
 from graflo.architecture.onto_sql import SchemaIntrospectionResult
 from graflo.architecture.schema import GraFloOutput, Schema
 from graflo.connections.onto import DBConfig, PostgresConfig, SparqlEndpointConfig
@@ -134,6 +135,51 @@ def introspect(
                 include_raw_tables=include_raw_tables,
             )
 
+    def sample_resources(
+        self,
+        source: "PostgresConfig | Bindings | Path | str | list[Path | str]",
+        *,
+        schema_name: str | None = None,
+        resources: list[str] | None = None,
+        max_docs: int = DEFAULT_MAX_DOCS,
+        source_name: str | None = None,
+    ) -> "SourceSample":
+        """Sample documents from a source, keeping connector provenance.
+
+        The returned :class:`~graflo.architecture.onto_sample.SourceSample` holds
+        pure JSON documents per resource. It is the shared input for schema
+        inference — algorithmic or agentic — and
+        ``SourceSample.samples_by_resource`` is the shape cross-resource identity
+        inference consumes directly.
+
+        Args:
+            source: A ``PostgresConfig``, a ``Bindings`` block, or a file/directory path.
+            schema_name: PostgreSQL schema to sample (PostgreSQL sources only).
+            resources: Restrict sampling to these resource names.
+            max_docs: Cap on documents fetched per resource.
+            source_name: Override the logical source name.
+
+        Returns:
+            SourceSample: Per-resource samples with connector provenance.
+        """
+        from graflo.hq.sampler import ResourceSampler
+
+        sampler = ResourceSampler(max_docs=max_docs)
+        if isinstance(source, PostgresConfig):
+            return sampler.sample_postgres(
+                source,
+                schema_name=schema_name,
+                tables=resources,
+                source_name=source_name,
+            )
+        if isinstance(source, Bindings):
+            return sampler.sample_bindings(
+                source,
+                resources=resources,
+                source_name=source_name or "bindings",
+            )
+        return sampler.sample_files(source, source_name=source_name)
+
     def infer_manifest(
         self,
         postgres_config: PostgresConfig,
diff --git a/graflo/hq/sampler.py b/graflo/hq/sampler.py
new file mode 100644
index 00000000..fe2e441e
--- /dev/null
+++ b/graflo/hq/sampler.py
@@ -0,0 +1,407 @@
+"""Connector-driven resource sampling.
+
+Pulls a bounded number of documents from each resource behind a set of
+connectors and returns them as a :class:`~graflo.architecture.onto_sample.SourceSample`
+— pure JSON, with the originating connector recorded on every sample.
+
+This is the input stage shared by every schema inferencer. ``infer_manifest``
+performs it privately for PostgreSQL; exposing it lets an agentic or algorithmic
+inferencer work from the same material, for any connector kind.
+
+Sampling is deliberately *not* profiling: this module fetches documents and does
+not describe them. See :func:`~graflo.architecture.onto_sample.profile_sample`
+for the derived typed view.
+"""
+
+from __future__ import annotations
+
+import logging
+import re
+from datetime import date, datetime, time
+from decimal import Decimal
+from pathlib import Path
+from typing import Any
+from uuid import UUID
+
+from graflo.architecture.contract.bindings.connectors import (
+    FileConnector,
+    ResourceConnector,
+    TableConnector,
+)
+from graflo.architecture.contract.bindings.core import Bindings
+from graflo.architecture.onto_sample import (
+    DEFAULT_MAX_DOCS,
+    ForeignKeyHint,
+    ResourceSample,
+    SourceSample,
+)
+from graflo.connections.onto import PostgresConfig
+from graflo.data_source.factory import DataSourceFactory
+from graflo.hq.registry_builder import RegistryBuilder
+
+logger = logging.getLogger(__name__)
+
+#: Values longer than this are clipped and the sample marked ``truncated``.
+DEFAULT_MAX_CELL_CHARS = 512
+
+
+def _jsonable(value: Any, *, max_cell_chars: int) -> tuple[Any, bool]:
+    """Coerce a fetched value to something JSON-serialisable.
+
+    Database and columnar readers hand back ``datetime``, ``Decimal``,
+    ``memoryview`` and numpy scalars. ``dict[str, Any]`` accepts them but only
+    serialises them best-effort, so normalise at sample time rather than leaving
+    it to whatever writes the JSON.
+
+    Returns the coerced value and whether it was clipped.
+    """
+    if value is None or isinstance(value, (bool, int)):
+        return value, False
+    if isinstance(value, float):
+        return value, False
+    if isinstance(value, str):
+        if len(value) > max_cell_chars:
+            return value[:max_cell_chars], True
+        return value, False
+    if isinstance(value, (datetime, date, time)):
+        return value.isoformat(), False
+    if isinstance(value, Decimal):
+        return float(value), False
+    if isinstance(value, UUID):
+        return str(value), False
+    if isinstance(value, (bytes, bytearray, memoryview)):
+        return f"<{len(bytes(value))} bytes>", True
+    if isinstance(value, dict):
+        out: dict[str, Any] = {}
+        clipped = False
+        for key, item in value.items():
+            out[str(key)], item_clipped = _jsonable(item, max_cell_chars=max_cell_chars)
+            clipped = clipped or item_clipped
+        return out, clipped
+    if isinstance(value, (list, tuple, set, frozenset)):
+        items: list[Any] = []
+        clipped = False
+        for item in value:
+            coerced, item_clipped = _jsonable(item, max_cell_chars=max_cell_chars)
+            items.append(coerced)
+            clipped = clipped or item_clipped
+        return items, clipped
+    # numpy scalars and anything else with a scalar view
+    item_method = getattr(value, "item", None)
+    if callable(item_method):
+        try:
+            return _jsonable(item_method(), max_cell_chars=max_cell_chars)
+        except Exception:
+            pass
+    return str(value), False
+
+
+class ResourceSampler:
+    """Fetch bounded document samples from connectors.
+
+    Args:
+        max_docs: Cap on documents fetched per resource.
+        max_cell_chars: Cap on the length of any single string value.
+    """
+
+    def __init__(
+        self,
+        *,
+        max_docs: int = DEFAULT_MAX_DOCS,
+        max_cell_chars: int = DEFAULT_MAX_CELL_CHARS,
+    ) -> None:
+        if max_docs < 1:
+            raise ValueError("max_docs must be at least 1")
+        self.max_docs = max_docs
+        self.max_cell_chars = max_cell_chars
+
+    # ------------------------------------------------------------------
+    # Normalisation
+    # ------------------------------------------------------------------
+
+    def _normalize_docs(
+        self, docs: list[dict[str, Any]]
+    ) -> tuple[list[dict[str, Any]], bool]:
+        normalized: list[dict[str, Any]] = []
+        truncated = False
+        for doc in docs[: self.max_docs]:
+            coerced, clipped = _jsonable(doc, max_cell_chars=self.max_cell_chars)
+            normalized.append(coerced)
+            truncated = truncated or clipped
+        if len(docs) > self.max_docs:
+            truncated = True
+        return normalized, truncated
+
+    def _read_data_source(self, data_source: Any) -> tuple[list[dict[str, Any]], bool]:
+        # Read one document past the cap so ``_normalize_docs`` can tell a source
+        # that happened to hold exactly ``max_docs`` from one that was truncated.
+        probe = self.max_docs + 1
+        docs: list[dict[str, Any]] = []
+        for batch in data_source.iter_batches(batch_size=probe, limit=probe):
+            docs.extend(batch)
+            if len(docs) >= probe:
+                break
+        return self._normalize_docs(docs)
+
+    # ------------------------------------------------------------------
+    # Files
+    # ------------------------------------------------------------------
+
+    def sample_file(
+        self,
+        path: Path | str,
+        *,
+        resource_name: str | None = None,
+        connector_name: str | None = None,
+    ) -> ResourceSample:
+        """Sample a single file, recording the connector that would read it."""
+        path = Path(path)
+        connector = FileConnector(
+            name=connector_name or path.stem,
+            regex=f"^{re.escape(path.name)}$",
+            sub_path=path.parent,
+        )
+        data_source = DataSourceFactory.create_file_data_source(path=path)
+        docs, truncated = self._read_data_source(data_source)
+        return ResourceSample(
+            resource_name=resource_name or path.stem,
+            connector=connector.name,
+            docs=docs,
+            truncated=truncated,
+            description=f"Sampled from file {path.name}",
+        )
+
+    def sample_files(
+        self,
+        paths_or_dir: Path | str | list[Path | str],
+        *,
+        source_name: str | None = None,
+    ) -> SourceSample:
+        """Sample every readable file in a directory, or an explicit file list.
+
+        Files that cannot be read, or that yield no documents, are skipped with a
+        warning rather than failing the whole sample — a source directory
+        routinely holds READMEs and notes beside its data.
+        """
+        if isinstance(paths_or_dir, (str, Path)):
+            root = Path(paths_or_dir)
+            if root.is_dir():
+                paths = sorted(p for p in root.iterdir() if p.is_file())
+                default_name = root.name
+            else:
+                paths = [root]
+                default_name = root.stem
+        else:
+            paths = [Path(p) for p in paths_or_dir]
+            default_name = paths[0].parent.name if paths else "source"
+
+        samples: list[ResourceSample] = []
+        for path in paths:
+            try:
+                sample = self.sample_file(path)
+            except (ValueError, OSError) as exc:
+                logger.warning("Skipping unsampleable file '%s': %s", path, exc)
+                continue
+            if not sample.docs:
+                logger.warning("Skipping file '%s': yielded no documents", path)
+                continue
+            samples.append(sample)
+
+        if not samples:
+            raise ValueError(f"No sampleable files found in {paths_or_dir}")
+
+        return SourceSample(source_name=source_name or default_name, samples=samples)
+
+    # ------------------------------------------------------------------
+    # PostgreSQL
+    # ------------------------------------------------------------------
+
+    def sample_postgres(
+        self,
+        config: PostgresConfig,
+        *,
+        schema_name: str | None = None,
+        tables: list[str] | None = None,
+        source_name: str | None = None,
+    ) -> SourceSample:
+        """Sample PostgreSQL tables, carrying declared keys through.
+
+        Primary and foreign keys come from schema introspection, not from column
+        naming — they are ground truth for edge inference, so an inferencer never
+        has to guess at ``*_id`` suffixes when a real constraint exists.
+        """
+        from graflo.db.postgres.conn import PostgresConnection
+        from graflo.hq.sql_inferencer import SQLInferenceManager
+
+        effective_schema = schema_name or config.schema_name or "public"
+        samples: list[ResourceSample] = []
+
+        with PostgresConnection(config) as conn:
+            inferencer = SQLInferenceManager(
+                conn=conn, target_db_flavor=self.postgres_target_flavor()
+            )
+            introspection = inferencer.introspect(
+                schema_name=effective_schema, include_raw_tables=True
+            )
+            for table in introspection.raw_tables:
+                if tables is not None and table.name not in tables:
+                    continue
+                connector = TableConnector(
+                    name=table.name,
+                    table_name=table.name,
+                    schema_name=table.schema_name,
+                )
+                rows = conn.get_table_sample_rows(
+                    table.name, schema_name=table.schema_name, limit=self.max_docs
+                )
+                docs, truncated = self._normalize_docs(rows)
+                samples.append(
+                    ResourceSample(
+                        resource_name=table.name,
+                        connector=connector.name,
+                        docs=docs,
+                        primary_key=list(table.primary_key),
+                        foreign_keys=[
+                            ForeignKeyHint(
+                                field=fk.column,
+                                references_resource=fk.references_table,
+                                references_field=fk.references_column,
+                            )
+                            for fk in table.foreign_keys
+                        ],
+                        truncated=truncated,
+                        total_estimate=table.row_count_estimate,
+                        description=f"Sampled from table {table.schema_name}.{table.name}",
+                    )
+                )
+
+        if not samples:
+            raise ValueError(
+                f"No tables found to sample in schema '{effective_schema}'"
+            )
+
+        return SourceSample(
+            source_name=source_name or config.database or effective_schema,
+            samples=samples,
+        )
+
+    @staticmethod
+    def postgres_target_flavor() -> Any:
+        """Target flavour used for introspection type mapping."""
+        from graflo.connections.onto import DBType
+
+        return DBType.ARANGO
+
+    # ------------------------------------------------------------------
+    # Bindings
+    # ------------------------------------------------------------------
+
+    def sample_connector(
+        self,
+        connector: ResourceConnector,
+        *,
+        resource_name: str,
+        config: PostgresConfig | None = None,
+    ) -> ResourceSample:
+        """Sample one resource through *connector*."""
+        if isinstance(connector, FileConnector):
+            files = RegistryBuilder.discover_files(
+                connector.sub_path.expanduser(), connector=connector, limit_files=1
+            )
+            if not files:
+                raise ValueError(
+                    f"FileConnector for resource '{resource_name}' matched no files "
+                    f"under '{connector.sub_path}'"
+                )
+            sample = self.sample_file(
+                files[0],
+                resource_name=resource_name,
+                connector_name=connector.name or resource_name,
+            )
+            return sample
+
+        if isinstance(connector, TableConnector):
+            if config is None:
+                raise ValueError(
+                    f"Sampling TableConnector for resource '{resource_name}' requires "
+                    "a PostgresConfig"
+                )
+            from graflo.db.postgres.conn import PostgresConnection
+
+            with PostgresConnection(config) as conn:
+                rows = conn.get_table_sample_rows(
+                    connector.table_name,
+                    schema_name=connector.schema_name,
+                    limit=self.max_docs,
+                )
+            docs, truncated = self._normalize_docs(rows)
+            return ResourceSample(
+                resource_name=resource_name,
+                connector=connector.name or resource_name,
+                docs=docs,
+                truncated=truncated,
+            )
+
+        raise ValueError(
+            f"Sampling is not implemented for {type(connector).__name__} "
+            f"(resource '{resource_name}')"
+        )
+
+    def sample_bindings(
+        self,
+        bindings: Bindings,
+        *,
+        resources: list[str] | None = None,
+        config: PostgresConfig | None = None,
+        source_name: str = "bindings",
+    ) -> SourceSample:
+        """Sample every resource wired up in *bindings*.
+
+        The ``resource_connector`` mapping is the authority on which connector
+        feeds which resource, so the provenance recorded on each sample is the
+        same relation ingestion will later use.
+        """
+        connectors_by_ref: dict[str, ResourceConnector] = {}
+        for connector in bindings.connectors:
+            if connector.name:
+                connectors_by_ref[connector.name] = connector
+            connectors_by_ref[connector.hash] = connector
+
+        samples: list[ResourceSample] = []
+        for mapping in bindings.resource_connector:
+            resource = (
+                mapping.get("resource")
+                if isinstance(mapping, dict)
+                else mapping.resource
+            )
+            connector_ref = (
+                mapping.get("connector")
+                if isinstance(mapping, dict)
+                else mapping.connector
+            )
+            if resources is not None and resource not in resources:
+                continue
+            connector = connectors_by_ref.get(str(connector_ref))
+            if connector is None:
+                logger.warning(
+                    "Skipping resource '%s': connector '%s' not found in bindings",
+                    resource,
+                    connector_ref,
+                )
+                continue
+            try:
+                samples.append(
+                    self.sample_connector(
+                        connector, resource_name=str(resource), config=config
+                    )
+                )
+            except ValueError as exc:
+                logger.warning("Skipping resource '%s': %s", resource, exc)
+
+        if not samples:
+            raise ValueError("No resources could be sampled from the given bindings")
+
+        return SourceSample(source_name=source_name, samples=samples)
+
+
+__all__ = ["DEFAULT_MAX_CELL_CHARS", "ResourceSampler"]
diff --git a/graflo/onto.py b/graflo/onto.py
index 01a4913c..d995923b 100644
--- a/graflo/onto.py
+++ b/graflo/onto.py
@@ -41,7 +41,7 @@ def __contains__(self, member: object) -> bool:
         """Check if an item is a valid member of the enum.
 
         Args:
-            item: Value to check for membership
+            member: Value to check for membership
 
         Returns:
             bool: True if the item is a valid enum member, False otherwise
diff --git a/graflo/rdf/deserializer.py b/graflo/rdf/deserializer.py
index 59a5fd4d..b340f504 100644
--- a/graflo/rdf/deserializer.py
+++ b/graflo/rdf/deserializer.py
@@ -115,14 +115,9 @@ def _parse_vertex_config(
         return vertex_config
 
     def _parse_vertex(self, graph: Graph, vertex_uri: URIRef | BNode) -> dict[str, Any]:
-        identities = []
-        for identity_node in self._related_nodes(graph, vertex_uri, ns.hasIdentity):
-            identity = self._literal(graph, identity_node, ns.identityName)
-            if identity is not None:
-                identities.append(identity)
         vertex: dict[str, Any] = {
             "name": self._literal(graph, vertex_uri, ns.name),
-            "identity": identities,
+            "identity": self._identity_names(graph, vertex_uri, ns.hasIdentity),
             "properties": self._ordered_nodes(
                 graph,
                 vertex_uri,
@@ -136,6 +131,22 @@ def _parse_vertex(self, graph: Graph, vertex_uri: URIRef | BNode) -> dict[str, A
         blank_value = self._literal(graph, vertex_uri, ns.blank)
         if blank_value is not None:
             vertex["blank"] = blank_value.lower() == "true"
+        assigned_value = self._literal(graph, vertex_uri, ns.assigned)
+        if assigned_value is not None:
+            vertex["assigned"] = assigned_value.lower() == "true"
+
+        hash_identity = self._identity_names(graph, vertex_uri, ns.hasHashIdentity)
+        if hash_identity:
+            vertex["hash_identity_properties"] = hash_identity
+
+        secondary = self._ordered_nodes(
+            graph,
+            vertex_uri,
+            ns.hasSecondaryIdentity,
+            self._parse_secondary_identity,
+        )
+        if secondary:
+            vertex["secondary_identities"] = secondary
 
         payload = parse_json_literal(self._literal(graph, vertex_uri, ns.vertexPayload))
         if isinstance(payload, dict):
@@ -143,6 +154,39 @@ def _parse_vertex(self, graph: Graph, vertex_uri: URIRef | BNode) -> dict[str, A
 
         return vertex
 
+    def _identity_names(
+        self,
+        graph: Graph,
+        owner: URIRef | BNode,
+        predicate: URIRef,
+    ) -> list[str]:
+        """Ordered identity token values under ``predicate``.
+
+        Graphs written before ``gf:artifactIndex`` reached identity nodes have no
+        index; ``_ordered_nodes`` treats those as index 0, so they degrade to
+        arbitrary order rather than failing.
+        """
+        names = self._ordered_nodes(
+            graph,
+            owner,
+            predicate,
+            lambda g, node: self._literal(g, node, ns.identityName),
+        )
+        return [name for name in names if name is not None]
+
+    def _parse_secondary_identity(
+        self, graph: Graph, node: URIRef | BNode
+    ) -> dict[str, Any]:
+        secondary: dict[str, Any] = {
+            "fields": self._identity_names(graph, node, ns.hasIdentity)
+        }
+        # Omit rather than null: an unnamed entry must reach SecondaryIdentity's
+        # auto-naming validator, which assigns ``secondary_{position}``.
+        name = self._literal(graph, node, ns.name)
+        if name is not None:
+            secondary["name"] = name
+        return secondary
+
     def _parse_field(
         self, graph: Graph, field_uri: URIRef | BNode
     ) -> dict[str, Any] | str:
diff --git a/graflo/rdf/namespace.py b/graflo/rdf/namespace.py
index 91020df5..f58c7637 100644
--- a/graflo/rdf/namespace.py
+++ b/graflo/rdf/namespace.py
@@ -16,7 +16,7 @@
 )
 
 GF_ONTOLOGY_IRI = "https://ontology.growgraph.dev/graflo"
-GF_VERSION = "1.0.0"
+GF_VERSION = "1.1.0"
 GF_VERSION_IRI = f"{GF_ONTOLOGY_IRI}/{GF_VERSION}"
 GF_BASE = "https://ontology.growgraph.dev/graflo/"
 GF = Namespace(GF_BASE)
@@ -33,6 +33,7 @@
 Edge = GF.Edge
 Field = GF.Field
 Identity = GF.Identity
+SecondaryIdentity = GF.SecondaryIdentity
 IngestionModel = GF.IngestionModel
 Resource = GF.Resource
 EdgeInferSpec = GF.EdgeInferSpec
@@ -71,6 +72,8 @@
 hasEdge = GF.hasEdge
 hasField = GF.hasField
 hasIdentity = GF.hasIdentity
+hasHashIdentity = GF.hasHashIdentity
+hasSecondaryIdentity = GF.hasSecondaryIdentity
 edgeSource = GF.edgeSource
 edgeTarget = GF.edgeTarget
 hasResource = GF.hasResource
@@ -103,6 +106,7 @@
 identityName = GF.identityName
 relation = GF.relation
 blank = GF.blank
+assigned = GF.assigned
 edgesOnDuplicate = GF.edgesOnDuplicate
 resourceName = GF.resourceName
 connectorName = GF.connectorName
diff --git a/graflo/rdf/ontology/graflo-context.jsonld b/graflo/rdf/ontology/graflo-context.jsonld
index cbf107ee..baa2e05a 100644
--- a/graflo/rdf/ontology/graflo-context.jsonld
+++ b/graflo/rdf/ontology/graflo-context.jsonld
@@ -14,6 +14,7 @@
         "Edge": "gf:Edge",
         "Field": "gf:Field",
         "Identity": "gf:Identity",
+        "SecondaryIdentity": "gf:SecondaryIdentity",
         "IngestionModel": "gf:IngestionModel",
         "Resource": "gf:Resource",
         "ProtoTransform": "gf:ProtoTransform",
@@ -85,6 +86,14 @@
             "@id": "gf:hasIdentity",
             "@type": "@id"
         },
+        "hasHashIdentity": {
+            "@id": "gf:hasHashIdentity",
+            "@type": "@id"
+        },
+        "hasSecondaryIdentity": {
+            "@id": "gf:hasSecondaryIdentity",
+            "@type": "@id"
+        },
         "edgeSource": {
             "@id": "gf:edgeSource",
             "@type": "@id"
@@ -201,6 +210,10 @@
             "@id": "gf:blank",
             "@type": "xsd:boolean"
         },
+        "assigned": {
+            "@id": "gf:assigned",
+            "@type": "xsd:boolean"
+        },
         "edgesOnDuplicate": {
             "@id": "gf:edgesOnDuplicate",
             "@type": "@id"
diff --git a/graflo/rdf/ontology/graflo.ttl b/graflo/rdf/ontology/graflo.ttl
index 558aed5a..2a1dd807 100644
--- a/graflo/rdf/ontology/graflo.ttl
+++ b/graflo/rdf/ontology/graflo.ttl
@@ -7,8 +7,8 @@
 @prefix prov: <http://www.w3.org/ns/prov#> .
 
 <https://ontology.growgraph.dev/graflo> a owl:Ontology ;
-    owl:versionIRI <https://ontology.growgraph.dev/graflo/1.0.0> ;
-    owl:versionInfo "1.0.0" ;
+    owl:versionIRI <https://ontology.growgraph.dev/graflo/1.1.0> ;
+    owl:versionInfo "1.1.0" ;
     rdfs:label "GraFlo Ontology"@en ;
     skos:prefLabel "GraFlo Ontology"@en ;
     rdfs:comment "OWL vocabulary describing GraFlo GraphManifest contracts: schema, ingestion, bindings, transforms, and pipeline actors."@en .
@@ -82,6 +82,12 @@ gf:Identity a owl:Class ;
     skos:prefLabel "Identity"@en ;
     rdfs:comment "Manifest-facing identity element (string) promoted to a first-class ontology node for traceability."@en .
 
+gf:SecondaryIdentity a owl:Class ;
+    rdfs:subClassOf gf:GrafloArtifact ;
+    rdfs:label "Secondary identity"@en ;
+    skos:prefLabel "SecondaryIdentity"@en ;
+    rdfs:comment "Alternate named field-set that identifies a vertex for lookup only. Edge endpoints may match on it; upserts always use the primary identity."@en .
+
 gf:IngestionModel a owl:Class ;
     rdfs:subClassOf gf:GrafloArtifact ;
     rdfs:label "Ingestion model"@en ;
@@ -285,9 +291,19 @@ gf:hasField a owl:ObjectProperty ;
     rdfs:range gf:Field .
 
 gf:hasIdentity a owl:ObjectProperty ;
-    rdfs:domain gf:Vertex , gf:Edge ;
+    rdfs:domain gf:Vertex , gf:Edge , gf:SecondaryIdentity ;
+    rdfs:range gf:Identity ;
+    rdfs:comment "Manifest-facing identity chain node for vertex, edge and secondary-identity field lists."@en .
+
+gf:hasHashIdentity a owl:ObjectProperty ;
+    rdfs:domain gf:Vertex ;
     rdfs:range gf:Identity ;
-    rdfs:comment "Manifest-facing identity chain node for vertex and edge identities."@en .
+    rdfs:comment "Source field whose value contributes to the deterministic SHA256 synthetic id (vertex.hash_identity_properties)."@en .
+
+gf:hasSecondaryIdentity a owl:ObjectProperty ;
+    rdfs:domain gf:Vertex ;
+    rdfs:range gf:SecondaryIdentity ;
+    rdfs:comment "Alternate lookup-only field-set declared in vertex.secondary_identities."@en .
 
 gf:edgeSource a owl:ObjectProperty ;
     rdfs:domain gf:Edge ;
@@ -416,6 +432,11 @@ gf:blank a owl:DatatypeProperty ;
     rdfs:domain gf:Vertex ;
     rdfs:range xsd:boolean .
 
+gf:assigned a owl:DatatypeProperty ;
+    rdfs:domain gf:Vertex ;
+    rdfs:range xsd:boolean ;
+    rdfs:comment "True when the vertex uses an intentional UUID primary key filled at assemble time. Mutually exclusive with gf:blank."@en .
+
 gf:edgesOnDuplicate a owl:ObjectProperty ;
     rdfs:domain gf:IngestionModel ;
     rdfs:range gf:EdgeDuplicatePolicy .
diff --git a/graflo/rdf/serializer.py b/graflo/rdf/serializer.py
index fb62bff8..448aa075 100644
--- a/graflo/rdf/serializer.py
+++ b/graflo/rdf/serializer.py
@@ -169,12 +169,22 @@ def _emit_vertex(self, graph: Graph, vertex_uri: URIRef, vertex: Vertex) -> None
         add_literal(graph, vertex_uri, ns.name, vertex.name)
         add_literal(graph, vertex_uri, ns.description, vertex.description)
         add_literal(graph, vertex_uri, ns.blank, vertex.blank)
+        add_literal(graph, vertex_uri, ns.assigned, vertex.assigned)
 
-        for identity in vertex.identity:
-            identity_node = BNode()
-            graph.add((vertex_uri, ns.hasIdentity, identity_node))
-            graph.add((identity_node, RDF.type, ns.Identity))
-            add_literal(graph, identity_node, ns.identityName, identity)
+        self._emit_identity_names(graph, vertex_uri, ns.hasIdentity, vertex.identity)
+        self._emit_identity_names(
+            graph, vertex_uri, ns.hasHashIdentity, vertex.hash_identity_properties
+        )
+
+        for index, secondary in enumerate(vertex.secondary_identities):
+            secondary_node = BNode()
+            graph.add((vertex_uri, ns.hasSecondaryIdentity, secondary_node))
+            graph.add((secondary_node, RDF.type, ns.SecondaryIdentity))
+            add_literal(graph, secondary_node, ns.artifactIndex, index)
+            add_literal(graph, secondary_node, ns.name, secondary.name)
+            self._emit_identity_names(
+                graph, secondary_node, ns.hasIdentity, secondary.fields
+            )
 
         payload = {}
         if vertex.filters:
@@ -190,6 +200,26 @@ def _emit_vertex(self, graph: Graph, vertex_uri: URIRef, vertex: Vertex) -> None
         for index, field in enumerate(vertex.properties):
             self._emit_field(graph, vertex_uri, field, index)
 
+    @staticmethod
+    def _emit_identity_names(
+        graph: Graph,
+        owner: URIRef | BNode,
+        predicate: URIRef,
+        names: list[str],
+    ) -> None:
+        """Emit an ordered chain of ``gf:Identity`` nodes under ``predicate``.
+
+        The index is what makes the chain ordered: identity lists are positional
+        and RDF triples are not, so a reader without ``gf:artifactIndex`` can
+        only recover the set, not the sequence.
+        """
+        for index, name in enumerate(names):
+            identity_node = BNode()
+            graph.add((owner, predicate, identity_node))
+            graph.add((identity_node, RDF.type, ns.Identity))
+            add_literal(graph, identity_node, ns.artifactIndex, index)
+            add_literal(graph, identity_node, ns.identityName, name)
+
     def _emit_field(
         self,
         graph: Graph,
diff --git a/graflo/util/transform.py b/graflo/util/transform.py
index 96bf1d6c..0345ae4c 100644
--- a/graflo/util/transform.py
+++ b/graflo/util/transform.py
@@ -10,7 +10,7 @@
     - cast_ibes_analyst: Parse and standardize analyst names
     - clear_first_level_nones: Clean dictionaries by removing None values
     - parse_multi_item: Parse complex multi-item strings
-    - pick_unique_dict: Remove duplicate dictionaries
+    - pick_unique_dict: Remove duplicate structures by content hash
 
 Example:
     >>> name = standardize("John. Doe, Smith")
@@ -24,6 +24,7 @@
 from collections import defaultdict
 from datetime import datetime
 from functools import lru_cache
+from typing import TypeVar
 
 ORDINAL_SUFFIX = ["st", "nd", "rd", "th"]
 _CAMEL_TO_SNAKE_STEP1_RE = re.compile(r"(.)([A-Z][a-z]+)")
@@ -33,8 +34,10 @@
 
 logger = logging.getLogger(__name__)
 
+T = TypeVar("T")
 
-def standardize(k):
+
+def standardize(k: str) -> str:
     """Standardizes a string key by removing periods and splitting.
 
     Handles comma and space-separated strings, normalizing their format.
@@ -51,17 +54,17 @@ def standardize(k):
         >>> standardize("John Doe Smith")
         'John,Doe,Smith'
     """
-    k = k.translate(str.maketrans({".": ""}))
+    cleaned = k.translate(str.maketrans({".": ""}))
     # try to split by ", "
-    k = k.split(", ")
-    if len(k) < 2:
-        k = k[0].split(" ")
+    parts = cleaned.split(", ")
+    if len(parts) < 2:
+        parts = parts[0].split(" ")
     else:
-        k[1] = k[1].translate(str.maketrans({" ": ""}))
-    return ",".join(k)
+        parts[1] = parts[1].translate(str.maketrans({" ": ""}))
+    return ",".join(parts)
 
 
-def parse_date_standard(input_str):
+def parse_date_standard(input_str: str) -> tuple[int, int, int]:
     """Parse a date string in YYYY-MM-DD format.
 
     Args:
@@ -78,7 +81,7 @@ def parse_date_standard(input_str):
     return dt.year, dt.month, dt.day
 
 
-def parse_date_conf(input_str):
+def parse_date_conf(input_str: str) -> tuple[int, int, int]:
     """Parse a date string in YYYYMMDD format.
 
     Args:
@@ -95,7 +98,7 @@ def parse_date_conf(input_str):
     return dt.year, dt.month, dt.day
 
 
-def parse_date_ibes(date0, time0):
+def parse_date_ibes(date0: str | int, time0: str) -> str:
     """Converts IBES date and time to ISO 8601 format datetime.
 
     Args:
@@ -116,7 +119,7 @@ def parse_date_ibes(date0, time0):
     return full_datetime
 
 
-def parse_date_yahoo(date0):
+def parse_date_yahoo(date0: str) -> str:
     """Convert Yahoo Finance date to ISO 8601 format.
 
     Args:
@@ -133,7 +136,7 @@ def parse_date_yahoo(date0):
     return full_datetime
 
 
-def round_str(x, **kwargs):
+def round_str(x: str, **kwargs) -> float:
     """Round a string number to specified precision.
 
     Args:
@@ -150,7 +153,7 @@ def round_str(x, **kwargs):
     return round(float(x), **kwargs)
 
 
-def parse_date_standard_to_epoch(input_str):
+def parse_date_standard_to_epoch(input_str: str) -> float:
     """Convert standard date string to Unix epoch timestamp.
 
     Args:
@@ -168,7 +171,7 @@ def parse_date_standard_to_epoch(input_str):
     return timestamp
 
 
-def cast_ibes_analyst(s):
+def cast_ibes_analyst(s: str) -> tuple[str, str]:
     """Splits and normalizes analyst name strings.
 
     Handles various name formats like 'ADKINS/NARRA' or 'ARFSTROM      J'.
@@ -203,7 +206,7 @@ def cast_ibes_analyst(s):
             return r[0], r[1][:1]
 
 
-def parse_date_reference(input_str):
+def parse_date_reference(input_str: str) -> int | str:
     """Extract year from a date reference string.
 
     Args:
@@ -219,7 +222,7 @@ def parse_date_reference(input_str):
     return _parse_date_reference(input_str)["year"]
 
 
-def _parse_date_reference(input_str):
+def _parse_date_reference(input_str: str) -> dict[str, int | str]:
     """Parse complex, human-written date references.
 
     Handles various date formats like:
@@ -296,7 +299,9 @@ def try_int(x):
         return x
 
 
-def clear_first_level_nones(docs, keys_keep_nones: list | None = None):
+def clear_first_level_nones(
+    docs: list[dict], keys_keep_nones: list | None = None
+) -> list[dict]:
     """Removes None values from dictionaries, with optional key exceptions.
 
     Args:
@@ -319,7 +324,7 @@ def clear_first_level_nones(docs, keys_keep_nones: list | None = None):
     return docs
 
 
-def parse_multi_item(s, mapper: dict, direct: list):
+def parse_multi_item(s: str, mapper: dict, direct: list) -> defaultdict[str, list]:
     """Parses complex multi-item strings into structured data.
 
     Supports parsing strings with quoted or bracketed items.
@@ -371,17 +376,18 @@ def parse_multi_item(s, mapper: dict, direct: list):
     return r
 
 
-def pick_unique_dict(docs):
-    """Removes duplicate dictionaries from a list.
+def pick_unique_dict(docs: list[T]) -> list[T]:
+    """Remove duplicate structures from a list by content hash.
 
-    Uses a hash-based approach to identify unique dictionaries, which is more
-    efficient than JSON serialization and preserves original object types.
+    Uses a hash-based approach that handles nested dicts, lists/tuples,
+    datetime objects, and Decimal types. Preserves original objects and
+    insertion order. Works for vertex dicts and edge triples alike.
 
     Args:
-        docs (list): List of dictionaries.
+        docs: List of structures to deduplicate.
 
     Returns:
-        list: List of unique dictionaries (preserving original objects).
+        List of unique structures (preserving original objects).
 
     Example:
         >>> docs = [{"a": 1}, {"a": 1}, {"b": 2}]
@@ -421,7 +427,7 @@ def make_hashable(obj):
             return obj
 
     # Use a dict to preserve insertion order and original objects
-    seen = {}
+    seen: dict = {}
     for doc in docs:
         # Create hashable representation
         hashable_repr = make_hashable(doc)
diff --git a/mkdocs.yml b/mkdocs.yml
index 88dc04ba..1c66e843 100644
--- a/mkdocs.yml
+++ b/mkdocs.yml
@@ -61,6 +61,7 @@ nav:
         -   Core components: concepts/architecture/core_components.md
         -   Capabilities: concepts/architecture/capabilities.md
     -   Schema and manifest:
+        -   Sampling and profiling: concepts/schema/sampling_and_profiling.md
         -   Vertex identity: concepts/schema/vertex_identity.md
         -   Backend indexes: concepts/schema/backend_indexes.md
         -   Manifest evolution: concepts/schema/manifest_evolution.md
@@ -81,6 +82,7 @@ nav:
     -   Guides index: guides/index.md
     -   Importing and layering: guides/importing.md
     -   Graph DB migration: guides/graph_db_migration.md
+    -   Graph namespace and schema: guides/graph_namespace_and_schema.md
     -   Graph export and replay: guides/graph_export_and_replay.md
     -   API env wiring: guides/api_env_wiring.md
     -   Identity inference: guides/identity_inference.md
diff --git a/pyproject.toml b/pyproject.toml
index d61ff317..da7045bd 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -41,7 +41,7 @@ description = "Manifest-driven Graph Schema & Transformation Language (GSTL): de
 name = "graflo"
 readme = "README.md"
 requires-python = ">=3.11"
-version = "1.10.0"
+version = "1.10.1"
 
 [project.optional-dependencies]
 dev = [
@@ -57,7 +57,8 @@ docs = [
   "mkdocs-material>=9.6.12",
   "mkdocs-table-reader-plugin>=3.1.0",
   "mkdocs>=1.6.1",
-  "mkdocstrings[python]>=0.29.1"
+  "mkdocstrings[python]>=0.29.1",
+  "ruff"
 ]
 plot = [
   "pygraphviz>=1.14"
diff --git a/test/architecture/test_layering.py b/test/architecture/test_layering.py
index 559be28d..0ce601ad 100644
--- a/test/architecture/test_layering.py
+++ b/test/architecture/test_layering.py
@@ -51,6 +51,7 @@
     "graflo.architecture.evolution": 4,
     "graflo.architecture.util": 1,  # helpers over graph_types only
     "graflo.architecture.onto_sql": 1,  # leaf pydantic models (SQL introspection)
+    "graflo.architecture.onto_sample": 2,  # leaf sample models; needs FieldType (L2)
     "graflo.data_source": 4,
     "graflo.db": 5,
     "graflo.object_storage": 5,
diff --git a/test/architecture/test_manifest_canonical_contract.py b/test/architecture/test_manifest_canonical_contract.py
index 8f57fa8d..b25d3376 100644
--- a/test/architecture/test_manifest_canonical_contract.py
+++ b/test/architecture/test_manifest_canonical_contract.py
@@ -86,6 +86,52 @@ def test_ingestion_model_strict_transform_reference_fails_fast() -> None:
         assert "was not found in ingestion_model.transforms" in str(exc)
 
 
+def test_strict_mode_rejects_pipeline_vertex_step_naming_unknown_vertex() -> None:
+    """A vertex step naming an undeclared vertex is otherwise silently dropped.
+
+    ``filter_vertex_config_for_resource`` intersects the resource's vertex names
+    with the schema's, so a pipeline that writes nothing used to validate clean.
+    A name mismatch between the vertex definition and the step is the common
+    cause, and it produces a manifest that ingests no data at all.
+    """
+    schema = _minimal_schema()
+    ingestion_model = IngestionModel.model_validate(
+        {
+            "resources": [{"name": "r1", "pipeline": [{"vertex": "A"}]}],
+            "transforms": [],
+        }
+    )
+
+    try:
+        ingestion_model.finish_init(schema.core_schema, strict_references=True)
+        assert False, "Expected strict vertex reference validation to fail"
+    except ValueError as exc:
+        assert "references undefined vertices: ['A']" in str(exc)
+
+
+def test_lenient_mode_still_tolerates_unknown_pipeline_vertex_step() -> None:
+    """Non-strict behaviour is unchanged: unknown vertex steps are filtered out."""
+    schema = _minimal_schema()
+    ingestion_model = IngestionModel.model_validate(
+        {
+            "resources": [{"name": "r1", "pipeline": [{"vertex": "A"}]}],
+            "transforms": [],
+        }
+    )
+    ingestion_model.finish_init(schema.core_schema, strict_references=False)
+
+
+def test_strict_mode_accepts_matching_pipeline_vertex_step() -> None:
+    schema = _minimal_schema()
+    ingestion_model = IngestionModel.model_validate(
+        {
+            "resources": [{"name": "r1", "pipeline": [{"vertex": "a"}]}],
+            "transforms": [],
+        }
+    )
+    ingestion_model.finish_init(schema.core_schema, strict_references=True)
+
+
 def test_registry_builder_strict_mode_aggregates_missing_connectors() -> None:
     schema = _minimal_schema()
     ingestion_model = IngestionModel.model_validate(
diff --git a/test/data/sample-source/NOTES.txt b/test/data/sample-source/NOTES.txt
new file mode 100644
index 00000000..6e0a74d5
--- /dev/null
+++ b/test/data/sample-source/NOTES.txt
@@ -0,0 +1 @@
+These are notes, not data.
diff --git a/test/data/sample-source/api_orders.json b/test/data/sample-source/api_orders.json
new file mode 100644
index 00000000..324d145e
--- /dev/null
+++ b/test/data/sample-source/api_orders.json
@@ -0,0 +1,37 @@
+[
+    {
+        "order_id": "o1",
+        "customer": {
+            "id": "c1",
+            "city": "Berlin"
+        },
+        "items": [
+            {
+                "sku": "A-1",
+                "qty": 2
+            },
+            {
+                "sku": "B-2",
+                "qty": 1
+            }
+        ],
+        "tags": [
+            "priority",
+            "gift"
+        ]
+    },
+    {
+        "order_id": "o2",
+        "customer": {
+            "id": "c2",
+            "city": null
+        },
+        "items": [
+            {
+                "sku": "C-3",
+                "qty": 5
+            }
+        ],
+        "tags": []
+    }
+]
diff --git a/test/data/sample-source/customers.csv b/test/data/sample-source/customers.csv
new file mode 100644
index 00000000..23c3ed17
--- /dev/null
+++ b/test/data/sample-source/customers.csv
@@ -0,0 +1,4 @@
+id,email,name
+c1,alice@example.com,Alice
+c2,bob@example.com,Bob
+c3,cara@example.com,Cara
diff --git a/test/data/sample-source/orders.csv b/test/data/sample-source/orders.csv
new file mode 100644
index 00000000..77d36e6c
--- /dev/null
+++ b/test/data/sample-source/orders.csv
@@ -0,0 +1,4 @@
+id,customer_id,total,paid
+o1,c1,9.50,true
+o2,c2,3.00,false
+o3,c1,4.25,true
diff --git a/test/db/postgres/test_postgres_config_dsn.py b/test/db/postgres/test_postgres_config_dsn.py
new file mode 100644
index 00000000..e8933a53
--- /dev/null
+++ b/test/db/postgres/test_postgres_config_dsn.py
@@ -0,0 +1,74 @@
+"""``PostgresConfig.from_dsn`` — DSN parsing into connectable config.
+
+``uri`` alone derives only host and port; ``username`` / ``password`` /
+``database`` are independent fields, so a config built from a bare URI cannot
+produce a connection string. These tests pin the parsing that closes that gap.
+"""
+
+import pytest
+
+from graflo.connections.onto import PostgresConfig
+
+
+def test_from_dsn_populates_credentials_and_database():
+    config = PostgresConfig.from_dsn("postgresql://alice:secret@db.example:5433/shop")
+    assert config.hostname == "db.example"
+    assert str(config.port) == "5433"
+    assert config.username == "alice"
+    assert config.password == "secret"
+    assert config.database == "shop"
+
+
+def test_from_dsn_yields_usable_sqlalchemy_connection_string():
+    """The regression this guards: a bare uri= config raised here."""
+    config = PostgresConfig.from_dsn("postgresql://alice:secret@db.example:5433/shop")
+    assert (
+        config.to_sqlalchemy_connection_string()
+        == "postgresql://alice:secret@db.example:5433/shop"
+    )
+
+
+def test_from_dsn_percent_decodes_credentials():
+    config = PostgresConfig.from_dsn("postgresql://user%40corp:p%40ss@h/db")
+    assert config.username == "user@corp"
+    assert config.password == "p@ss"
+
+
+def test_from_dsn_applies_default_port():
+    config = PostgresConfig.from_dsn("postgresql://user@localhost/db")
+    assert str(config.port) == "5432"
+
+
+@pytest.mark.parametrize(
+    "dsn,expected",
+    [
+        ("postgresql://u@h/db?schema=sales", "sales"),
+        ("postgresql://u@h/db?options=-csearch_path=sales", "sales"),
+        ("postgresql://u@h/db?options=-c%20search_path=sales", "sales"),
+        ("postgresql://u@h/db", None),
+    ],
+)
+def test_from_dsn_reads_schema_from_query_or_libpq_options(dsn, expected):
+    assert PostgresConfig.from_dsn(dsn).schema_name == expected
+
+
+def test_from_dsn_overrides_win_over_parsed_values():
+    config = PostgresConfig.from_dsn(
+        "postgresql://u@h/db", database="other", schema_name="analytics"
+    )
+    assert config.database == "other"
+    assert config.schema_name == "analytics"
+
+
+def test_from_dsn_rejects_dsn_without_host():
+    with pytest.raises(ValueError, match="no host"):
+        PostgresConfig.from_dsn("not-a-dsn")
+
+
+def test_from_dsn_tolerates_missing_database():
+    """A DSN may legitimately omit the database; the error should surface later,
+    at connection-string construction, not at parse time."""
+    config = PostgresConfig.from_dsn("postgresql://u@h")
+    assert config.database is None
+    with pytest.raises(ValueError, match="database name is required"):
+        config.to_sqlalchemy_connection_string()
diff --git a/test/hq/test_sampler.py b/test/hq/test_sampler.py
new file mode 100644
index 00000000..ec789477
--- /dev/null
+++ b/test/hq/test_sampler.py
@@ -0,0 +1,312 @@
+"""Resource sampling and profiling.
+
+Covers the substrate shared by algorithmic and agentic schema inference: samples
+stay pure JSON, connector provenance survives, and nesting is described by path
+rather than flattened away at the boundary.
+"""
+
+from datetime import datetime
+from decimal import Decimal
+from pathlib import Path
+
+import pytest
+
+from graflo.architecture.onto_sample import (
+    ForeignKeyHint,
+    ResourceSample,
+    SourceSample,
+    infer_field_type,
+    iter_paths,
+    profile_sample,
+    profile_source,
+)
+from graflo.architecture.schema.vertex import FieldType
+from graflo.hq.sampler import ResourceSampler, _jsonable
+
+SOURCE_DIR = Path(__file__).parent.parent / "data" / "sample-source"
+
+
+@pytest.fixture
+def sampler() -> ResourceSampler:
+    return ResourceSampler(max_docs=10)
+
+
+@pytest.fixture
+def file_sample(sampler: ResourceSampler) -> SourceSample:
+    return sampler.sample_files(SOURCE_DIR)
+
+
+# ----------------------------------------------------------------------
+# Sampling
+# ----------------------------------------------------------------------
+
+
+def test_sample_files_records_connector_provenance(file_sample):
+    """Every sample must know which connector produced it -- that relation is
+    what later becomes a resource_connector binding."""
+    assert file_sample.source_name == "sample-source"
+    for sample in file_sample.samples:
+        assert sample.connector, f"{sample.resource_name} lost its connector"
+        assert sample.connector == sample.resource_name
+
+
+def test_sample_files_skips_non_data_files(file_sample):
+    """A source directory routinely holds notes beside its data."""
+    names = {sample.resource_name for sample in file_sample.samples}
+    assert names == {"customers", "orders", "api_orders"}
+
+
+def test_sample_files_keeps_documents_verbatim(file_sample):
+    """Tabular rows arrive as flat dicts, nested documents keep their structure."""
+    customers = file_sample.get("customers")
+    assert customers is not None
+    assert customers.docs[0]["email"] == "alice@example.com"
+
+    api = file_sample.get("api_orders")
+    assert api is not None
+    assert api.docs[0]["customer"] == {"id": "c1", "city": "Berlin"}
+    assert api.docs[0]["items"][0]["sku"] == "A-1"
+    assert api.docs[0]["tags"] == ["priority", "gift"]
+
+
+def test_samples_by_resource_matches_inferencer_input_shape(file_sample):
+    """dict[str, list[dict]] is what cross-resource identity inference consumes."""
+    by_resource = file_sample.samples_by_resource
+    assert set(by_resource) == {"customers", "orders", "api_orders"}
+    assert all(
+        isinstance(docs, list) and all(isinstance(doc, dict) for doc in docs)
+        for docs in by_resource.values()
+    )
+
+
+def test_max_docs_caps_and_marks_truncated():
+    sampler = ResourceSampler(max_docs=2)
+    sample = sampler.sample_file(SOURCE_DIR / "orders.csv")
+    assert len(sample.docs) == 2
+    assert sample.truncated is True
+
+
+def test_max_docs_below_one_is_rejected():
+    with pytest.raises(ValueError, match="at least 1"):
+        ResourceSampler(max_docs=0)
+
+
+def test_sample_single_file_directly(sampler):
+    sample = sampler.sample_file(SOURCE_DIR / "customers.csv")
+    assert sample.resource_name == "customers"
+    assert len(sample.docs) == 3
+    assert sample.truncated is False
+
+
+def test_sample_files_rejects_directory_without_data(tmp_path, sampler):
+    (tmp_path / "readme.txt").write_text("nothing here")
+    with pytest.raises(ValueError, match="No sampleable files"):
+        sampler.sample_files(tmp_path)
+
+
+def test_source_sample_requires_at_least_one_resource():
+    with pytest.raises(ValueError):
+        SourceSample(source_name="empty", samples=[])
+
+
+# ----------------------------------------------------------------------
+# JSON coercion
+# ----------------------------------------------------------------------
+
+
+@pytest.mark.parametrize(
+    "value,expected",
+    [
+        (datetime(2026, 7, 30, 12, 0), "2026-07-30T12:00:00"),
+        (Decimal("9.50"), 9.5),
+        (None, None),
+        (True, True),
+        (7, 7),
+        ("plain", "plain"),
+    ],
+)
+def test_jsonable_coerces_database_types(value, expected):
+    """DB and columnar readers hand back types json.dumps cannot serialise."""
+    coerced, _ = _jsonable(value, max_cell_chars=64)
+    assert coerced == expected
+
+
+def test_jsonable_clips_long_values_and_reports_it():
+    coerced, clipped = _jsonable("x" * 100, max_cell_chars=10)
+    assert coerced == "x" * 10
+    assert clipped is True
+
+
+def test_jsonable_recurses_into_containers():
+    coerced, _ = _jsonable(
+        {"when": datetime(2026, 1, 1), "amounts": [Decimal("1.5")]}, max_cell_chars=64
+    )
+    assert coerced == {"when": "2026-01-01T00:00:00", "amounts": [1.5]}
+
+
+def test_jsonable_summarises_binary():
+    coerced, clipped = _jsonable(b"\x00\x01\x02", max_cell_chars=64)
+    assert coerced == "<3 bytes>"
+    assert clipped is True
+
+
+# ----------------------------------------------------------------------
+# Path walking and type inference
+# ----------------------------------------------------------------------
+
+
+def test_iter_paths_describes_nesting_by_path():
+    doc = {
+        "id": 1,
+        "customer": {"id": "c1", "address": {"city": "Berlin"}},
+        "items": [{"sku": "A"}, {"sku": "B"}],
+        "tags": ["x"],
+    }
+    paths = {path for path, _depth, _value in iter_paths(doc)}
+    assert paths == {
+        "id",
+        "customer.id",
+        "customer.address.city",
+        "items[].sku",
+        "tags",
+    }
+
+
+def test_iter_paths_reports_depth():
+    doc = {"a": 1, "b": {"c": {"d": 2}}}
+    depths = {path: depth for path, depth, _value in iter_paths(doc)}
+    assert depths == {"a": 0, "b.c.d": 2}
+
+
+@pytest.mark.parametrize(
+    "values,expected",
+    [
+        ([True, False], FieldType.BOOL),
+        ([1, 2, 3], FieldType.INT),
+        ([1.5, 2], FieldType.FLOAT),
+        (["a", "b"], FieldType.STRING),
+        ([datetime(2026, 1, 1)], FieldType.DATETIME),
+        (["2026-07-30", "2026-07-31"], FieldType.DATETIME),
+        (["4f8c1d2e-1111-4222-8333-abcdefabcdef"], FieldType.UUID),
+        ([None, None], FieldType.STRING),
+    ],
+)
+def test_infer_field_type(values, expected):
+    field_type, _item_type = infer_field_type(values)
+    assert field_type == expected
+
+
+def test_infer_field_type_checks_bool_before_int():
+    """bool is an int subclass, so the naive order mistypes booleans as INT."""
+    assert infer_field_type([True, False])[0] == FieldType.BOOL
+    assert infer_field_type([0, 1])[0] == FieldType.INT
+
+
+def test_infer_field_type_lists_carry_item_type():
+    field_type, item_type = infer_field_type([["a", "b"], ["c"]])
+    assert field_type == FieldType.LIST
+    assert item_type == FieldType.STRING
+
+
+# ----------------------------------------------------------------------
+# Profiling
+# ----------------------------------------------------------------------
+
+
+def test_profile_sample_types_tabular_columns(sampler):
+    profile = profile_sample(sampler.sample_file(SOURCE_DIR / "orders.csv"))
+    assert profile.max_depth == 0
+    assert profile.nested is False
+    types = {field.path: field.type for field in profile.fields}
+    assert types["id"] == FieldType.STRING
+    assert types["customer_id"] == FieldType.STRING
+
+
+def test_profile_sample_describes_nesting(sampler):
+    """Nesting depth is the signal an ingestion model needs descend steps."""
+    profile = profile_sample(sampler.sample_file(SOURCE_DIR / "api_orders.json"))
+    assert profile.nested is True
+    assert profile.max_depth == 1
+    assert "items[].sku" in profile.field_paths
+    assert "customer.city" in profile.field_paths
+
+
+def test_profile_sample_tracks_nulls_and_cardinality(sampler):
+    profile = profile_sample(sampler.sample_file(SOURCE_DIR / "api_orders.json"))
+    fields = {field.path: field for field in profile.fields}
+    assert fields["customer.city"].null_count == 1
+    assert fields["order_id"].unique is True
+    assert fields["items[].sku"].present == 3
+
+
+def test_profile_sample_types_scalar_lists(sampler):
+    profile = profile_sample(sampler.sample_file(SOURCE_DIR / "api_orders.json"))
+    tags = next(field for field in profile.fields if field.path == "tags")
+    assert tags.type == FieldType.LIST
+    assert tags.item_type == FieldType.STRING
+
+
+def test_profile_carries_declared_keys_not_guesses():
+    """Declared PK/FK are ground truth for edge inference."""
+    sample = ResourceSample(
+        resource_name="orders",
+        connector="orders",
+        docs=[{"id": "o1", "customer_id": "c1"}],
+        primary_key=["id"],
+        foreign_keys=[
+            ForeignKeyHint(
+                field="customer_id",
+                references_resource="customers",
+                references_field="id",
+            )
+        ],
+    )
+    profile = profile_sample(sample)
+    assert profile.primary_key == ["id"]
+    assert profile.foreign_keys[0].references_resource == "customers"
+
+
+def test_profile_max_paths_caps_and_marks_truncated():
+    sample = ResourceSample(
+        resource_name="wide",
+        docs=[{f"col_{i}": i for i in range(50)}],
+    )
+    profile = profile_sample(sample, max_paths=10)
+    assert len(profile.fields) == 10
+    assert profile.truncated is True
+
+
+def test_flat_docs_makes_nested_sources_eligible_for_identity_inference(sampler):
+    """Identity inference operates on flat records; this is the bridge."""
+    sample = sampler.sample_file(SOURCE_DIR / "api_orders.json")
+    profile = profile_sample(sample)
+    flat = profile.flat_docs(sample.docs)
+    assert flat[0]["customer.id"] == "c1"
+    assert flat[0]["items[].sku"] == "A-1"
+    assert all(not isinstance(value, dict) for value in flat[0].values())
+
+
+def test_profile_source_covers_every_resource(file_sample):
+    profiles = profile_source(file_sample)
+    assert {profile.resource_name for profile in profiles} == {
+        "customers",
+        "orders",
+        "api_orders",
+    }
+    assert all(profile.connector for profile in profiles)
+
+
+# ----------------------------------------------------------------------
+# Serialization
+# ----------------------------------------------------------------------
+
+
+def test_source_sample_round_trips_through_json(file_sample):
+    """The sample crosses an HTTP boundary, so it must survive model_dump/validate
+    with nested documents intact."""
+    payload = file_sample.model_dump(mode="json")
+    restored = SourceSample.model_validate(payload)
+    assert restored.samples_by_resource == file_sample.samples_by_resource
+    api = restored.get("api_orders")
+    assert api is not None
+    assert api.docs[0]["items"][0]["sku"] == "A-1"
diff --git a/test/rdf/test_manifest_rdf.py b/test/rdf/test_manifest_rdf.py
index b04b9b88..a32280ba 100644
--- a/test/rdf/test_manifest_rdf.py
+++ b/test/rdf/test_manifest_rdf.py
@@ -4,6 +4,7 @@
 
 import pathlib
 
+import pytest
 import yaml
 from rdflib import Graph, URIRef
 from rdflib.namespace import OWL, RDF
@@ -31,6 +32,14 @@ def _canonical(manifest: GraphManifest) -> dict:
     return manifest.to_minimal_canonical_dict()
 
 
+def _round_trip(manifest: GraphManifest) -> GraphManifest:
+    serializer = ManifestRdfSerializer(include_ontology=False)
+    deserializer = ManifestRdfDeserializer()
+    return deserializer.from_turtle(
+        serializer.to_turtle(manifest, BASE_URI), BASE_URI.rstrip("/")
+    )
+
+
 def test_ontology_file_exists_and_loads() -> None:
     path = ontology_path()
     assert path.is_file()
@@ -145,6 +154,68 @@ def test_context_has_new_vertex_config_and_label_terms() -> None:
     assert '"prefLabel": "skos:prefLabel"' in payload
 
 
+def test_context_has_identity_mode_terms() -> None:
+    context_path = pathlib.Path(ontology_path()).parent / "graflo-context.jsonld"
+    payload = context_path.read_text(encoding="utf-8")
+    assert '"assigned"' in payload
+    assert '"hasHashIdentity"' in payload
+    assert '"hasSecondaryIdentity"' in payload
+    assert '"SecondaryIdentity": "gf:SecondaryIdentity"' in payload
+
+
+def test_round_trip_example_16_secondary_identities() -> None:
+    original = _load_example_manifest("16-secondary-identities")
+    restored = _round_trip(original)
+
+    assert _canonical(restored) == _canonical(original)
+
+    vertices = restored.graph_schema.core_schema.vertex_config.vertices
+    by_isin = vertices[0].secondary_identities[0]
+    assert by_isin.name == "by_isin"
+    assert by_isin.fields == ["isin"]
+
+
+@pytest.mark.parametrize(
+    "mutate, expected_mode",
+    [
+        (lambda v: None, "natural"),
+        (lambda v: setattr(v, "hash_identity_properties", ["isin", "name"]), "hash"),
+        (lambda v: setattr(v, "blank", True), "blank"),
+        (
+            lambda v: (setattr(v, "assigned", True), setattr(v, "identity", [])),
+            "assigned",
+        ),
+    ],
+    ids=["natural", "hash", "blank", "assigned"],
+)
+def test_round_trip_preserves_vertex_identity_modes(mutate, expected_mode) -> None:
+    """All four identity modes survive RDF. Previously everything read back as natural."""
+    original = _load_example_manifest("16-secondary-identities")
+    vertex = original.graph_schema.core_schema.vertex_config.vertices[0]
+    # blank and assigned vertices reject secondary identities.
+    vertex.secondary_identities = []
+    mutate(vertex)
+    assert vertex.identity_mode == expected_mode
+
+    restored = _round_trip(original)
+
+    restored_vertex = restored.graph_schema.core_schema.vertex_config.vertices[0]
+    assert restored_vertex.identity_mode == expected_mode
+    assert restored_vertex.hash_identity_properties == vertex.hash_identity_properties
+
+
+def test_round_trip_preserves_identity_field_order() -> None:
+    """Identity lists are positional; RDF triples are not, hence gf:artifactIndex."""
+    original = _load_example_manifest("16-secondary-identities")
+    vertex = original.graph_schema.core_schema.vertex_config.vertices[0]
+    vertex.identity = ["sid", "isin", "name"]
+
+    restored = _round_trip(original)
+
+    restored_vertex = restored.graph_schema.core_schema.vertex_config.vertices[0]
+    assert restored_vertex.identity == ["sid", "isin", "name"]
+
+
 def test_profile_and_transform_actor_semantic_links_are_emitted() -> None:
     manifest = _load_example_manifest("2-ingest-self-references")
     assert manifest.graph_schema is not None
diff --git a/uv.lock b/uv.lock
index 3ddb3afb..683078e3 100644
--- a/uv.lock
+++ b/uv.lock
@@ -73,11 +73,11 @@ dependencies = [
   {name = "s3transfer"}
 ]
 name = "boto3"
-sdist = {url = "https://files.pythonhosted.org/packages/18/95/bd6276870084c9c1a94e17d1dc73b2de275e59bd7a0ad8bfff0a1598cec4/boto3-1.43.58.tar.gz", hash = "sha256:12871fb50c383f1b9aa4ed6dd386ba689062baef730552e79d5a9cd782b53058", size = 112685, upload-time = "2026-07-28T19:35:09.336Z"}
+sdist = {url = "https://files.pythonhosted.org/packages/dc/c8/ef9de1d7413da3adcdb6363258ba6b5cc703593409d8c1957825b20a69d3/boto3-1.43.59.tar.gz", hash = "sha256:4e9b14f89adc1a533c89312e86d8e00455a6f15d398796d92f9191b06e56b401", size = 112653, upload-time = "2026-07-29T19:33:25.703Z"}
 source = {registry = "https://pypi.org/simple"}
-version = "1.43.58"
+version = "1.43.59"
 wheels = [
-  {url = "https://files.pythonhosted.org/packages/1c/85/b0709066efb4ce7b86aba4092c739ad0e458be9d62cf32550fc4c130c93f/boto3-1.43.58-py3-none-any.whl", hash = "sha256:ce1a20cbcfaa1d0b3c8f568e2b6c7fbd34b842ea00e729d49a4b4de522828db9", size = 140026, upload-time = "2026-07-28T19:35:06.993Z"}
+  {url = "https://files.pythonhosted.org/packages/2b/10/c5999e72b020012f2e0ccccf2a15632329edd34cb95b02b1ccfb1712ec08/boto3-1.43.59-py3-none-any.whl", hash = "sha256:58b9635deebf075c1c3d76df78df08eb2979c2a74283194676783a0bff3b4557", size = 140024, upload-time = "2026-07-29T19:33:23.751Z"}
 ]
 
 [[package]]
@@ -87,11 +87,11 @@ dependencies = [
   {name = "urllib3"}
 ]
 name = "botocore"
-sdist = {url = "https://files.pythonhosted.org/packages/5e/64/5cd46a7e72b0647e6d78fc8da016259ad66b9ae0818f4c5d629c75e7ca49/botocore-1.43.58.tar.gz", hash = "sha256:e110ca53f65c128fe98df4d6d36a459b1db17c6114671c8a18becf151bf20909", size = 15742412, upload-time = "2026-07-28T19:34:57.468Z"}
+sdist = {url = "https://files.pythonhosted.org/packages/e5/37/3712a70796583570a5a2e426163e13762ba5ec615a73e966ad18e5933954/botocore-1.43.59.tar.gz", hash = "sha256:8016da69ecc1d705249a8ef13548d3c95eec87ac1cd26133a8bdfa73ca175be0", size = 15788291, upload-time = "2026-07-29T19:33:14.871Z"}
 source = {registry = "https://pypi.org/simple"}
-version = "1.43.58"
+version = "1.43.59"
 wheels = [
-  {url = "https://files.pythonhosted.org/packages/3d/82/6f8fbbea47b773734ba0199643d2d851e5c2f75bc3699fe99db8af344d96/botocore-1.43.58-py3-none-any.whl", hash = "sha256:f516159f0732da8249206163ccea3bd1f82ad2a9d184fe6ed447e1abdba4330e", size = 15426503, upload-time = "2026-07-28T19:34:53.508Z"}
+  {url = "https://files.pythonhosted.org/packages/f6/cd/62d749f824b25c152144665f7c5eb8b5ca8be967a87e0c63577b7d4501ae/botocore-1.43.59-py3-none-any.whl", hash = "sha256:21393c35d23b19d7ba95cc4156b59f4013f80696d667997e1abd9d4e29651708", size = 15471171, upload-time = "2026-07-29T19:33:10.864Z"}
 ]
 
 [[package]]
@@ -434,11 +434,11 @@ wheels = [
 
 [[package]]
 name = "filelock"
-sdist = {url = "https://files.pythonhosted.org/packages/c0/80/8232b582c4b318b817cf1274ba74976b07b34d35ef439b3eb948f98645a1/filelock-3.32.0.tar.gz", hash = "sha256:7be2ad23a14607ccc71808e68fe30848aeace7058ace17852f68e2a68e310402", size = 213757, upload-time = "2026-07-21T13:17:42.898Z"}
+sdist = {url = "https://files.pythonhosted.org/packages/f6/57/3ba6e6cb097f85b855b00163d169f35365f44277df044dcf96d55b8f62a3/filelock-3.32.2.tar.gz", hash = "sha256:c33351e1f49cae33414acbc6d56784e6ecee82514ec90795da1161fc4836b5b8", size = 217172, upload-time = "2026-07-29T22:46:04.895Z"}
 source = {registry = "https://pypi.org/simple"}
-version = "3.32.0"
+version = "3.32.2"
 wheels = [
-  {url = "https://files.pythonhosted.org/packages/06/79/b4c714bef36bc4ec2beeae1e0c124f0223888cd8c6feb1cdc56038116920/filelock-3.32.0-py3-none-any.whl", hash = "sha256:d396bea984af47333ef05e50eae7eff88c84256de6112aea0ec48a233c064fe3", size = 97732, upload-time = "2026-07-21T13:17:41.55Z"}
+  {url = "https://files.pythonhosted.org/packages/c1/e8/72f8cef9fdfeffe06213fe8508039396ee48daa0e3259457ed766173bfd6/filelock-3.32.2-py3-none-any.whl", hash = "sha256:87dd94cf281e586d135fa51132b8e3d9a598b316e90377a288663c9321036c82", size = 98830, upload-time = "2026-07-29T22:46:03.52Z"}
 ]
 
 [[package]]
@@ -491,7 +491,7 @@ dependencies = [
 ]
 name = "graflo"
 source = {editable = "."}
-version = "1.10.0"
+version = "1.10.1"
 
 [package.metadata]
 provides-extras = ["dev", "docs", "plot"]
@@ -526,6 +526,7 @@ requires-dist = [
   {name = "rdflib", specifier = ">=7.0.0"},
   {name = "redis", specifier = ">=5.0.0"},
   {name = "requests", specifier = ">=2.31.0"},
+  {name = "ruff", marker = "extra == 'docs'"},
   {name = "sparqlwrapper", specifier = ">=2.0.0"},
   {name = "sqlalchemy", specifier = ">=2.0.0"},
   {name = "suthing", specifier = ">=0.5.1"},
@@ -548,7 +549,8 @@ docs = [
   {name = "mkdocs-literate-nav"},
   {name = "mkdocs-material"},
   {name = "mkdocs-table-reader-plugin"},
-  {name = "mkdocstrings", extra = ["python"]}
+  {name = "mkdocstrings", extra = ["python"]},
+  {name = "ruff"}
 ]
 plot = [
   {name = "pygraphviz"}
@@ -2094,6 +2096,31 @@ wheels = [
   {url = "https://files.pythonhosted.org/packages/82/3b/64d4899d73f91ba49a8c18a8ff3f0ea8f1c1d75481760df8c68ef5235bf5/rich-15.0.0-py3-none-any.whl", hash = "sha256:33bd4ef74232fb73fe9279a257718407f169c09b78a87ad3d296f548e27de0bb", size = 310654, upload-time = "2026-04-12T08:24:02.83Z"}
 ]
 
+[[package]]
+name = "ruff"
+sdist = {url = "https://files.pythonhosted.org/packages/4d/94/1e5e4967626faf12fa56999cd6222dff6992ceb086ad7945756baf70c7a7/ruff-0.16.0.tar.gz", hash = "sha256:e460aafd5495ec89efaa6ced2e4a9a581116451e1c88b9d37ef497e0f8e93982", size = 4790557, upload-time = "2026-07-23T19:11:30.981Z"}
+source = {registry = "https://pypi.org/simple"}
+version = "0.16.0"
+wheels = [
+  {url = "https://files.pythonhosted.org/packages/4b/81/1c8818fee7ce1a04cd7d1b3172e0a8f8e4f1dc4feb7fc390e16daa8af323/ruff-0.16.0-py3-none-linux_armv6l.whl", hash = "sha256:e5115729eb08c585e5121978ba5d5b60caeae394ce21b9fb5e6cd33a1c6c9b1e", size = 10754633, upload-time = "2026-07-23T19:10:46.415Z"},
+  {url = "https://files.pythonhosted.org/packages/23/df/beaf59c09d68db84304d555f188b276a77132a5d5b0b67a5c762aa143628/ruff-0.16.0-py3-none-macosx_10_12_x86_64.whl", hash = "sha256:3c954b1d580bfa035b41654f7858cc7e71d5fc3ac5b723dd62bd9133830ed522", size = 10969164, upload-time = "2026-07-23T19:10:50.271Z"},
+  {url = "https://files.pythonhosted.org/packages/42/ce/741cd197496a1abbf51352710fd15ed995d2a2be87189c1da26a450d6e83/ruff-0.16.0-py3-none-macosx_11_0_arm64.whl", hash = "sha256:e01c21d10eb1b29f47b7454e1f4056db9a3f0260c646aa88457c610291db9f81", size = 10488846, upload-time = "2026-07-23T19:10:52.639Z"},
+  {url = "https://files.pythonhosted.org/packages/52/2a/a2db8e88cade358f5cdcb05674a917751074109315d014eb6352d9a893f7/ruff-0.16.0-py3-none-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:6e364e5ed22ed8dc05082fd78e35308618260907ac2d3c1d637b2e682415b6c9", size = 10889729, upload-time = "2026-07-23T19:10:54.89Z"},
+  {url = "https://files.pythonhosted.org/packages/42/65/62a771694ebd63029dc953e27dbad40e1588bd4860ff9fe881018fddaa49/ruff-0.16.0-py3-none-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:d327b8fc113a1d4421a04f3839d3752057c8dd1ee320223a6f3f52d04ada462a", size = 10568275, upload-time = "2026-07-23T19:10:56.993Z"},
+  {url = "https://files.pythonhosted.org/packages/3f/e2/ced249fe8af5f086c5c58cc21cc3356d50f32f7401c5df87050c999620a7/ruff-0.16.0-py3-none-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:a9b50c55e263103586b3dcf5f73d479eb8cb5fdb6098fec59a62891dab653717", size = 11385112, upload-time = "2026-07-23T19:10:59.615Z"},
+  {url = "https://files.pythonhosted.org/packages/87/0b/05154977a8fd69eeb6c103271f55403bfd8711f5c0f8ed07489d95a504e7/ruff-0.16.0-py3-none-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:0ff4a79ce3ec0172f3241943835de1c4cb4e2dcd07f0f8c2d02603dbbbee4b17", size = 12207008, upload-time = "2026-07-23T19:11:02.154Z"},
+  {url = "https://files.pythonhosted.org/packages/fb/29/98225831a3a1eab0e02f4acc6ca6559a98611dcc68b6965ff4b7234627c1/ruff-0.16.0-py3-none-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:e95c448fca1fb2a18372a9440926c5a6ee789639bb975c72e7ae6d0b04218ab4", size = 11650842, upload-time = "2026-07-23T19:11:04.557Z"},
+  {url = "https://files.pythonhosted.org/packages/91/66/6bd3cf90500653d55dc0ffc8507aa8300bd49d0214b2e8cb4d3fef2943ba/ruff-0.16.0-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:4f11a8d11010301d0a398a2fdef67691feca7294da6aef55e2150e8fa2cd520b", size = 11400718, upload-time = "2026-07-23T19:11:09.233Z"},
+  {url = "https://files.pythonhosted.org/packages/8e/a2/a54eb4eae05d66364050a5d3b8a9c5ef88196531b3cbe7109d873f87f819/ruff-0.16.0-py3-none-manylinux_2_31_riscv64.whl", hash = "sha256:48044c678e9cb8698246c99b14aaccfa6601dea7379eb48a6f8f73f7a6d86cd0", size = 11426177, upload-time = "2026-07-23T19:11:11.994Z"},
+  {url = "https://files.pythonhosted.org/packages/1a/be/16e3eea4b2a478a496919f5e36f17c4559e54620bd3bbac5d6affa068006/ruff-0.16.0-py3-none-musllinux_1_2_aarch64.whl", hash = "sha256:7aa0959bad8eb8bef50340154fc9b58678dae31fa4293afa38b44b6e552c0213", size = 10856126, upload-time = "2026-07-23T19:11:14.221Z"},
+  {url = "https://files.pythonhosted.org/packages/a2/84/252eb8b868a16eec7257c14f504f77537e734b2d69c762e639e588e304a3/ruff-0.16.0-py3-none-musllinux_1_2_armv7l.whl", hash = "sha256:28ea2b7df8ebf7f9da6b7d47b230ab48f387c0a29be3b474c4d0740e197bb9af", size = 10571208, upload-time = "2026-07-23T19:11:16.378Z"},
+  {url = "https://files.pythonhosted.org/packages/21/09/817a482f542f7570cbb4554b26e896610c7114f539b1d9e2d2145bf6bef6/ruff-0.16.0-py3-none-musllinux_1_2_i686.whl", hash = "sha256:33a3dfac8c35f81498dea9181bccc2f4c4bc8f1521a1dd9406e77643e0f0fb09", size = 11063329, upload-time = "2026-07-23T19:11:19.173Z"},
+  {url = "https://files.pythonhosted.org/packages/2e/23/9403c180ca1cb9b1f7335f5c3e5305c09d49ea5b345196682a36028bde4a/ruff-0.16.0-py3-none-musllinux_1_2_x86_64.whl", hash = "sha256:a5237a0bda500d30d81b8e07a6973a5cbc772864cbf746ae2f4e8a2e01c9f4ed", size = 11489751, upload-time = "2026-07-23T19:11:21.74Z"},
+  {url = "https://files.pythonhosted.org/packages/b2/1d/1b2ef7bcde851c78d7f17f1cca13fd6dc695fc4b3d6197941e72cae5b132/ruff-0.16.0-py3-none-win32.whl", hash = "sha256:7fab76fa065c873f41ff744347c6e77bcc3dfec4bcc754dc26b63d23c0f7f5fb", size = 10785885, upload-time = "2026-07-23T19:11:23.947Z"},
+  {url = "https://files.pythonhosted.org/packages/b2/a3/d5e4ef7a56be3f928ffb90b94c25ba7d3cb9c7fe0736aeaaedf361770712/ruff-0.16.0-py3-none-win_amd64.whl", hash = "sha256:429c117f022bf481fabd9d551e7a3952b24c65e6ef44337ea09d90bebef14472", size = 11923141, upload-time = "2026-07-23T19:11:26.409Z"},
+  {url = "https://files.pythonhosted.org/packages/cb/9a/8415f2657cbe200f41a4531ccededf135505a92d4a012229121f885b26f9/ruff-0.16.0-py3-none-win_arm64.whl", hash = "sha256:14296fedcd2705c77ab8235439278bbb38f285cf7da5528b00b3e330c3d4872d", size = 11273407, upload-time = "2026-07-23T19:11:28.705Z"}
+]
+
 [[package]]
 dependencies = [
   {name = "botocore"}
@@ -2265,27 +2292,27 @@ wheels = [
 
 [[package]]
 name = "ty"
-sdist = {url = "https://files.pythonhosted.org/packages/8e/aa/14c9965d3b173105692473897cc89c34cd91241368b2044e43167e1c17ff/ty-0.0.64.tar.gz", hash = "sha256:d12ddbb05f15158bb518af619378b385486450def95fb06f8ab98037febe9f2c", size = 6350966, upload-time = "2026-07-27T18:32:45.403Z"}
-source = {registry = "https://pypi.org/simple"}
-version = "0.0.64"
-wheels = [
-  {url = "https://files.pythonhosted.org/packages/ae/4c/c54937e4ff3fa7b34a99ea3387ec766bf0ad98dc8df8d792e89b388e658e/ty-0.0.64-py3-none-linux_armv6l.whl", hash = "sha256:3830a6675ab43635ced1c4c557f380ac4a49e9414e03975e4e4e8db644c64944", size = 12118357, upload-time = "2026-07-27T18:32:08.702Z"},
-  {url = "https://files.pythonhosted.org/packages/ce/aa/a839ee2bc78e943d079e6abe199a97b4eeffb7e5c9a57326d69de452186a/ty-0.0.64-py3-none-macosx_10_12_x86_64.whl", hash = "sha256:3ff07d7bc32a2135f58afe57393789a32ea2fed1a66216129a8e535e76043903", size = 11790882, upload-time = "2026-07-27T18:32:10.997Z"},
-  {url = "https://files.pythonhosted.org/packages/4c/de/19f14357888a7198438926303753cf749428e3d62e8980ff1e9a72a78402/ty-0.0.64-py3-none-macosx_11_0_arm64.whl", hash = "sha256:4f6d1c7f897cca05d12bacbf1435150d5ffa496099515aa6ed303c8b29e1d0bb", size = 11317394, upload-time = "2026-07-27T18:32:13.162Z"},
-  {url = "https://files.pythonhosted.org/packages/08/2f/f54462300535ab99b551eda733177be2eef5dbc2997d3fdb357c4ddd760a/ty-0.0.64-py3-none-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:138d6c37ad4bf8583aa7a9b29d90954151d7856f9910a03eae3eb34b34c57215", size = 11863042, upload-time = "2026-07-27T18:32:15.307Z"},
-  {url = "https://files.pythonhosted.org/packages/5e/95/dbecf745520ebe8bd7b02fc55eee6441c9be312ebf6addce605eb52740dd/ty-0.0.64-py3-none-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:1ed9719d1b7b66fb8efe073d860208a44c40af1f6cd5c2364aa9b323a1e579b4", size = 11910730, upload-time = "2026-07-27T18:32:17.467Z"},
-  {url = "https://files.pythonhosted.org/packages/3b/26/12cfd40028e51ceed7b3cb645281c61c02eb64ff9fb0c09231d65c30ff25/ty-0.0.64-py3-none-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:d68b23e5169e2137b5f1de7169ab0cebecab6c8eda374c34c1f6394308f58242", size = 12631936, upload-time = "2026-07-27T18:32:19.533Z"},
-  {url = "https://files.pythonhosted.org/packages/a0/d0/65ffc2b0a686347193c6f98e9421a7fc2a96fc3cd0b1001cf7cf284baff9/ty-0.0.64-py3-none-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:f41cb07d89d32626fcaf3ed4d262778fcb28b2628b6ed1e172cd7b18820668d8", size = 13171049, upload-time = "2026-07-27T18:32:22.026Z"},
-  {url = "https://files.pythonhosted.org/packages/f6/a4/975a5961842dcd6fa60f0770a102c0bba7509da909ab652919bfcdcd4fc7/ty-0.0.64-py3-none-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:5f24e1504ab9e212f92356b82fe088fdeb3a39f9a2f4ff25e505d2e1d0db9056", size = 12826438, upload-time = "2026-07-27T18:32:24.178Z"},
-  {url = "https://files.pythonhosted.org/packages/af/ef/dfb9b7f9bcc032d3b540b0d1f55f532a336e2fb41b1bd539c05ae81a151a/ty-0.0.64-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:86db830cb914bb33bb8247b66ccea58de4496c2391bd42658aba744b439f3290", size = 12440880, upload-time = "2026-07-27T18:32:26.341Z"},
-  {url = "https://files.pythonhosted.org/packages/ee/d0/bedac20505e8a8f5501ad73d7d15d8e421a563fef59993909a23036929a4/ty-0.0.64-py3-none-manylinux_2_31_riscv64.whl", hash = "sha256:652ee3d6d03bea76cd2fe8949c78bb5970394ebd7c5fd270e9518c0dee1b931d", size = 12782439, upload-time = "2026-07-27T18:32:28.642Z"},
-  {url = "https://files.pythonhosted.org/packages/79/d5/795733f13ceff1378f08b3de0c49d0f518df220ed856b0dfac869f3b7c81/ty-0.0.64-py3-none-musllinux_1_2_aarch64.whl", hash = "sha256:4838768295774a86e95f9ec5633e739d016adbbb69dcbdc62f3549578a11f624", size = 11814821, upload-time = "2026-07-27T18:32:30.632Z"},
-  {url = "https://files.pythonhosted.org/packages/52/3e/9d99cd1e1831003434f508ed9f258a56543194afc3bbe051eba2545fa676/ty-0.0.64-py3-none-musllinux_1_2_armv7l.whl", hash = "sha256:5a3d700669868599edf39ce5125196682f15a8880cc8c669bc2a83b99984d99b", size = 11928678, upload-time = "2026-07-27T18:32:32.888Z"},
-  {url = "https://files.pythonhosted.org/packages/b8/3d/448f49a3503fb119a34348a5714bb92001f252fadbbb12d645f2e744b557/ty-0.0.64-py3-none-musllinux_1_2_i686.whl", hash = "sha256:b161f0a82a8e2f2432db3bf7702b4d3924fa9486ba0014f6710a160fc157df0d", size = 12202249, upload-time = "2026-07-27T18:32:34.905Z"},
-  {url = "https://files.pythonhosted.org/packages/dd/9b/75768e562cec990d189dc05807ae72890b20ffbd1e1f43597bb98db63c60/ty-0.0.64-py3-none-musllinux_1_2_x86_64.whl", hash = "sha256:39b9dd42908df47c2dc57dda87e656fab97097ffd2618474bbdea986af0d6a9d", size = 12548817, upload-time = "2026-07-27T18:32:36.995Z"},
-  {url = "https://files.pythonhosted.org/packages/34/89/44cc276ea6ca0245495014758ffeadde1798fb0b5840c9cf36d8d2ed3250/ty-0.0.64-py3-none-win32.whl", hash = "sha256:d0676ab0e0935795e5843baa28dd5e366dc343d7c0945a996df9eab8e0644885", size = 11545474, upload-time = "2026-07-27T18:32:39.389Z"},
-  {url = "https://files.pythonhosted.org/packages/01/7e/d1c8a871a38d17c8f168b9a6975f6247f7660f8334e517656a5e4b4a4858/ty-0.0.64-py3-none-win_amd64.whl", hash = "sha256:dcb9bd31f54097e362b776c26ab4564d4564cdd1355cb883481167a26c03cc3f", size = 12542987, upload-time = "2026-07-27T18:32:41.412Z"},
-  {url = "https://files.pythonhosted.org/packages/35/4d/6d18640d0204cacd69abbaca95ad6a34c6d7e9169e9051d0117b17b827ec/ty-0.0.64-py3-none-win_arm64.whl", hash = "sha256:82cc34c1ad9a8feb6059aef193bebcecc656e548f6fab3d518bcd8b57d198d39", size = 11899263, upload-time = "2026-07-27T18:32:43.366Z"}
+sdist = {url = "https://files.pythonhosted.org/packages/d6/54/cf561927e8e9ab5c1892a833b664aa9cd6f051a75f6280c66d8047246bda/ty-0.0.65.tar.gz", hash = "sha256:b7134bffcc00b715fa8291e84d845782ced810a998dc1f7f11d71c85c4046325", size = 6460098, upload-time = "2026-07-29T18:31:03.27Z"}
+source = {registry = "https://pypi.org/simple"}
+version = "0.0.65"
+wheels = [
+  {url = "https://files.pythonhosted.org/packages/7b/4e/71e2d325d2b53a1afad81624ad076b2ede413213fc4a18cb05b78c568571/ty-0.0.65-py3-none-linux_armv6l.whl", hash = "sha256:dc556c9f05408bef4c4ef02b2cc382e4e5f797b4b20d64410289848f0d76705f", size = 12298466, upload-time = "2026-07-29T18:30:12.744Z"},
+  {url = "https://files.pythonhosted.org/packages/57/77/fec8f29647c55794efa430a7f365e44f5ce7ffb6459d9445a87fac569bec/ty-0.0.65-py3-none-macosx_10_12_x86_64.whl", hash = "sha256:29d2e0d34cc0a28a17ef0cf81135c5ebabc3562131f9079138ba5e7bae0f56bd", size = 11942421, upload-time = "2026-07-29T18:30:16.076Z"},
+  {url = "https://files.pythonhosted.org/packages/13/09/7f3766aef9dc627e2698cf4e3e59cf53389dcae3812040d33c1aa931230f/ty-0.0.65-py3-none-macosx_11_0_arm64.whl", hash = "sha256:685f49a9312bbf69d5b65bbb66384fed1f927403ea030c217b9289092d7e46c4", size = 11451922, upload-time = "2026-07-29T18:30:19.155Z"},
+  {url = "https://files.pythonhosted.org/packages/cb/7b/1a77cd50e0befb50f55b8bf9bd3ed3eddf184bf28c61b56727039e0774fc/ty-0.0.65-py3-none-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:6f564b5ebe78e2f3a8e7b8eacb1292eb88b7c0f3c8630671cfca31abc0709cd9", size = 11994999, upload-time = "2026-07-29T18:30:22.315Z"},
+  {url = "https://files.pythonhosted.org/packages/63/7b/feda16f3a4a0a99be27431e0c9598eeeec0db1eb2fec9a15976698209418/ty-0.0.65-py3-none-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:c983e156fe9e113fb56389e13d327b6b8549fe866de9b269684723a88e9b732d", size = 12090662, upload-time = "2026-07-29T18:30:24.93Z"},
+  {url = "https://files.pythonhosted.org/packages/ed/3e/3f69bf9c9307dbdc0771719f65ce5b556e7bdeeaccbdd599d4f57866d801/ty-0.0.65-py3-none-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:e3663b7396e8b1a9954e20e732de7ccb0192bf4118473069b4945920d6923921", size = 12822094, upload-time = "2026-07-29T18:30:28.012Z"},
+  {url = "https://files.pythonhosted.org/packages/90/38/8fa791b3bb503ee2b46ad81690cd1bdd54519582df6d805cee57fe143e85/ty-0.0.65-py3-none-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:306ed01f29d6e108e98feb233dbbf5878a027603b71bd3743b343977933a9f16", size = 13357833, upload-time = "2026-07-29T18:30:31.122Z"},
+  {url = "https://files.pythonhosted.org/packages/c1/73/4dda396a201e1dd0ed3594a9b48e559cb41c4bc048c6cd4c4d1b39eb4313/ty-0.0.65-py3-none-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:28bcfc8898c94f079a9100e684bcf312b6a64ad3a7d4ebb35a4591546030a2cd", size = 12977303, upload-time = "2026-07-29T18:30:33.944Z"},
+  {url = "https://files.pythonhosted.org/packages/a5/26/c250c2c569adc53a8591716641388397bcb2a442e4a30b952ae81b50c0e0/ty-0.0.65-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:5a75bd0c245c38802a8f488378e74f92feb7dd33db7d63fbdd6fdf82791ba730", size = 12579338, upload-time = "2026-07-29T18:30:37.199Z"},
+  {url = "https://files.pythonhosted.org/packages/d3/94/4a5647d44753ca218fc930d7e4d9bf468d0ed4a0ad4b3d57588bc1bbacf7/ty-0.0.65-py3-none-manylinux_2_31_riscv64.whl", hash = "sha256:9e5e1bdea9662d2b5312b4e99f319f4e6e2ea427511b5fbc546141b79ec53f76", size = 12957731, upload-time = "2026-07-29T18:30:39.937Z"},
+  {url = "https://files.pythonhosted.org/packages/36/b6/1e22fa11a1e0dfb20b1c7f3cbfd8170273aada2a82f9ecd3055275370c44/ty-0.0.65-py3-none-musllinux_1_2_aarch64.whl", hash = "sha256:03a88493d4842889f65280ae241e06b399d57eb3c63571054cad21a4c33b3b69", size = 11938625, upload-time = "2026-07-29T18:30:42.603Z"},
+  {url = "https://files.pythonhosted.org/packages/5c/0a/fe5f22ef62b193201bc5566762e22049762cd485bfafb5095a7050760054/ty-0.0.65-py3-none-musllinux_1_2_armv7l.whl", hash = "sha256:600b8bf6f4940cf7ffb2f43d3716faaf38dcb97cd8617c55771451bc0276408f", size = 12105592, upload-time = "2026-07-29T18:30:45.419Z"},
+  {url = "https://files.pythonhosted.org/packages/76/fd/922b3a6e9d697452cdbb4b7e3f636868add5ec652154518a736e4364f3b7/ty-0.0.65-py3-none-musllinux_1_2_i686.whl", hash = "sha256:0c28007bc79d648c1ddaf1e65885d07baec48eb87240da442f608e4107c1b7d8", size = 12387335, upload-time = "2026-07-29T18:30:48.405Z"},
+  {url = "https://files.pythonhosted.org/packages/77/22/a1a08ebc84c083db2fb55e3b5cd186db0c067692f4921146f601360231e2/ty-0.0.65-py3-none-musllinux_1_2_x86_64.whl", hash = "sha256:c852da96091ad22361e6586b7c7ba98e1334dcd4d8ffb67e47f4fb673de33f77", size = 12682710, upload-time = "2026-07-29T18:30:51.364Z"},
+  {url = "https://files.pythonhosted.org/packages/81/14/eaaa410a25bbdea19722109b5422380a0e211b3afcf3071d15953ddbd5db/ty-0.0.65-py3-none-win32.whl", hash = "sha256:cf529d538f1403b14b0511e6ec3cdb95d3d974adabf24cc76cedc533368c3edc", size = 11692341, upload-time = "2026-07-29T18:30:54.35Z"},
+  {url = "https://files.pythonhosted.org/packages/bc/0f/6d48f206dce9d7e53fe3b5ea0f0ab5800dd9d2365b2b48f736783436c43f/ty-0.0.65-py3-none-win_amd64.whl", hash = "sha256:234a321e33c7cbbfbd67bfa0b01b685dd9c21f1841781a21e5ca1fa0b25f1d5d", size = 12729355, upload-time = "2026-07-29T18:30:57.275Z"},
+  {url = "https://files.pythonhosted.org/packages/96/aa/7446f7725e303cf78e058c893af1f0552b9451895454908706f4c6c3494b/ty-0.0.65-py3-none-win_arm64.whl", hash = "sha256:b9424be1ec56d93ff18609fb1c0a0a2283fe1282cd6d1c7604f97d73b94d61f2", size = 12051375, upload-time = "2026-07-29T18:31:00.579Z"}
 ]
 
 [[package]]