src/ix/contracts/__init__.py

@@ -0,0 +1,27 @@
+"""Pydantic v2 data contracts shared by the pipeline, adapters, and store.
+
+Re-exports the public symbols from sibling modules so call sites can write
+``from ix.contracts import RequestIX`` without chasing the submodule layout.
+"""
+
+from __future__ import annotations
+
+from ix.contracts.request import (
+    Context,
+    FileRef,
+    GenAIOptions,
+    OCROptions,
+    Options,
+    ProvenanceOptions,
+    RequestIX,
+)
+
+__all__ = [
+    "Context",
+    "FileRef",
+    "GenAIOptions",
+    "OCROptions",
+    "Options",
+    "ProvenanceOptions",
+    "RequestIX",
+]

src/ix/contracts/request.py

@@ -0,0 +1,103 @@
+"""Incoming request contracts — :class:`RequestIX` + nested option blocks.
+
+Mirrors MVP spec §3 exactly. Dropped spec fields (e.g. ``use_vision``,
+``reasoning_effort``, ``version``) are intentionally absent from this module;
+if a caller sends one ``RequestIX`` rejects the payload (``extra="forbid"``).
+
+The file-URL union keeps plain strings as ``str`` and dict entries as
+:class:`FileRef` so callers can mix auth-required URLs with plain ones in a
+single list without wrapping every entry.
+"""
+
+from __future__ import annotations
+
+from typing import Literal
+
+from pydantic import BaseModel, ConfigDict, Field
+
+
+class FileRef(BaseModel):
+    """A file entry with optional auth headers and per-file size override.
+
+    Used when the file URL needs authentication (e.g. Paperless ``Token``) or a
+    tighter size cap than :envvar:`IX_FILE_MAX_BYTES`. Plain URLs that need no
+    headers can stay as bare ``str`` values in :attr:`Context.files`.
+    """
+
+    model_config = ConfigDict(extra="forbid")
+
+    url: str
+    headers: dict[str, str] = Field(default_factory=dict)
+    max_bytes: int | None = None
+
+
+class Context(BaseModel):
+    """Document payload: files to fetch + OCR + any pre-extracted texts."""
+
+    model_config = ConfigDict(extra="forbid")
+
+    files: list[str | FileRef] = Field(default_factory=list)
+    texts: list[str] = Field(default_factory=list)
+
+
+class OCROptions(BaseModel):
+    """Knobs for the OCR step.
+
+    ``service`` is kept so the adapter point stays visible in the contract even
+    though MVP only wires Surya. Future engines plug in behind the same name.
+    """
+
+    model_config = ConfigDict(extra="forbid")
+
+    use_ocr: bool = True
+    ocr_only: bool = False
+    include_ocr_text: bool = False
+    include_geometries: bool = False
+    service: Literal["surya"] = "surya"
+
+
+class GenAIOptions(BaseModel):
+    """Knobs for the LLM step."""
+
+    model_config = ConfigDict(extra="forbid")
+
+    gen_ai_model_name: str | None = None
+
+
+class ProvenanceOptions(BaseModel):
+    """Knobs for the provenance + reliability steps."""
+
+    model_config = ConfigDict(extra="forbid")
+
+    include_provenance: bool = True
+    max_sources_per_field: int = 10
+
+
+class Options(BaseModel):
+    """Aggregate options block nested into :class:`RequestIX`."""
+
+    model_config = ConfigDict(extra="forbid")
+
+    ocr: OCROptions = Field(default_factory=OCROptions)
+    gen_ai: GenAIOptions = Field(default_factory=GenAIOptions)
+    provenance: ProvenanceOptions = Field(default_factory=ProvenanceOptions)
+
+
+class RequestIX(BaseModel):
+    """Top-level job request.
+
+    ``ix_id`` is a transport-assigned 16-char hex handle. Callers MUST NOT set
+    it; the REST adapter / pg-queue adapter populates it on insert. The field
+    is kept here so the contract is closed-over-construction round-trips
+    (e.g. when the worker re-hydrates a job out of the store).
+    """
+
+    model_config = ConfigDict(extra="forbid")
+
+    use_case: str
+    ix_client_id: str
+    request_id: str
+    ix_id: str | None = None
+    context: Context
+    options: Options = Field(default_factory=Options)
+    callback_url: str | None = None

tests/unit/test_contracts.py

@@ -0,0 +1,168 @@
+"""Pydantic contracts — RequestIX and its nested option structures (spec §3)."""
+
+from __future__ import annotations
+
+import json
+
+import pytest
+from pydantic import ValidationError
+
+from ix.contracts import (
+    Context,
+    FileRef,
+    GenAIOptions,
+    OCROptions,
+    Options,
+    ProvenanceOptions,
+    RequestIX,
+)
+
+
+class TestFileRef:
+    def test_minimal(self) -> None:
+        fr = FileRef(url="https://example.com/x.pdf")
+        assert fr.url == "https://example.com/x.pdf"
+        assert fr.headers == {}
+        assert fr.max_bytes is None
+
+    def test_with_headers_and_max_bytes(self) -> None:
+        fr = FileRef(
+            url="https://paperless/x.pdf",
+            headers={"Authorization": "Token abc"},
+            max_bytes=1_000_000,
+        )
+        assert fr.headers == {"Authorization": "Token abc"}
+        assert fr.max_bytes == 1_000_000
+
+
+class TestOptionDefaults:
+    def test_ocr_defaults_match_spec(self) -> None:
+        o = OCROptions()
+        assert o.use_ocr is True
+        assert o.ocr_only is False
+        assert o.include_ocr_text is False
+        assert o.include_geometries is False
+        assert o.service == "surya"
+
+    def test_genai_defaults_match_spec(self) -> None:
+        g = GenAIOptions()
+        assert g.gen_ai_model_name is None
+
+    def test_provenance_defaults_match_spec(self) -> None:
+        p = ProvenanceOptions()
+        assert p.include_provenance is True
+        assert p.max_sources_per_field == 10
+
+    def test_options_default_nests_each_block(self) -> None:
+        opts = Options()
+        assert isinstance(opts.ocr, OCROptions)
+        assert isinstance(opts.gen_ai, GenAIOptions)
+        assert isinstance(opts.provenance, ProvenanceOptions)
+
+
+class TestContextFiles:
+    def test_plain_string_entry_preserved_as_str(self) -> None:
+        ctx = Context(files=["https://example.com/a.pdf"])
+        assert ctx.files == ["https://example.com/a.pdf"]
+        assert isinstance(ctx.files[0], str)
+
+    def test_dict_entry_parsed_as_fileref(self) -> None:
+        ctx = Context(files=[{"url": "https://x/a.pdf", "headers": {"H": "v"}}])
+        assert len(ctx.files) == 1
+        entry = ctx.files[0]
+        assert isinstance(entry, FileRef)
+        assert entry.url == "https://x/a.pdf"
+        assert entry.headers == {"H": "v"}
+
+    def test_mixed_entries(self) -> None:
+        ctx = Context(
+            files=[
+                "file:///tmp/x.pdf",
+                {"url": "https://paperless/y.pdf", "headers": {"Authorization": "Token t"}},
+            ],
+            texts=["extra ocr text"],
+        )
+        assert isinstance(ctx.files[0], str)
+        assert isinstance(ctx.files[1], FileRef)
+        assert ctx.texts == ["extra ocr text"]
+
+    def test_empty_defaults(self) -> None:
+        ctx = Context()
+        assert ctx.files == []
+        assert ctx.texts == []
+
+
+class TestRequestIX:
+    def _minimal_payload(self) -> dict:
+        return {
+            "use_case": "bank_statement_header",
+            "ix_client_id": "mammon",
+            "request_id": "req-1",
+            "context": {"files": ["https://example/x.pdf"]},
+        }
+
+    def test_minimal_valid(self) -> None:
+        r = RequestIX(**self._minimal_payload())
+        assert r.use_case == "bank_statement_header"
+        assert r.ix_id is None
+        assert r.callback_url is None
+        assert r.options.provenance.include_provenance is True
+
+    def test_roundtrip_json_mixed_files(self) -> None:
+        payload = {
+            "use_case": "bank_statement_header",
+            "ix_client_id": "mammon",
+            "request_id": "req-42",
+            "context": {
+                "files": [
+                    "file:///tmp/x.pdf",
+                    {
+                        "url": "https://paperless/y.pdf",
+                        "headers": {"Authorization": "Token t"},
+                        "max_bytes": 2_000_000,
+                    },
+                ],
+                "texts": ["paperless ocr text"],
+            },
+            "options": {
+                "ocr": {"include_ocr_text": True},
+                "gen_ai": {"gen_ai_model_name": "gpt-oss:20b"},
+                "provenance": {"max_sources_per_field": 5},
+            },
+            "callback_url": "https://mammon/ix-callback",
+        }
+        r = RequestIX.model_validate(payload)
+        assert isinstance(r.context.files[0], str)
+        assert isinstance(r.context.files[1], FileRef)
+        assert r.context.files[1].headers == {"Authorization": "Token t"}
+        assert r.options.ocr.include_ocr_text is True
+        assert r.options.gen_ai.gen_ai_model_name == "gpt-oss:20b"
+        assert r.options.provenance.max_sources_per_field == 5
+        assert r.callback_url == "https://mammon/ix-callback"
+
+        # Round-trip through JSON and back: FileRef dicts survive as FileRef.
+        dumped = r.model_dump_json()
+        r2 = RequestIX.model_validate_json(dumped)
+        assert isinstance(r2.context.files[1], FileRef)
+        assert r2.context.files[1].headers == {"Authorization": "Token t"}
+
+        # dumped JSON is valid JSON
+        json.loads(dumped)
+
+    def test_unknown_fields_rejected(self) -> None:
+        payload = self._minimal_payload()
+        payload["not_a_field"] = "x"
+        with pytest.raises(ValidationError):
+            RequestIX.model_validate(payload)
+
+    def test_ix_id_optional_and_documented(self) -> None:
+        # The docstring contract: caller MUST NOT set; transport assigns.
+        # Here we only assert the field exists and defaults to None — the
+        # "MUST NOT set" is a convention enforced at the transport layer.
+        r = RequestIX(**self._minimal_payload())
+        assert r.ix_id is None
+        assert "transport" in RequestIX.__doc__.lower() or "MUST NOT" in (RequestIX.__doc__ or "")
+
+    def test_missing_required_fields(self) -> None:
+        with pytest.raises(ValidationError):
+            RequestIX.model_validate({"use_case": "x"})