refactor(evidence): migrate acquisition and preprocessing (#30)

harness/tht/corpus/pipeline.py

@@ -15,7 +15,8 @@ from tht.corpus.chunk import ChunkPolicy, chunk
 from tht.corpus.models import CanonicalChunk, CanonicalDocument, CorpusManifest
 from tht.corpus.normalize import normalize
 from tht.corpus.store import CorpusStore
-from tht.ports.evidence import EvidenceSource, SourceObject, canonical_provenance_uri
+import tht.evidence.acquisition as evidence_acquisition
+from tht.evidence.contracts import EvidenceSource, SourceObject, canonical_provenance_uri
 from tht.ports.vector import VectorStore, VectorWriteRecord
 from tht.vectorstore.records import VectorRecord
 from tht.jobs.models import JobSpec
@@ -228,7 +229,7 @@ class CorpusPipeline:
         discovered = []
         seen = set()
         for source in self.sources:
-            for item in source.discover():
+            for item in evidence_acquisition.discover(source):
                 if item.source_id in seen:
                     raise PipelineError("duplicate Evidence source identity")
                 seen.add(item.source_id)
@@ -455,7 +456,9 @@ class CorpusPipeline:
             documents = [prior[source_id] for source_id in plan["unchanged"]]
             for source_id in plan["changed"]:
                 source, item = source_by_id[source_id]
-                documents.append(normalize(source.acquire(item), self.pipeline_version))
+                documents.append(normalize(
+                    evidence_acquisition.acquire(source, item), self.pipeline_version,
+                ))
             documents.sort(key=lambda value: value.source_id)
             chunks = [part for document in documents for part in chunk(document, self.chunk_policy)]
             previous_generations = dict(previous.metadata.get("document_generations", {})) if previous else {}
@@ -710,7 +713,9 @@ class CorpusPipeline:
         try:
             for source, item in discovered:
                 if item.source_id in changed_set:
-                    documents.append(normalize(source.acquire(item), self.pipeline_version))
+                    documents.append(normalize(
+                        evidence_acquisition.acquire(source, item), self.pipeline_version,
+                    ))
             documents.sort(key=lambda document: document.source_id)
             chunks: list[CanonicalChunk] = []
             for document in documents: