Tell a document which database it came from

Document.source names the configured database, as SearchResult and
Citation already do. A listing spanning databases is unreadable
without it, and `--database NAME list` could not name the one it
opened.

CHANGELOG.md

@@ -5,8 +5,7 @@
 ### Added
 
 - `api_key` on model and embedding-model config, overriding the provider's environment variable. Honored on the `openai` and `ollama` providers, `vllm` embedders and rerankers, the picture-description VLM endpoint, and `doctor`'s endpoint probes; other providers raise.
-- `lancedb.databases`: a name-to-location mapping for searching several databases at once, mutually exclusive with `lancedb.uri`. `client.search(..., sources=[...])` selects which to search, `sources=None` searches all of them, and `SearchResult.source` carries the configured name a result came from. Candidates are fused by the configured reranker over the union, or by reciprocal rank fusion when none is configured. `SearchResult.format_for_agent` names the database, so the model can attribute evidence to one while it answers. `haiku-rag search`, `ask` and `analyze` cover the configured set and label each result and citation with its database; every other command works on one, named with `--database NAME` or `--db PATH`.
+- `lancedb.databases`: a name-to-location mapping for searching several databases at once, mutually exclusive with `lancedb.uri`. `client.search(..., sources=[...])` selects which to search, `sources=None` searches all of them, and `SearchResult.source` carries the configured name a result came from. `Document.source` names it on a document from a listing or a lookup. Candidates are fused by the configured reranker over the union, or by reciprocal rank fusion when none is configured. `SearchResult.format_for_agent` names the database, so the model can attribute evidence to one while it answers. `haiku-rag search`, `ask`, `analyze` and `chat` cover the configured set and label each result and citation with its database; every other command works on one, named with `--database NAME` or `--db PATH`.
-- `lancedb.databases`: a name-to-location mapping for searching several databases at once, mutually exclusive with `lancedb.uri`. `client.search(..., sources=[...])` selects which to search, `sources=None` searches all of them, and `SearchResult.source` carries the configured name a result came from. Candidates are fused by the configured reranker over the union, or by reciprocal rank fusion when none is configured. `SearchResult.format_for_agent` names the database, so the model can attribute evidence to one while it answers. `haiku-rag search`, `ask`, `analyze` and `chat` cover the configured set and label each result and citation with its database; every other command works on one, named with `--database NAME` or `--db PATH`.
 - `client.ask(..., sources=[...])` asks across the selected databases, and `Citation.source` names the one a cited chunk came from. The cite fallback for an id absent from the run's results looks only in the selected databases, so a question scoped to some cannot cite another.
 - `client.analyze(..., sources=[...])` analyzes across the selected databases: the sandbox mounts their documents under one flat `/documents/{id}/` namespace, resolving each id to the database holding it, and in-code `search()` covers the same selection.
 

docs/configuration/storage.md

@@ -211,7 +211,9 @@ results = await client.search("query", sources=["medic"])   # one of them
 Candidates from each database are fused into one ranked list, by the configured
 reranker where there is one and by reciprocal rank fusion otherwise. Each result
 carries `source`, the name of the database it came from, and so does each
-citation.
+citation. A document from `list_documents`, `get_document_by_id`,
+`get_document_by_uri` or `resolve_document` carries it too, whether it came from
+a set or from one database named in `lancedb.databases`.
 
 **Configure a reranker when searching several databases.** Reciprocal rank fusion
 compares ranks, not scores, so every database contributes its own best matches

haiku_rag_slim/haiku/rag/app.py

@@ -850,10 +850,16 @@ class HaikuRAGApp:
             if doc.title
             else ""
         )
+        database_part = (
+            f" [repr.attrib_name]database[/repr.attrib_name]: {doc.source}"
+            if doc.source
+            else ""
+        )
         self.console.print(
             f"[repr.attrib_name]id[/repr.attrib_name]: {doc.id} "
             f"[repr.attrib_name]uri[/repr.attrib_name]: {doc.uri}"
             + title_part
+            + database_part
             + f" [repr.attrib_name]meta[/repr.attrib_name]: {doc.metadata}"
         )
         self.console.print(

haiku_rag_slim/haiku/rag/client/__init__.py

@@ -509,7 +509,7 @@ class HaikuRAG:
             return await self._from_any_covered(
                 lambda owner: owner.get_document_by_id(document_id)
             )
-        return await self.document_repository.get_by_id(document_id)
+        return self._name(await self.document_repository.get_by_id(document_id))
 
     async def get_chunk_by_id(self, chunk_id: str) -> Chunk | None:
         """Get a chunk by its ID.
@@ -565,7 +565,7 @@ class HaikuRAG:
             return await self._from_any_covered(
                 lambda owner: owner.get_document_by_uri(uri)
             )
-        return await self.document_repository.get_by_uri(uri)
+        return self._name(await self.document_repository.get_by_uri(uri))
 
     async def resolve_document(self, id_or_title: str) -> Document | None:
         """Resolve a document by ID, title, or URI (in that order).
@@ -671,9 +671,12 @@ class HaikuRAG:
             ]
             start = offset or 0
             return merged[start:] if limit is None else merged[start : start + limit]
-        return await self.document_repository.list_all(
+        documents = await self.document_repository.list_all(
             limit=limit, offset=offset, filter=filter, include_content=include_content
         )
+        for document in documents:
+            document.source = self._source
+        return documents
 
     async def count_documents(self, filter: str | None = None) -> int:
         """Count documents with optional filtering.
@@ -694,6 +697,15 @@ class HaikuRAG:
             return sum(counts)
         return await self.document_repository.count(filter=filter)
 
+    def _name(self, document: Document | None) -> Document | None:
+        """`document`, told which configured database it came from.
+
+        None where no database is named, as with a single ``lancedb.uri``.
+        """
+        if document is not None:
+            document.source = self._source
+        return document
+
     async def _from_any_covered(
         self, lookup: "Callable[[HaikuRAG], Coroutine[Any, Any, Any]]"
     ) -> Any:

haiku_rag_slim/haiku/rag/store/models/document.py

@@ -13,12 +13,18 @@ if TYPE_CHECKING:
 class Document(BaseModel):
     """
     Represents a document with an ID, content, and metadata.
+
+    ``source`` names the configured database a document came from: the name
+    from ``lancedb.databases``, never a path or URI. It is None where no
+    database is named, as with the single ``lancedb.uri``, and is never
+    persisted.
     """
 
     id: str | None = None
     content: str
     uri: str | None = None
     title: str | None = None
+    source: str | None = None
     metadata: dict = {}
     docling_document: bytes | None = Field(default=None, exclude=True)
     docling_pages: bytes | None = Field(default=None, exclude=True)

tests/test_app.py

@@ -64,6 +64,28 @@ async def test_list_documents_prints_each_document(app, client):
     assert "second" in out(app)
 
 
+async def test_list_documents_names_the_database_of_each(app, client):
+    """A listing spanning databases has to say which one each document is from."""
+    client.list_documents.return_value = [
+        _doc("first", source="arxiv"),
+        _doc("second", source="wiki"),
+    ]
+
+    await app.list_documents()
+
+    printed = out(app)
+    assert "database: arxiv" in printed
+    assert "database: wiki" in printed
+
+
+async def test_list_documents_omits_the_database_when_unnamed(app, client):
+    client.list_documents.return_value = [_doc("only")]
+
+    await app.list_documents()
+
+    assert "database:" not in out(app)
+
+
 async def test_add_document_from_text_reports_the_new_id(app, client):
     client.create_document.return_value = _doc("added body")
 

tests/test_multi_db.py

@@ -279,6 +279,81 @@ class TestLookupByIdentifier:
             assert await rag.get_document_by_uri("test://nowhere") is None
 
 
+class TestDocumentsNameTheirDatabase:
+    """A listing that spans databases is unreadable when the documents do not
+    say which one they came from, the same reason a search result carries one."""
+
+    @pytest.mark.asyncio
+    async def test_a_listing_names_each_document_s_database(self, tmp_path):
+        config = _config(tmp_path, ["alpha", "beta"])
+        await _seed(config, "alpha", ["alpha one", "alpha two"])
+        await _seed(config, "beta", ["beta one"])
+
+        async with HaikuRAG(config=config) as rag:
+            docs = await rag.list_documents()
+
+        assert {d.uri: d.source for d in docs} == {
+            "test://alpha/alpha one": "alpha",
+            "test://alpha/alpha two": "alpha",
+            "test://beta/beta one": "beta",
+        }
+
+    @pytest.mark.asyncio
+    async def test_a_looked_up_document_names_its_database(self, tmp_path):
+        config = _config(tmp_path, ["alpha", "beta"])
+        await _seed(config, "alpha", ["alpha one"])
+        await _seed(config, "beta", ["beta one"])
+
+        async with HaikuRAG(config=config) as rag:
+            beta = (await rag.clients_for(["beta"]))[0]
+            [target] = await beta.document_repository.list_all(limit=1)
+            assert target.id is not None
+
+            by_id = await rag.get_document_by_id(target.id)
+            by_uri = await rag.get_document_by_uri("test://alpha/alpha one")
+            resolved = await rag.resolve_document(target.id)
+
+        assert by_id is not None and by_id.source == "beta"
+        assert by_uri is not None and by_uri.source == "alpha"
+        assert resolved is not None and resolved.source == "beta"
+
+    @pytest.mark.asyncio
+    async def test_one_named_database_still_names_itself(self, tmp_path):
+        """`haiku-rag --database alpha list` opens one database, and its name is
+        the whole reason the option exists."""
+        config = _config(tmp_path, ["alpha", "beta"])
+        await _seed(config, "alpha", ["alpha one"])
+        await _seed(config, "beta", ["beta one"])
+
+        async with HaikuRAG(config=config, sources=["alpha"]) as rag:
+            [listed] = await rag.list_documents()
+            assert listed.id is not None
+            by_id = await rag.get_document_by_id(listed.id)
+            by_uri = await rag.get_document_by_uri("test://alpha/alpha one")
+
+        assert listed.source == "alpha"
+        assert by_id is not None and by_id.source == "alpha"
+        assert by_uri is not None and by_uri.source == "alpha"
+
+    @pytest.mark.asyncio
+    async def test_one_database_leaves_the_source_unset(self, tmp_path, temp_db_path):
+        """Nothing names the database when there is only one to name."""
+        async with HaikuRAG(temp_db_path, create=True) as rag:
+            dim = get_config().embeddings.model.vector_dim
+            doc = DoclingDocument(name="solo")
+            doc.add_text(label=DocItemLabel.TEXT, text="solo")
+            await rag.import_document(
+                doc,
+                [Chunk(content="solo", embedding=[0.1] * dim, order=0)],
+                uri="test://solo",
+            )
+
+            [listed] = await rag.list_documents()
+            assert listed.source is None
+            assert listed.id is not None
+            assert (await rag.get_document_by_id(listed.id)).source is None
+
+
 class TestFederatedSearch:
     @pytest.mark.asyncio
     async def test_results_carry_their_source(self, tmp_path):