fix translate images/pdfs

bpm-ai/bpm_ai/translate/translate.py

@@ -31,8 +31,8 @@ async def translate_llm(
     else:
         input_items = await ocr_documents(input_items, ocr)
     input_items = await transcribe_audio(input_items, asr)
-    input_data = prepare_text_blobs(input_data)
-    assert_all_files_processed(input_data)
+    input_items = prepare_text_blobs(input_items)
+    assert_all_files_processed(input_items)
 
     prompt = Prompt.from_file(
         "translate",

bpm-ai/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "bpm-ai"
-version = "1.7.2"
+version = "1.7.3"
 description = "AI task automation for BPM engines."
 authors = ["Bennet Krause <[email protected]>"]
 repository = "https://github.com/holunda-io/bpm-ai"

bpm-ai/tests/test_translate.py

@@ -80,6 +80,30 @@ async def test_translate_none(llm):
     assert result["subject"] is None
 
 
+async def test_translate_image(llm):
+    input_data = {
+        "doc": "files/invoice.png",
+    }
+    llm = llm or FakeLLM(
+        name="openai",
+        supports_images=True,
+        responses=[
+            AssistantMessage(
+                content={"doc": "Rechnung\n\nVon:\nDEMO - Sliced Invoices\nSuite 5A-1204\n123 Somewhere Street\nYour City AZ 12345\n[email protected]\n\nRechnungsnummer: INV-3337"}
+            )
+        ]
+    )
+    result = await translate_llm(
+        llm=llm,
+        input_data=input_data,
+        target_language="German",
+    )
+    #if isinstance(llm, FakeLLM):
+    #    llm.assert_last_request_contains("[email protected]")
+
+    assert "Rechnung" in result["doc"]
+
+
 async def test_translate_empty(llm):
     input_data = {}
     llm = llm or FakeLLM(name="openai")