better error handling

VERSION

@@ -1 +1 @@
-v0.1.5
+v0.1.6

impl/main.py

@@ -5,7 +5,7 @@
 import httpx
 import openai
 import uvicorn
-from fastapi import FastAPI, Request
+from fastapi import FastAPI, Request, HTTPException
 from fastapi.responses import JSONResponse
 from prometheus_client import Counter, Summary, Histogram
 from prometheus_fastapi_instrumentator import Instrumentator
@@ -158,6 +158,8 @@ async def generic_exception_handler(request: Request, exc: Exception):
     # Log the error
     logger.error(f"Unexpected error: {exc}")
 
+    if isinstance(exc, HTTPException):
+        raise exec
     # Return an error response, not sure if we want to return all errors but at least this surfaces things like bad embedding model. Though that should be a 4xx error?
     return JSONResponse(
         status_code=501, content={"message": str(exc)}

impl/routes/files.py

@@ -13,7 +13,8 @@
     Path,
     Request,
     Query,
-    UploadFile, HTTPException,
+    UploadFile,
+    HTTPException,
 )
 from litellm import utils
 

impl/services/file.py

@@ -6,10 +6,11 @@
 
 import docx2txt
 import pptx
-from fastapi import UploadFile
+from fastapi import UploadFile, HTTPException
 from loguru import logger
 from PyPDF2 import PdfReader
 
+from impl.astra_vector import HandledResponse
 from impl.models import Document
 
 
@@ -32,7 +33,7 @@ def extract_text_from_filepath(filepath: str, mimetype: Optional[str] = None) ->
         if filepath.endswith(".md"):
             mimetype = "text/markdown"
         else:
-            raise Exception("Unsupported file type")
+            raise HTTPException(status_code=400, detail="Unsupported file type")
 
     try:
         with open(filepath, "rb") as file:
@@ -82,12 +83,17 @@ def extract_text_from_file(file: BufferedReader, mimetype: str) -> str:
                             extracted_text += run.text + " "
                     extracted_text += "\n"
     else:
-        extension = mimetypes.guess_extension(mimetype)[1:]
-        if extension in ("c", "cpp", "css", "html", "java", "js", "json", "md", "php", "py", "rb", "ts", "xml"):
-            extracted_text = file.read().decode("utf-8")
+        raw_extension = mimetypes.guess_extension(mimetype)
+        if raw_extension is not None:
+            extension = raw_extension[1:]
+            if extension in ("c", "cpp", "css", "html", "java", "js", "json", "md", "php", "py", "rb", "ts", "xml"):
+                extracted_text = file.read().decode("utf-8")
         else:
             # Unsupported file type
-            raise ValueError("Unsupported file type: {}".format(mimetype))
+            raise HTTPException(
+                status_code=400,
+                detail="Unsupported file type: {}".format(mimetype),
+            )
     return extracted_text
 
 

tests/streaming-assistants/test_file_embedding.py

@@ -21,4 +21,18 @@ def test_file_embedding_python(patched_openai_client):
         purpose="assistants",
         embedding_model="text-embedding-3-large",
     )
-    logger.info(file)
+    logger.info(file)
+
+def test_file_embedding_python(patched_openai_client):
+    try:
+        patched_openai_client.files.create(
+            file=open(
+                "./tests/fixtures/invalid",
+                "rb",
+            ),
+            purpose="assistants",
+            embedding_model="text-embedding-3-large",
+        )
+    except Exception as e:
+        assert e.status_code == 400
+        logger.info(e)