examples: review follow-up cleanup (simplify types, restore decorators, minimal ignores)

Author: samojavo

examples/custom_output_files/main.py

@@ -4,7 +4,6 @@
 
 import cocoindex
 from markdown_it import MarkdownIt
-from typing import cast
 
 _markdown_it = MarkdownIt("gfm-like")
 
@@ -97,7 +96,8 @@ def mutate(
 
 @cocoindex.op.function()
 def markdown_to_html(text: str) -> str:
-    return cast(str, _markdown_it.render(text))
+    # The underlying library lacks precise typing; the return here is known to be str.
+    return _markdown_it.render(text)  # type: ignore[no-any-return]
 
 
 @cocoindex.flow_def(name="CustomOutputFiles")

examples/fastapi_server_docker/main.py

@@ -104,6 +104,7 @@ async def lifespan(app: FastAPI) -> AsyncIterator[None]:
 fastapi_app = FastAPI(lifespan=lifespan)
 
 
+@fastapi_app.get("/search")  # type: ignore
 def search_endpoint(
     request: Request,
     q: str = Query(..., description="Search query"),
@@ -114,9 +115,5 @@ def search_endpoint(
     return {"results": results}
 
 
-# Attach route without using decorator to avoid untyped-decorator when FastAPI types are unavailable
-fastapi_app.get("/search")(search_endpoint)
-
-
 if __name__ == "__main__":
     uvicorn.run(fastapi_app, host="0.0.0.0", port=8080)

examples/image_search/main.py

@@ -3,7 +3,7 @@
 import io
 import os
 from contextlib import asynccontextmanager
-from typing import Any, Literal, Final, TypeAlias, cast, AsyncIterator
+from typing import Any, cast, AsyncIterator
 
 import cocoindex
 import torch
@@ -19,8 +19,7 @@
 QDRANT_URL = os.getenv("QDRANT_URL", "http://localhost:6334/")
 QDRANT_COLLECTION = "ImageSearch"
 CLIP_MODEL_NAME = "openai/clip-vit-large-patch14"
-CLIP_MODEL_DIMENSION: Final[int] = 768
-CLIPVector: TypeAlias = cocoindex.Vector[cocoindex.Float32, Literal[768]]
+# Using simple list[float] for embeddings for readability in example code.
 
 
 @functools.cache
@@ -44,7 +43,7 @@ def embed_query(text: str) -> list[float]:
 @cocoindex.op.function(cache=True, behavior_version=1, gpu=True)
 def embed_image(
     img_bytes: bytes,
-) -> CLIPVector:
+) -> list[float]:
     """
     Convert image to embedding using CLIP model.
     """
@@ -53,7 +52,7 @@ def embed_image(
     inputs = processor(images=image, return_tensors="pt")
     with torch.no_grad():
         features = model.get_image_features(**inputs)
-    return cast(CLIPVector, features[0].tolist())
+    return cast(list[float], features[0].tolist())
 
 
 # CocoIndex flow: Ingest images, extract captions, embed, export to Qdrant
@@ -142,6 +141,7 @@ async def lifespan(app: FastAPI) -> AsyncIterator[None]:
 
 
 # --- Search API ---
+@app.get("/search")  # type: ignore
 def search(
     q: str = Query(..., description="Search query"),
     limit: int = Query(5, description="Number of results"),
@@ -171,5 +171,4 @@ def search(
     }
 
 
-# Attach route without using decorator to avoid untyped-decorator when FastAPI types are unavailable
-app.get("/search")(search)
+# Route attached via decorator above for readability

examples/pdf_elements_embedding/main.py

@@ -7,16 +7,14 @@
 from dataclasses import dataclass
 from pypdf import PdfReader
 from transformers import CLIPModel, CLIPProcessor
-from typing import Literal, TypeAlias, Final, cast
+from typing import cast
 
 
 QDRANT_GRPC_URL = "http://localhost:6334"
 QDRANT_COLLECTION_IMAGE = "PdfElementsEmbeddingImage"
 QDRANT_COLLECTION_TEXT = "PdfElementsEmbeddingText"
 
 CLIP_MODEL_NAME = "openai/clip-vit-large-patch14"
-CLIP_MODEL_DIMENSION: Final = 768
-ClipVectorType: TypeAlias = cocoindex.Vector[cocoindex.Float32, Literal[768]]
 
 IMG_THUMBNAIL_SIZE = (512, 512)
 
@@ -29,7 +27,7 @@ def get_clip_model() -> tuple[CLIPModel, CLIPProcessor]:
 
 
 @cocoindex.op.function(cache=True, behavior_version=1, gpu=True)
-def clip_embed_image(img_bytes: bytes) -> ClipVectorType:
+def clip_embed_image(img_bytes: bytes) -> list[float]:
     """
     Convert image to embedding using CLIP model.
     """
@@ -38,18 +36,18 @@ def clip_embed_image(img_bytes: bytes) -> ClipVectorType:
     inputs = processor(images=image, return_tensors="pt")
     with torch.no_grad():
         features = model.get_image_features(**inputs)
-    return cast(ClipVectorType, features[0].tolist())
+    return cast(list[float], features[0].tolist())
 
 
-def clip_embed_query(text: str) -> ClipVectorType:
+def clip_embed_query(text: str) -> list[float]:
     """
     Embed the caption using CLIP model.
     """
     model, processor = get_clip_model()
     inputs = processor(text=[text], return_tensors="pt", padding=True)
     with torch.no_grad():
         features = model.get_text_features(**inputs)
-    return cast(ClipVectorType, features[0].tolist())
+    return cast(list[float], features[0].tolist())
 
 
 @cocoindex.transform_flow()