FastAPI RAG backend (Docker)

app.py

@@ -28,6 +28,9 @@ class AskRequest(BaseModel):
     question: str
     top_k: int = 3
 
+@app.get("/")
+def root():
+    return {"status": "ok", "service": "rag-backend", "docs": "/docs"}
 
 @app.post("/ingest")
 async def ingest(file: UploadFile = File(...)) -> Dict[str, Any]:

rag.py

@@ -22,8 +22,8 @@ EMBEDDER = SentenceTransformer("all-MiniLM-L6-v2")
 # For MVP: use a smallish instruct model if possible
 # NOTE: Mistral 7B is heavy; if you can't run it locally, use a smaller HF model.
 GENERATOR = pipeline(
-    "text-generation",
-    model="google/flan-t5-base",   # safe CPU model for MVP
+    "text2text-generation",
+    model="google/flan-t5-base",
     max_new_tokens=256
 )
 
@@ -80,13 +80,8 @@ def generate_answer(question: str, context: str) -> str:
         f"Context:\n{context}\n\nQuestion:\n{question}\n\nAnswer:"
     )
 
-    # For flan-t5 pipeline: use text2text-generation instead normally,
-    # but keeping this simple - if needed swap pipeline.
     out = GENERATOR(prompt)
-    # pipeline output format differs by model; handle safely:
-    if isinstance(out, list) and out and "generated_text" in out[0]:
-        return out[0]["generated_text"]
-    return str(out)
+    return out[0]["generated_text"].strip()
 
 
 def create_session(chunks: List[str]) -> str: