ts startup

app/main.py

@@ -20,9 +20,9 @@ import tempfile
 from utils import getconfig
 
 config = getconfig("params.cfg")
-RETRIEVER = config.get("retriever", "RETRIEVER")
-GENERATOR = config.get("generator", "GENERATOR")
-INGESTOR = config.get("ingestor", "INGESTOR")
+RETRIEVER = config.get("retriever", "RETRIEVER", fallback="https://giz-chatfed-retriever.hf.space")
+GENERATOR = config.get("generator", "GENERATOR", fallback="https://giz-chatfed-generator.hf.space")
+INGESTOR = config.get("ingestor", "INGESTOR", fallback="https://mtyrrell-chatfed-ingestor.hf.space")
 MAX_CONTEXT_CHARS = config.get("general", "MAX_CONTEXT_CHARS")
 
 logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
@@ -84,7 +84,7 @@ def ingest_node(state: GraphState) -> GraphState:
         try:
             # Call the ingestor's ingest endpoint - use gradio_client.file() for proper formatting
             ingestor_context = client.predict(
-                file=tmp_file_path,
+                file(tmp_file_path),  # Use gradio_client.file() to properly format
                 api_name="/ingest"
             )
             
@@ -122,6 +122,52 @@ def ingest_node(state: GraphState) -> GraphState:
             "ingestion_error": str(e)
         })
         return {"ingestor_context": "", "metadata": metadata}
+    
+    try:
+        client = Client(INGESTOR)
+        
+        # Create a temporary file to upload
+        with tempfile.NamedTemporaryFile(delete=False, suffix=os.path.splitext(state["filename"])[1]) as tmp_file:
+            tmp_file.write(state["file_content"])
+            tmp_file_path = tmp_file.name
+        
+        try:
+            # Call the ingestor's ingest endpoint - returns context directly
+            ingestor_context = client.predict(
+                file=tmp_file_path,
+                api_name="/ingest"
+            )
+            
+            logger.info(f"Ingest result length: {len(ingestor_context) if ingestor_context else 0}")
+            
+        finally:
+            # Clean up temporary file
+            os.unlink(tmp_file_path)
+        
+        duration = (datetime.now() - start_time).total_seconds()
+        metadata = state.get("metadata", {})
+        metadata.update({
+            "ingestion_duration": duration,
+            "ingestor_context_length": len(ingestor_context) if ingestor_context else 0,
+            "ingestion_success": True
+        })
+        
+        return {
+            "ingestor_context": ingestor_context,
+            "metadata": metadata
+        }
+        
+    except Exception as e:
+        duration = (datetime.now() - start_time).total_seconds()
+        logger.error(f"Ingestion failed: {str(e)}")
+        
+        metadata = state.get("metadata", {})
+        metadata.update({
+            "ingestion_duration": duration,
+            "ingestion_success": False,
+            "ingestion_error": str(e)
+        })
+        return {"ingestor_context": "", "metadata": metadata}
 
 def retrieve_node(state: GraphState) -> GraphState:
     start_time = datetime.now()
@@ -170,8 +216,8 @@ def generate_node(state: GraphState) -> GraphState:
         ingestor_context = state.get("ingestor_context", "")
         
         # Limit context size to prevent token overflow
-        MAX_CONTEXT_CHARS = int(MAX_CONTEXT_CHARS)  # Adjust based on your model's limits
-
+        MAX_CONTEXT_CHARS = int(config.get("general", "MAX_CONTEXT_CHARS"))
+        
         combined_context = ""
         if ingestor_context and retrieved_context:
             # Prioritize ingestor context, truncate if needed
@@ -355,7 +401,6 @@ def process_query_langserve(input_data: ChatFedInput) -> ChatFedOutput:
     )
     return ChatFedOutput(result=result["result"], metadata=result["metadata"])
 
-# This is not working currently... Problematic because HF doesn't allow > 1 port open at the same time
 def create_gradio_interface():
     with gr.Blocks(title="ChatFed Orchestrator") as demo:
         gr.Markdown("# ChatFed Orchestrator")
@@ -416,25 +461,42 @@ async def root():
         }
     }
 
-# LangServe routes (these are the main endpoints)
-add_routes(
-    app,
-    RunnableLambda(process_query_langserve),
-    path="/chatfed",
-    input_type=ChatFedInput,
-    output_type=ChatFedOutput
-)
-
-add_routes(
-    app,
-    RunnableLambda(chatui_adapter),
-    path="/chatfed-ui-stream",
-    input_type=ChatUIInput,
-    output_type=str,
-    enable_feedback_endpoint=True,
-    enable_public_trace_link_endpoint=True,
-)
+# Additional endpoint for file uploads via API
+@app.post("/chatfed-with-file")
+async def chatfed_with_file(
+    query: str = Form(...),
+    file: Optional[UploadFile] = File(None),
+    reports_filter: Optional[str] = Form(""),
+    sources_filter: Optional[str] = Form(""),
+    subtype_filter: Optional[str] = Form(""),
+    year_filter: Optional[str] = Form(""),
+    session_id: Optional[str] = Form(None),
+    user_id: Optional[str] = Form(None)
+):
+    """Endpoint for queries with optional file attachments"""
+    file_content = None
+    filename = None
+    
+    if file:
+        file_content = await file.read()
+        filename = file.filename
+    
+    result = process_query_core(
+        query=query,
+        reports_filter=reports_filter,
+        sources_filter=sources_filter,
+        subtype_filter=subtype_filter,
+        year_filter=year_filter,
+        file_content=file_content,
+        filename=filename,
+        session_id=session_id,
+        user_id=user_id,
+        return_metadata=True
+    )
+    
+    return ChatFedOutput(result=result["result"], metadata=result["metadata"])
 
+# Additional endpoint for file uploads via API
 @app.post("/chatfed-with-file")
 async def chatfed_with_file(
     query: str = Form(...),
@@ -469,6 +531,25 @@ async def chatfed_with_file(
     
     return ChatFedOutput(result=result["result"], metadata=result["metadata"])
 
+# LangServe routes (these are the main endpoints)
+add_routes(
+    app,
+    RunnableLambda(process_query_langserve),
+    path="/chatfed",
+    input_type=ChatFedInput,
+    output_type=ChatFedOutput
+)
+
+add_routes(
+    app,
+    RunnableLambda(chatui_adapter),
+    path="/chatfed-ui-stream",
+    input_type=ChatUIInput,
+    output_type=str,
+    enable_feedback_endpoint=True,
+    enable_public_trace_link_endpoint=True,
+)
+
 def run_gradio_server():
     demo = create_gradio_interface()
     demo.launch(