diff --git a/trustgraph-flow/trustgraph/prompt/template/service.py b/trustgraph-flow/trustgraph/prompt/template/service.py
index c9c0c87b..c599ce77 100755
--- a/trustgraph-flow/trustgraph/prompt/template/service.py
+++ b/trustgraph-flow/trustgraph/prompt/template/service.py
@@ -128,12 +128,14 @@ class Processor(FlowProcessor):
 
                         # Always send a message if there's content OR if it's the final message
                         if resp.response or is_final:
-                            # Forward each chunk immediately
                             r = PromptResponse(
                                 text=resp.response if resp.response else "",
                                 object=None,
                                 error=None,
                                 end_of_stream=is_final,
+                                in_token=resp.in_token,
+                                out_token=resp.out_token,
+                                model=resp.model,
                             )
                             await flow("response").send(r, properties={"id": id})
 
@@ -155,6 +157,8 @@ class Processor(FlowProcessor):
                 return
 
             # Non-streaming path (original behavior)
+            usage = {}
+
             async def llm(system, prompt):
 
                 logger.debug(f"System prompt: {system}")
@@ -164,6 +168,9 @@ class Processor(FlowProcessor):
                     result = await flow("text-completion-request").text_completion(
                         system = system, prompt = prompt,
                     )
+                    usage["in_token"] = result.in_token
+                    usage["out_token"] = result.out_token
+                    usage["model"] = result.model
                     return result.text
                 except Exception as e:
                     logger.error(f"LLM Exception: {e}", exc_info=True)
@@ -186,6 +193,9 @@ class Processor(FlowProcessor):
                     object=None,
                     error=None,
                     end_of_stream=True,
+                    in_token=usage.get("in_token", 0),
+                    out_token=usage.get("out_token", 0),
+                    model=usage.get("model", ""),
                 )
 
                 await flow("response").send(r, properties={"id": id})
@@ -202,6 +212,9 @@ class Processor(FlowProcessor):
                     object=json.dumps(resp),
                     error=None,
                     end_of_stream=True,
+                    in_token=usage.get("in_token", 0),
+                    out_token=usage.get("out_token", 0),
+                    model=usage.get("model", ""),
                 )
 
                 await flow("response").send(r, properties={"id": id})