fixes

llmproxymetrics.go

@@ -2,6 +2,7 @@ package main
 
 import (
 	"bufio"
+	"bytes"
 	"encoding/json"
 	"fmt"
 	"io"
@@ -39,6 +40,9 @@ func createProxy(target *url.URL) func(http.ResponseWriter, *http.Request) {
 			req.URL.Scheme = target.Scheme
 			req.URL.Host = target.Host
 		}
+		data, err := io.ReadAll(r.Body)
+		if err == nil {
+			r.Body = io.NopCloser(bytes.NewReader(data))
 			var jsonData map[string]interface{}
 			json.NewDecoder(r.Body).Decode(&jsonData)
 
@@ -49,6 +53,10 @@ func createProxy(target *url.URL) func(http.ResponseWriter, *http.Request) {
 					streamed.Inc()
 				}
 			}
+		} else {
+			fmt.Println("Error reading body:", err)
+		}
+		r.Body = io.NopCloser(bytes.NewReader(data))
 
 		modifyResponse := func(response *http.Response) error {
 			pr, pw := io.Pipe()
@@ -89,6 +97,9 @@ func createProxy(target *url.URL) func(http.ResponseWriter, *http.Request) {
 }
 
 func handleJsonLine(line []byte) {
+	if len(line) == 0 {
+		return
+	}
 	var jsonData map[string]interface{}
 	err := json.Unmarshal([]byte(line), &jsonData)
 	if err != nil {
@@ -100,7 +111,8 @@ func handleJsonLine(line []byte) {
 		duration := jsonData["eval_duration"].(float64) / 1000000000.0
 		fmt.Printf("Duration: %.2f seconds\n", duration)
 		opsProcessed.Inc()
-		tokens_out.Add(jsonData["prompt_eval_count"].(float64))
+		tokens_out.Add(jsonData["eval_count"].(float64))
+		tokens_in.Add(jsonData["prompt_eval_count"].(float64))
 		eval_time.Observe(duration)
 	}
 }
@@ -122,7 +134,10 @@ var tokens_out = promauto.NewCounter(prometheus.CounterOpts{
 	Name: "llmproxymetrics_tokens_out",
 	Help: "Tokens generated.",
 })
-
+var tokens_in = promauto.NewCounter(prometheus.CounterOpts{
+	Name: "llmproxymetrics_tokens_in",
+	Help: "Tokens input.",
+})
 var eval_time = promauto.NewHistogram(prometheus.HistogramOpts{
 	Name:    "llmproxymetrics_eval_time",
 	Help:    "Tokens generated.",