From d7e1b2a2318e4f32cdc200ad013c2a94eb16d116 Mon Sep 17 00:00:00 2001
From: "Laisky.Cai" <github@laisky.com>
Date: Sun, 5 Jan 2025 02:42:37 +0000
Subject: [PATCH 1/2] fix: o1 do not support system prompt and max_tokens

---
 relay/adaptor/openai/adaptor.go | 15 +++++++++++++++
 relay/billing/ratio/model.go    |  2 +-
 2 files changed, 16 insertions(+), 1 deletion(-)

diff --git a/relay/adaptor/openai/adaptor.go b/relay/adaptor/openai/adaptor.go
index 6946e402..ede9fc13 100644
--- a/relay/adaptor/openai/adaptor.go
+++ b/relay/adaptor/openai/adaptor.go
@@ -82,6 +82,21 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *model.G
 		}
 		request.StreamOptions.IncludeUsage = true
 	}
+
+	// o1/o1-mini/o1-preview do not support system prompt and max_tokens
+	if strings.HasPrefix(request.Model, "o1") {
+		request.MaxTokens = 0
+		request.Messages = func(raw []model.Message) (filtered []model.Message) {
+			for i := range raw {
+				if raw[i].Role != "system" {
+					filtered = append(filtered, raw[i])
+				}
+			}
+
+			return
+		}(request.Messages)
+	}
+
 	return request, nil
 }
 
diff --git a/relay/billing/ratio/model.go b/relay/billing/ratio/model.go
index f83aa70c..e7c89ca7 100644
--- a/relay/billing/ratio/model.go
+++ b/relay/billing/ratio/model.go
@@ -466,7 +466,7 @@ func GetCompletionRatio(name string, channelType int) float64 {
 		}
 		return 2
 	}
-	// including o1, o1-preview, o1-mini
+	// including o1/o1-preview/o1-mini
 	if strings.HasPrefix(name, "o1") {
 		return 4
 	}

From 3e17184c1e36135fe951ab302cb9eedd4275fd64 Mon Sep 17 00:00:00 2001
From: "Laisky.Cai" <github@laisky.com>
Date: Sat, 25 Jan 2025 07:20:07 +0000
Subject: [PATCH 2/2] fix: update o1 model handling to include temperature
 support

---
 relay/adaptor/openai/adaptor.go | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/relay/adaptor/openai/adaptor.go b/relay/adaptor/openai/adaptor.go
index ede9fc13..050070c9 100644
--- a/relay/adaptor/openai/adaptor.go
+++ b/relay/adaptor/openai/adaptor.go
@@ -83,8 +83,10 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *model.G
 		request.StreamOptions.IncludeUsage = true
 	}
 
-	// o1/o1-mini/o1-preview do not support system prompt and max_tokens
+	// o1/o1-mini/o1-preview do not support system prompt/max_tokens/temperature
 	if strings.HasPrefix(request.Model, "o1") {
+		temperature := float64(1)
+		request.Temperature = &temperature // Only the default (1) value is supported
 		request.MaxTokens = 0
 		request.Messages = func(raw []model.Message) (filtered []model.Message) {
 			for i := range raw {