Mirror Zhanlu client: GLM tool_stream, shared v4 UUID request id, drop chat state:ERROR

internal/openai/types.go

@@ -1,6 +1,9 @@
 package openai
 
-import "encoding/json"
+import (
+	"encoding/json"
+	"strings"
+)
 
 type ChatCompletionRequest struct {
 	Model    string                     `json:"model"`
@@ -40,6 +43,11 @@ func (r ChatCompletionRequest) MarshalForUpstream() ([]byte, error) {
 			m[k] = anyValue
 		}
 	}
+	// Mirror the Zhanlu plugin: GLM models require {tool_stream:true} to
+	// stream tool calls back to the client (matches /glm/i.test(model)).
+	if strings.Contains(strings.ToLower(r.Model), "glm") {
+		m["tool_stream"] = true
+	}
 	return json.Marshal(m)
 }
 

internal/server/server.go

@@ -440,7 +440,7 @@ func (s *Server) postPhoneAPI(endpoint string, payload map[string]string, out *p
 	req.Header.Set("Content-Type", "application/json")
 	req.Header.Set("plugin_type", "zhanlu_ide")
 	req.Header.Set("plugin_version", s.cfg.PluginVersion)
-	req.Header.Set("request", randomRequestID())
+	req.Header.Set("request", util.RandomRequestID())
 	resp, err := s.upstreamHTTPClient().Do(req)
 	if err != nil {
 		return err
@@ -1093,10 +1093,6 @@ func forEachSSEChunk(r io.Reader, fn func([]byte) error) error {
 		if payload == "" || payload == "[DONE]" {
 			continue
 		}
-		var upstreamError map[string]any
-		if json.Unmarshal([]byte(payload), &upstreamError) == nil && upstreamError["state"] == "ERROR" {
-			return fmt.Errorf("zhanlu upstream error: %v", upstreamError["errorMessage"])
-		}
 		if err := fn([]byte(payload)); err != nil {
 			return err
 		}

internal/util/util.go

@@ -2,7 +2,12 @@
 // avoid divergent same-named copies.
 package util
 
-import "strings"
+import (
+	"crypto/rand"
+	"fmt"
+	"strings"
+	"time"
+)
 
 // FirstNonEmpty returns the first trimmed-non-empty value, or "" when none of
 // the values are non-empty. Callers that need a specific fallback for the
@@ -15,3 +20,17 @@ func FirstNonEmpty(values ...string) string {
 	}
 	return ""
 }
+
+// RandomRequestID returns a random v4 UUID string, matching the format the
+// Zhanlu plugin sends in the `request` header of every gateway call
+// (crypto.randomUUID()). Extracted so the login and phone-code paths share
+// one implementation.
+func RandomRequestID() string {
+	b := make([]byte, 16)
+	if _, err := rand.Read(b); err != nil {
+		return fmt.Sprintf("%d", time.Now().UnixNano())
+	}
+	b[6] = (b[6] & 0x0f) | 0x40
+	b[8] = (b[8] & 0x3f) | 0x80
+	return fmt.Sprintf("%x-%x-%x-%x-%x", b[0:4], b[4:6], b[6:8], b[8:10], b[10:])
+}

internal/zhanlu/client.go

@@ -210,7 +210,7 @@ func (c *Client) setPluginHeaders(req *http.Request) {
 	req.Header.Set("Content-Type", "application/json")
 	req.Header.Set("plugin_type", "zhanlu_ide")
 	req.Header.Set("plugin_version", c.PluginVersion)
-	req.Header.Set("request", randomRequestID())
+	req.Header.Set("request", util.RandomRequestID())
 }
 
 func decodeJSON(resp *http.Response, out any) error {
@@ -249,12 +249,4 @@ func randomAlnum(n int) string {
 	return string(b)
 }
 
-func randomRequestID() string {
-	b := make([]byte, 16)
-	if _, err := rand.Read(b); err != nil {
-		return fmt.Sprintf("%d", time.Now().UnixNano())
-	}
-	b[6] = (b[6] & 0x0f) | 0x40
-	b[8] = (b[8] & 0x3f) | 0x80
-	return fmt.Sprintf("%x-%x-%x-%x-%x", b[0:4], b[4:6], b[6:8], b[8:10], b[10:])
-}
+