fix(hermes): prefer /v1/runs over /v1/responses to reuse session_id (#275)

- /v1/responses ignores body.session_id and generates a fresh server-side
  session for each request, causing the Hermes sessions list to grow by
  one entry per ClawPanel message.
- /v1/runs honors body.session_id directly, so reversing the call priority
  (try /v1/runs first; fall back to /v1/responses on HTTP 404 for older
  Hermes Agent builds) lets sessions group naturally without breaking
  backward compatibility.

scripts/dev-api.js

@@ -8617,9 +8617,8 @@ async function _handleHermesAgentRunStream(req, res, args = {}) {
     if (args.conversationHistory) payload.conversation_history = args.conversationHistory
     if (args.instructions) payload.instructions = args.instructions
 
-    const handledByResponses = await _tryHermesResponsesStream(gwUrl, apiKey, payload, args, controller, res)
-    if (handledByResponses) return
-
+    // 优先 /v1/runs：支持 body.session_id 复用，避免 Hermes session 暴增（#275）。
+    // /v1/responses 上游强制每次生成新 UUID 作 session id，只作为老版本兼容的 fallback。
     const startedResp = await globalThis.fetch(`${gwUrl}/v1/runs`, {
       method: 'POST',
       headers,
@@ -8627,6 +8626,13 @@ async function _handleHermesAgentRunStream(req, res, args = {}) {
       signal: controller.signal,
     })
     if (!startedResp.ok) {
+      // 404 → 老版本 Hermes Agent 没 /v1/runs，降级 /v1/responses
+      if (startedResp.status === 404) {
+        try { await startedResp.body?.cancel() } catch {}
+        const handledByResponses = await _tryHermesResponsesStream(gwUrl, apiKey, payload, args, controller, res)
+        if (handledByResponses) return
+        throw new Error('HTTP 404: /v1/runs 不存在，且 /v1/responses fallback 失败')
+      }
       const text = await startedResp.text()
       throw new Error(`HTTP ${startedResp.status}: ${text}`)
     }

src-tauri/src/commands/hermes.rs

@@ -3394,12 +3394,9 @@ pub async fn hermes_agent_run(
         payload["instructions"] = Value::String(inst.clone());
     }
 
-    if let Some(response_run_id) =
-        try_hermes_responses_run(&app, &gw_url, &api_key, &payload, session_id.as_deref()).await?
-    {
-        return Ok(response_run_id);
-    }
-
+    // 优先 /v1/runs：该端点显式支持 body.session_id，按 client 传的 session id 复用 session，
+    // 避免 Hermes 服务端 `sessions list` 中每条消息生成一个新 session（issue #275）。
+    // /v1/responses 会忽略 body.session_id 并对每次请求新建 session_id，所以不作为主路径。
     let client = hermes_gateway_http_client(std::time::Duration::from_secs(10))
         .map_err(|e| format!("HTTP 客户端创建失败: {e}"))?;
 
@@ -3425,6 +3422,16 @@ pub async fn hermes_agent_run(
     };
     if !resp.status().is_success() {
         let status = resp.status().as_u16();
+        // 404 → 老版本 Hermes Agent 没有 /v1/runs，降级到 /v1/responses 兼容
+        // （代价：session 会暴增，但至少能用；建议用户升级 Hermes Agent）
+        if status == 404 {
+            if let Some(response_run_id) =
+                try_hermes_responses_run(&app, &gw_url, &api_key, &payload, session_id.as_deref())
+                    .await?
+            {
+                return Ok(response_run_id);
+            }
+        }
         let text = resp.text().await.unwrap_or_default();
         return Err(format!("HTTP {status}: {text}"));
     }