@elvatis_com/openclaw-cli-bridge-elvatis 2.6.0 → 2.6.2

package/README.md

@@ -2,7 +2,7 @@
 
 > OpenClaw plugin that bridges locally installed AI CLIs (Codex, Gemini, Claude Code, OpenCode, Pi) as model providers — with slash commands for instant model switching, restore, health testing, and model listing.
 
-**Current version:** `2.6.0`
+**Current version:** `2.6.2`
 
 ---
 
@@ -296,6 +296,26 @@ In `~/.openclaw/openclaw.json` → `plugins.entries.openclaw-cli-bridge-elvatis.
 }
 ```
 
+### Required: OpenClaw LLM idle timeout
+
+**Important:** OpenClaw's embedded agent has a default **LLM idle timeout of 60 seconds**. CLI subprocesses (especially Claude/Gemini with large prompts) often need longer than 60s before producing the first token. Without this setting, you'll see `exit 143` / `status:408` / `FailoverError: LLM request timed out.`
+
+Add to `~/.openclaw/openclaw.json`:
+
+```json5
+{
+  "agents": {
+    "defaults": {
+      "llm": {
+        "idleTimeoutSeconds": 300  // 5 min — must be >= longest per-model timeout
+      }
+    }
+  }
+}
+```
+
+> **Note:** Cron-triggered agents automatically have `idleTimeoutSeconds: 0` (disabled) in OpenClaw, so crons are not affected.
+
 ---
 
 ## Model Allowlist
@@ -386,6 +406,16 @@ npm run ci          # lint + typecheck + test
 
 ## Changelog
 
+### v2.6.2
+- **fix:** Codex CLI `--quiet` flag removed in latest Codex version — replaced with `codex exec` subcommand for non-interactive execution. All `openai-codex/*` models were failing with "unexpected argument '--quiet'" error.
+- **fix:** Agent model routing — 10 agents referenced non-existent `google-gemini-cli` provider. Remapped to `vllm/cli-gemini/*` (OAuth-based, stable) for reliable Gemini access.
+- **fix:** Main agent fallback `openai-codex/gpt-5.1` routed via direct OpenAI API (broken OAuth scopes) — now routes through CLI bridge (`vllm/openai-codex/gpt-5.1`)
+
+### v2.6.1
+- **fix:** Root cause of Exit 143 / 408 timeouts identified — OpenClaw's `agents.defaults.llm.idleTimeoutSeconds` defaults to 60s, which is too short for CLI subprocesses that need time before producing the first token
+- **feat:** Startup warning when `idleTimeoutSeconds` is not set or < 120s — tells you exactly what to add to `openclaw.json`
+- **docs:** Added "Required: OpenClaw LLM idle timeout" section to README with recommended config
+
 ### v2.6.0
 - **feat:** Provider session registry (`src/provider-sessions.ts`) — persistent sessions that survive across runs. When a CLI run times out, the session is preserved (not deleted) so follow-up runs can resume in the same context. Sessions are stored in `~/.openclaw/cli-bridge/sessions.json`.
 - **feat:** Centralized config module (`src/config.ts`) — all magic numbers, timeouts, paths, ports, and model defaults extracted into one file. No more scattered hardcoded values.

package/SKILL.md

@@ -68,4 +68,4 @@ On gateway restart, if any session has expired, a **WhatsApp alert** is sent aut
 
 See `README.md` for full configuration reference and architecture diagram.
 
-**Version:** 2.6.0
+**Version:** 2.6.2

package/index.ts

@@ -1406,6 +1406,27 @@ const plugin = {
           api.logger.info(
             `[cli-bridge] proxy ready on :${port} — vllm/cli-gemini/* and vllm/cli-claude/* available`
           );
+
+          // Warn if OpenClaw's LLM idle timeout is too low for CLI models.
+          // CLI subprocesses (especially with large prompts) need time before producing
+          // the first token. The default 60s causes premature 408 timeouts.
+          const llmIdleTimeout = (api.pluginConfig as Record<string, unknown>)?._resolvedAgentDefaults?.llm?.idleTimeoutSeconds;
+          if (llmIdleTimeout === undefined) {
+            // Can't read the resolved config — check the raw file instead
+            try {
+              const ocConfig = JSON.parse(readFileSync(join(OPENCLAW_DIR, "openclaw.json"), "utf-8")) as Record<string, unknown>;
+              const agentDefaults = (ocConfig?.agents as Record<string, unknown>)?.defaults as Record<string, unknown> | undefined;
+              const llm = agentDefaults?.llm as Record<string, unknown> | undefined;
+              const idle = llm?.idleTimeoutSeconds as number | undefined;
+              if (idle === undefined || (idle > 0 && idle < 120)) {
+                api.logger.warn(
+                  `[cli-bridge] ⚠️  agents.defaults.llm.idleTimeoutSeconds is ${idle ?? "not set (default 60s)"} — ` +
+                  `CLI models need at least 120–300s. Set to 300 in openclaw.json to prevent premature 408 timeouts.`
+                );
+              }
+            } catch { /* non-fatal — just skip the check */ }
+          }
+
           const result = patchOpencllawConfig(port);
           if (result.patched) {
             api.logger.info(

package/openclaw.plugin.json

@@ -2,7 +2,7 @@
   "id": "openclaw-cli-bridge-elvatis",
   "slug": "openclaw-cli-bridge-elvatis",
   "name": "OpenClaw CLI Bridge",
-  "version": "2.6.0",
+  "version": "2.6.2",
   "license": "MIT",
   "description": "Phase 1: openai-codex auth bridge. Phase 2: local HTTP proxy routing model calls through gemini/claude CLIs (vllm provider).",
   "providers": [

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@elvatis_com/openclaw-cli-bridge-elvatis",
-  "version": "2.6.0",
+  "version": "2.6.2",
   "description": "Bridges gemini, claude, and codex CLI tools as OpenClaw model providers. Reads existing CLI auth without re-login.",
   "type": "module",
   "openclaw": {

package/src/cli-runner.ts

@@ -592,7 +592,7 @@ export async function runCodex(
   opts?: { tools?: ToolDefinition[]; mediaFiles?: MediaFile[]; log?: (msg: string) => void }
 ): Promise<string> {
   const model = stripPrefix(modelId);
-  const args = ["--model", model, "--quiet", "--full-auto"];
+  const args = ["exec", "--model", model, "--full-auto"];
 
   // Codex supports native image input via -i flag
   if (opts?.mediaFiles?.length) {

package/src/session-manager.ts

@@ -332,7 +332,7 @@ export class SessionManager {
       }
       return {
         cmd: "codex",
-        args: ["--model", modelName, "--quiet", "--full-auto"],
+        args: ["exec", "--model", modelName, "--full-auto"],
         cwd,
         useStdin: true,
       };

package/test/cli-runner-extended.test.ts

@@ -87,7 +87,7 @@ describe("runCodex()", () => {
     expect(result).toBe("codex result");
     expect(mockSpawn).toHaveBeenCalledWith(
       "codex",
-      ["--model", "gpt-5.3-codex", "--quiet", "--full-auto"],
+      ["exec", "--model", "gpt-5.3-codex", "--full-auto"],
       expect.any(Object)
     );
   });