@elvatis_com/openclaw-cli-bridge-elvatis 3.1.1 → 3.1.2

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:** `3.1.1`
+**Current version:** `3.1.2`
 
 ---
 
@@ -406,6 +406,10 @@ npm run ci          # lint + typecheck + test
 
 ## Changelog
 
+### v3.1.2
+- **fix:** fallback models returning text instead of tool_calls in a tool loop now trigger the next model in the chain. Previously Haiku would say "Lass mich das starten:" as text but never call a tool — conversation died.
+- **feat:** `[FALLBACK-NO-TOOLS]` debug log category for tool-format violations
+
 ### v3.1.1
 - **fix:** empty-response detection — models returning zero content now trigger the next fallback instead of silently stopping the chain. Previously Haiku would return empty (0 bytes) and the bridge treated it as success, leaving the user with no response.
 - **feat:** `[EMPTY]` and `[FALLBACK-EMPTY]` debug log categories for diagnosing empty model responses

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:** 3.1.1
+**Version:** 3.1.2

package/openclaw.plugin.json

@@ -2,7 +2,7 @@
   "id": "openclaw-cli-bridge-elvatis",
   "slug": "openclaw-cli-bridge-elvatis",
   "name": "OpenClaw CLI Bridge",
-  "version": "3.1.1",
+  "version": "3.1.2",
   "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": "3.1.1",
+  "version": "3.1.2",
   "description": "Bridges gemini, claude, and codex CLI tools as OpenClaw model providers. Reads existing CLI auth without re-login.",
   "type": "module",
   "openclaw": {

package/src/proxy-server.ts

@@ -949,6 +949,15 @@ async function handleRequest(
               debugLog("FALLBACK-EMPTY", `${fallbackModel} returned empty`, {});
               throw new Error(`empty response from ${fallbackModel}`);
             }
+            // If tools were requested and the last message was a tool result (gateway expects
+            // tool continuation), but the fallback model returned text instead of tool_calls —
+            // it ignored the JSON format. Try next model in chain.
+            const lastMsg = cleanMessages[cleanMessages.length - 1];
+            const inToolLoop = lastMsg?.role === "tool" || lastMsg?.role === "function";
+            if (hasTools && inToolLoop && !result.tool_calls?.length && result.content) {
+              debugLog("FALLBACK-NO-TOOLS", `${fallbackModel} returned text instead of tool_calls in tool loop`, { contentLen: result.content.length, preview: result.content.slice(0, 80) });
+              throw new Error(`${fallbackModel} returned text instead of tool_calls`);
+            }
             const fbCompTokens = estimateTokens(result.content ?? "");
             metrics.recordRequest(fallbackModel, Date.now() - fallbackStart, true, estPromptTokens, fbCompTokens, promptPreview);
             metrics.recordFallback(model, fallbackModel, isTimeout ? "timeout" : "error", primaryDuration, true);