correct response url for groq

Author: bishalprasad321

README.md

@@ -33,14 +33,16 @@ Go to your repository:
 - **Settings** → **Secrets and variables** → **Actions**
 - Add a provider-specific secret, e.g. `GROQ_API_KEY`
 
-### 3. Optional base URL for Groq
+### 3. Optional base URL for Groq OpenAI-compatible API
 
-If you are using Groq OpenAI-compatible models, set `llm_provider_url` to the Groq OpenAI endpoint:
+If you are using Groq's OpenAI-compatible models, set `llm_provider_url` to:
 
 ```yaml
 llm_provider_url: https://api.groq.com/openai/v1
 ```
 
+This will route requests to `https://api.groq.com/openai/v1/chat/completions` for chat-based model inference.
+
 ### 4. Use the Action
 
 Create `.github/workflows/ai-review.yml`:
@@ -103,7 +105,7 @@ with:
   debug: "false" # Verbose logging
 ```
 
-Use stable provider-specific model IDs. For Gemini, use `v1beta` model codes. For Groq, use free-tier model IDs such as `groq-1.5-mini` or `groq-1.5-small`.
+Use stable provider-specific model IDs. For Gemini, use `v1beta` model codes. For Groq, use models like `llama-3.1-8b-instant`, `openai/gpt-oss-20b`, or `llama-3.3-70b-versatile` (check [Groq console](https://console.groq.com/keys) for available models).
 
 ### Groq Example
 
@@ -315,7 +317,7 @@ The free tier has limits:
 If you hit quota errors:
 
 1. **Upgrade to paid plan** — Recommended for production
-2. **Use lighter models** — `gemini-2.5-flash-lite` or `groq-1.5-mini` use fewer tokens
+2. **Use lighter models** — `gemini-2.5-flash-lite` or Groq's smaller models like `llama-3.1-8b-instant` use fewer tokens
 3. **Reduce reviewer count** — Use 2 instead of 3 reviewers
 4. **Schedule reviews** — Spread runs across off-peak hours
 

action.yml

@@ -15,7 +15,7 @@ inputs:
     description: "API key for the chosen LLM provider. Use a provider-specific API key such as GROQ_API_KEY for groq."
     required: false
   llm_provider_url:
-    description: "Optional base URL for the chosen LLM provider. For Groq, use https://api.groq.com/openai/v1 if needed."
+    description: "Optional base URL for the chosen LLM provider. For Groq, use https://api.groq.com/openai/v1 (uses /chat/completions endpoint)."
     required: false
   reviewer_models:
     description: "Comma-separated list of 3 review models. Use provider-specific model IDs such as llama-3.1-8b-instant or openai/gpt-oss-20b."

dist/index.js

@@ -947,8 +947,12 @@ const FALLBACK_MODELS = {
         ],
     },
     groq: {
-        reviewer: ["groq-1.5-mini", "groq-1.5-small"],
-        judge: ["groq-1.5-small", "groq-1.5-mini"],
+        reviewer: [
+            "llama-3.1-8b-instant",
+            "openai/gpt-oss-20b",
+            "llama-3.3-70b-versatile",
+        ],
+        judge: ["openai/gpt-oss-120b", "llama-3.3-70b-versatile"],
     },
 };
 class LLMClient {
@@ -1247,16 +1251,20 @@ RESPOND ONLY WITH THE JSON OBJECT. NO OTHER TEXT.`;
         const useOpenAICompat = trimmedBase.includes("/openai/v1");
         const openAIBase = trimmedBase.replace(/\/models$/, "");
         const url = useOpenAICompat
-            ? `${openAIBase}/completions`
+            ? `${openAIBase}/chat/completions`
             : `${trimmedBase}/${model}/generate`;
         const body = useOpenAICompat
             ? {
                 model,
-                prompt,
+                messages: [
+                    {
+                        role: "user",
+                        content: prompt,
+                    },
+                ],
                 temperature: 0.7,
                 top_p: 0.95,
                 max_tokens: 2048,
-                top_k: 40,
             }
             : {
                 input: prompt,

dist/index.js.map

@@ -126,8 +126,12 @@ const FALLBACK_MODELS: Record<
     ],
   },
   groq: {
-    reviewer: ["groq-1.5-mini", "groq-1.5-small"],
-    judge: ["groq-1.5-small", "groq-1.5-mini"],
+    reviewer: [
+      "llama-3.1-8b-instant",
+      "openai/gpt-oss-20b",
+      "llama-3.3-70b-versatile",
+    ],
+    judge: ["openai/gpt-oss-120b", "llama-3.3-70b-versatile"],
   },
 };
 
@@ -531,17 +535,21 @@ RESPOND ONLY WITH THE JSON OBJECT. NO OTHER TEXT.`;
     const useOpenAICompat = trimmedBase.includes("/openai/v1");
     const openAIBase = trimmedBase.replace(/\/models$/, "");
     const url = useOpenAICompat
-      ? `${openAIBase}/completions`
+      ? `${openAIBase}/chat/completions`
       : `${trimmedBase}/${model}/generate`;
 
     const body = useOpenAICompat
       ? {
           model,
-          prompt,
+          messages: [
+            {
+              role: "user",
+              content: prompt,
+            },
+          ],
           temperature: 0.7,
           top_p: 0.95,
           max_tokens: 2048,
-          top_k: 40,
         }
       : {
           input: prompt,