code-review-ai-cli 2.2.4__tar.gz → 2.2.5__tar.gz

{code_review_ai_cli-2.2.4 → code_review_ai_cli-2.2.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: code-review-ai-cli
-Version: 2.2.4
+Version: 2.2.5
 Summary: Automated AI-powered code review CLI for Azure DevOps / TFS Pull Requests
 License: MIT
 Keywords: code-review,ai,azure-devops,tfs,pull-request,llm

{code_review_ai_cli-2.2.4 → code_review_ai_cli-2.2.5}/code_review_ai_cli.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: code-review-ai-cli
-Version: 2.2.4
+Version: 2.2.5
 Summary: Automated AI-powered code review CLI for Azure DevOps / TFS Pull Requests
 License: MIT
 Keywords: code-review,ai,azure-devops,tfs,pull-request,llm

{code_review_ai_cli-2.2.4 → code_review_ai_cli-2.2.5}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "code-review-ai-cli"
-version = "2.2.4"
+version = "2.2.5"
 description = "Automated AI-powered code review CLI for Azure DevOps / TFS Pull Requests"
 readme = "README.md"
 requires-python = ">=3.10"

{code_review_ai_cli-2.2.4 → code_review_ai_cli-2.2.5}/src/llm_client.py

@@ -591,8 +591,9 @@ class LLMClient:
                 {"role": "user", "content": user_message},
             ],
             "max_tokens": self.config.max_tokens,
-            "temperature": self.config.temperature,
         }
+        if self.config.temperature != 99:
+            payload["temperature"] = self.config.temperature
 
         return self._http_openai_compatible(url, headers, payload)
 
@@ -637,10 +638,11 @@ class LLMClient:
                 "parts": [{"text": system_prompt}]
             },
             "generationConfig": {
-                "temperature": self.config.temperature,
                 "maxOutputTokens": self.config.max_tokens,
             },
         }
+        if self.config.temperature != 99:
+            payload["generationConfig"]["temperature"] = self.config.temperature
 
         try:
             resp = requests.post(url, headers=headers, json=payload, timeout=180)
@@ -715,12 +717,13 @@ class LLMClient:
         payload = {
             "model": model,
             "max_tokens": self.config.max_tokens,
-            "temperature": self.config.temperature,
             "system": system_prompt,
             "messages": [
                 {"role": "user", "content": user_message},
             ],
         }
+        if self.config.temperature != 99:
+            payload["temperature"] = self.config.temperature
 
         try:
             resp = requests.post(url, headers=headers, json=payload, timeout=180)
@@ -793,9 +796,10 @@ class LLMClient:
                 {"role": "system", "content": system_prompt},
                 {"role": "user", "content": user_message},
             ],
-            "temperature": self.config.temperature,
             "stream": False,
         }
+        if self.config.temperature != 99:
+            payload["temperature"] = self.config.temperature
 
         # Ollama does not require an API key, but we add max_tokens if configured
         if self.config.max_tokens:
@@ -912,8 +916,9 @@ class LLMClient:
                 {"role": "system", "content": system_prompt},
                 {"role": "user", "content": user_message},
             ],
-            "temperature": self.config.temperature,
         }
+        if self.config.temperature != 99:
+            payload["temperature"] = self.config.temperature
 
         # Add max_tokens if configured (some Copilot models
         # may not support this parameter)
@@ -1028,7 +1033,7 @@ class LLMClient:
         system_prompt: str,
         user_message: str,
     ) -> str:
-        """Calls Bedrock InvokeModel using an HTTP Bearer token (long-term API key).
+        """Calls Bedrock Converse using an HTTP Bearer token (long-term API key).
 
         Args:
             region: AWS region, e.g. ``us-east-1``.
@@ -1051,14 +1056,19 @@ class LLMClient:
 
         # The model ARN contains ':' and '/' that must be URL-encoded in the path
         model_encoded = urllib.parse.quote(self.config.model, safe="")
-        url = f"https://bedrock-runtime.{region}.amazonaws.com/model/{model_encoded}/invoke"
+        url = f"https://bedrock-runtime.{region}.amazonaws.com/model/{model_encoded}/converse"
+
+        inference_config = {"maxTokens": self.config.max_tokens}
+        if self.config.temperature != 99:
+            inference_config["temperature"] = self.config.temperature
 
         payload = json.dumps({
-            "anthropic_version": "bedrock-2023-05-31",
-            "max_tokens": self.config.max_tokens,
-            "temperature": self.config.temperature,
-            "system": system_prompt,
-            "messages": [{"role": "user", "content": user_message}],
+            "system": [{"text": system_prompt}],
+            "messages": [{
+                "role": "user",
+                "content": [{"text": user_message}],
+            }],
+            "inferenceConfig": inference_config,
         }, separators=(",", ":"))
 
         headers = {
@@ -1084,10 +1094,9 @@ class LLMClient:
         except ValueError as exc:
             raise LLMError(f"Invalid JSON from Bedrock: {resp.text[:500]}") from exc
 
-        content = data.get("content", [])
+        content = data.get("output", {}).get("message", {}).get("content", [])
         text_parts = [
-            item["text"] for item in content
-            if item.get("type") == "text" and "text" in item
+            item["text"] for item in content if "text" in item
         ]
         text = "\n".join(part for part in text_parts if part).strip()
         if not text:
@@ -1103,7 +1112,7 @@ class LLMClient:
         system_prompt: str,
         user_message: str,
     ) -> str:
-        """Calls Bedrock InvokeModel with manual AWS SigV4 HMAC-SHA256 signing.
+        """Calls Bedrock Converse with manual AWS SigV4 HMAC-SHA256 signing.
 
         Equivalent to the C# BedrockLlmClient implementation.
 
@@ -1135,15 +1144,20 @@ class LLMClient:
 
         host = f"bedrock-runtime.{region}.amazonaws.com"
         model_encoded = urllib.parse.quote(self.config.model, safe="")
-        endpoint = f"https://{host}/model/{model_encoded}/invoke"
+        endpoint = f"https://{host}/model/{model_encoded}/converse"
         service = "bedrock"
 
+        inference_config = {"maxTokens": self.config.max_tokens}
+        if self.config.temperature != 99:
+            inference_config["temperature"] = self.config.temperature
+
         payload = json.dumps({
-            "anthropic_version": "bedrock-2023-05-31",
-            "max_tokens": self.config.max_tokens,
-            "temperature": self.config.temperature,
-            "system": system_prompt,
-            "messages": [{"role": "user", "content": user_message}],
+            "system": [{"text": system_prompt}],
+            "messages": [{
+                "role": "user",
+                "content": [{"text": user_message}],
+            }],
+            "inferenceConfig": inference_config,
         }, separators=(",", ":"))
 
         now = datetime.datetime.now(datetime.timezone.utc)
@@ -1157,7 +1171,7 @@ class LLMClient:
         # Splitting by '/' and encoding each segment mirrors the C# SigV4 implementation.
         canonical_uri = "/".join(
             urllib.parse.quote(seg, safe="")
-            for seg in f"/model/{self.config.model}/invoke".split("/")
+            for seg in f"/model/{self.config.model}/converse".split("/")
         )
 
         headers_to_sign = {
@@ -1227,10 +1241,9 @@ class LLMClient:
         except ValueError as exc:
             raise LLMError(f"Invalid JSON from Bedrock: {resp.text[:500]}") from exc
 
-        content = data.get("content", [])
+        content = data.get("output", {}).get("message", {}).get("content", [])
         text_parts = [
-            item["text"] for item in content
-            if item.get("type") == "text" and "text" in item
+            item["text"] for item in content if "text" in item
         ]
         text = "\n".join(part for part in text_parts if part).strip()
         if not text:
@@ -1277,6 +1290,10 @@ class LLMClient:
             session = boto3.Session(**session_kwargs)
             client = session.client("bedrock-runtime", region_name=region)
 
+            inference_config = {"maxTokens": self.config.max_tokens}
+            if self.config.temperature != 99:
+                inference_config["temperature"] = self.config.temperature
+
             response = client.converse(
                 modelId=self.config.model,
                 system=[{"text": system_prompt}],
@@ -1286,10 +1303,7 @@ class LLMClient:
                         "content": [{"text": user_message}],
                     }
                 ],
-                inferenceConfig={
-                    "temperature": self.config.temperature,
-                    "maxTokens": self.config.max_tokens,
-                },
+                inferenceConfig=inference_config,
             )
 
             content = (

{code_review_ai_cli-2.2.4 → code_review_ai_cli-2.2.5}/src/prompts/config.yaml.template

@@ -26,7 +26,7 @@ llm:
   # Maximum tokens in response
   max_tokens: 4096
 
-  # Temperature (0.0 = deterministic, 1.0 = creative)
+  # Temperature (0.0 = deterministic, 1.0 = creative, 99 omits it for all models that dont support this)
   temperature: 0.0
 
 # ---------------------------------------------------------------------------

{code_review_ai_cli-2.2.4 → code_review_ai_cli-2.2.5}/tests/test_llm_client.py

@@ -5,6 +5,7 @@
 
 from __future__ import annotations
 
+import json
 import re
 import sys
 from pathlib import Path
@@ -606,7 +607,10 @@ def test_bedrock_boto3_success_and_error_paths(monkeypatch: pytest.MonkeyPatch)
     class FakeBedrockClient:
         """Stub Bedrock runtime client."""
 
+        last_converse_kwargs: dict[str, object] = {}
+
         def converse(self, **kwargs: object) -> dict:
+            FakeBedrockClient.last_converse_kwargs = kwargs
             return {"output": {"message": {"content": [{"text": "bedrock ok"}]}}}
 
     class FakeSession:
@@ -638,6 +642,12 @@ def test_bedrock_boto3_success_and_error_paths(monkeypatch: pytest.MonkeyPatch)
     assert client._call_bedrock("sys", "user") == "bedrock ok"
     assert client._call_bedrock_boto3("us-east-1", "sys", "user") == "bedrock ok"
 
+    client.config.temperature = 99
+    assert client._call_bedrock_boto3("us-east-1", "sys", "user") == "bedrock ok"
+    assert FakeBedrockClient.last_converse_kwargs["inferenceConfig"] == {
+        "maxTokens": 256,
+    }
+
     class BrokenSession(FakeSession):
         """Session that raises provider-side failures."""
 
@@ -689,8 +699,8 @@ def test_call_bedrock_routing_dispatches_correct_method(mocker) -> None:
 
 
 def test_call_bedrock_bearer_success_and_errors(monkeypatch: pytest.MonkeyPatch) -> None:
-    """_call_bedrock_bearer should use Bearer auth, URL-encode the model ARN and parse content."""
-    model_arn = "arn:aws:bedrock:us-east-1::foundation-model/anthropic.claude-3-sonnet"
+    """_call_bedrock_bearer should use the provider-agnostic Converse API."""
+    model_arn = "arn:aws:bedrock:us-east-1::foundation-model/openai.gpt-oss-120b-1:0"
     client = LLMClient(
         make_llm_config(
             llm_provider="bedrock",
@@ -703,15 +713,31 @@ def test_call_bedrock_bearer_success_and_errors(monkeypatch: pytest.MonkeyPatch)
     )
 
     # Success path: response parsed correctly, model ARN URL-encoded in URL
-    success_resp = FakeResponse(json_data={"content": [{"type": "text", "text": "bearer ok"}]})
+    success_resp = FakeResponse(
+        json_data={"output": {"message": {"content": [{"text": "bearer ok"}]}}}
+    )
     req_module = install_requests_raw(monkeypatch, success_resp)
     result = client._call_bedrock_bearer("us-east-1", "my-bearer-key", "sys", "user")
     assert result == "bearer ok"
     call = req_module._calls[0]
     assert "arn%3Aaws%3Abedrock" in call["url"]  # ':' encoded as %3A
-    assert "anthropic.claude-3-sonnet" in call["url"]
+    assert "openai.gpt-oss-120b-1%3A0" in call["url"]
+    assert call["url"].endswith("/converse")
     assert call["headers"]["Authorization"] == "Bearer my-bearer-key"
     assert call["headers"]["Content-Type"] == "application/json"
+    payload = json.loads(call["data"])
+    assert "anthropic_version" not in payload
+    assert payload == {
+        "system": [{"text": "sys"}],
+        "messages": [{"role": "user", "content": [{"text": "user"}]}],
+        "inferenceConfig": {"temperature": 0.2, "maxTokens": 256},
+    }
+
+    client.config.temperature = 99
+    req_module = install_requests_raw(monkeypatch, success_resp)
+    assert client._call_bedrock_bearer("us-east-1", "my-bearer-key", "sys", "user") == "bearer ok"
+    sentinel_payload = json.loads(req_module._calls[0]["data"])
+    assert sentinel_payload["inferenceConfig"] == {"maxTokens": 256}
 
     # 401 → specific authentication error message
     install_requests_raw(monkeypatch, FakeResponse(status_code=401))
@@ -724,12 +750,12 @@ def test_call_bedrock_bearer_success_and_errors(monkeypatch: pytest.MonkeyPatch)
         client._call_bedrock_bearer("us-east-1", "my-bearer-key", "sys", "user")
 
     # Empty content list → Unexpected response error
-    install_requests_raw(monkeypatch, FakeResponse(json_data={"content": []}))
+    install_requests_raw(monkeypatch, FakeResponse(json_data={"output": {"message": {"content": []}}}))
     with pytest.raises(LLMError, match="Unexpected Bedrock response"):
         client._call_bedrock_bearer("us-east-1", "my-bearer-key", "sys", "user")
 
     # Content items with no text type → Unexpected response error
-    install_requests_raw(monkeypatch, FakeResponse(json_data={"content": [{"type": "image", "data": "..."}]}))
+    install_requests_raw(monkeypatch, FakeResponse(json_data={"output": {"message": {"content": [{"image": {}}]}}}))
     with pytest.raises(LLMError, match="Unexpected Bedrock response"):
         client._call_bedrock_bearer("us-east-1", "my-bearer-key", "sys", "user")
 
@@ -741,18 +767,20 @@ def test_call_bedrock_bearer_success_and_errors(monkeypatch: pytest.MonkeyPatch)
 
 
 def test_call_bedrock_sigv4_success_and_errors(monkeypatch: pytest.MonkeyPatch) -> None:
-    """_call_bedrock_sigv4 should add SigV4 headers and parse content correctly."""
+    """_call_bedrock_sigv4 should use Converse with correctly signed requests."""
     client = LLMClient(
         make_llm_config(
             llm_provider="bedrock",
             api_key="",
-            model="anthropic.claude-3-sonnet",
+            model="openai.gpt-oss-120b-1:0",
             bedrock_region="us-east-1",
             bedrock_access_key_id="AKID",
             bedrock_secret_access_key="secret",
         )
     )
-    success_resp = FakeResponse(json_data={"content": [{"type": "text", "text": "sigv4 ok"}]})
+    success_resp = FakeResponse(
+        json_data={"output": {"message": {"content": [{"text": "sigv4 ok"}]}}}
+    )
 
     # Success path: SigV4 headers present in request
     req_module = install_requests_raw(monkeypatch, success_resp)
@@ -763,6 +791,16 @@ def test_call_bedrock_sigv4_success_and_errors(monkeypatch: pytest.MonkeyPatch)
     assert "x-amz-date" in call["headers"]
     assert "x-amz-content-sha256" in call["headers"]
     assert "x-amz-security-token" not in call["headers"]
+    assert call["url"].endswith("/converse")
+    payload = json.loads(call["data"])
+    assert "anthropic_version" not in payload
+    assert payload["inferenceConfig"] == {"temperature": 0.2, "maxTokens": 256}
+
+    client.config.temperature = 99
+    req_module = install_requests_raw(monkeypatch, success_resp)
+    assert client._call_bedrock_sigv4("us-east-1", "AKID", "secret", "", "sys", "user") == "sigv4 ok"
+    sentinel_payload = json.loads(req_module._calls[0]["data"])
+    assert sentinel_payload["inferenceConfig"] == {"maxTokens": 256}
 
     # Session token is forwarded as x-amz-security-token
     req_module = install_requests_raw(monkeypatch, success_resp)
@@ -782,7 +820,7 @@ def test_call_bedrock_sigv4_success_and_errors(monkeypatch: pytest.MonkeyPatch)
         client._call_bedrock_sigv4("us-east-1", "AKID", "secret", "", "sys", "user")
 
     # Empty content → Unexpected response error
-    install_requests_raw(monkeypatch, FakeResponse(json_data={"content": []}))
+    install_requests_raw(monkeypatch, FakeResponse(json_data={"output": {"message": {"content": []}}}))
     with pytest.raises(LLMError, match="Unexpected Bedrock response"):
         client._call_bedrock_sigv4("us-east-1", "AKID", "secret", "", "sys", "user")