git-copilot-commit 0.6.1__py3-none-any.whl → 0.7.0__py3-none-any.whl

git_copilot_commit/cli.py

@@ -81,8 +81,8 @@ BaseUrlOption = Annotated[
     cyclopts.Parameter(
         name="--base-url",
         help=(
-            "Base URL for an OpenAI-compatible provider, for example "
-            "http://127.0.0.1:11434/v1."
+            "Endpoint URL for an OpenAI-compatible provider, for example "
+            "http://127.0.0.1:11434/v1/chat/completions."
         ),
     ),
 ]
@@ -371,6 +371,8 @@ def ask_llm_with_system_prompt(
     model: str | None = None,
     provider_config: providers.ProviderConfig | None = None,
     http_client_config: llm.HttpClientConfig | None = None,
+    disable_thinking: bool = True,
+    max_tokens: int = 1024,
 ) -> str:
     """Send a prepared prompt to the selected LLM provider."""
     return providers.ask(
@@ -386,6 +388,8 @@ def ask_llm_with_system_prompt(
         provider_config=provider_config,
         model=normalize_model_name(model),
         http_client_config=http_client_config,
+        disable_thinking=disable_thinking,
+        max_tokens=max_tokens,
     )
 
 
@@ -394,6 +398,8 @@ def generate_commit_message_for_prompt(
     model: str | None = None,
     provider_config: providers.ProviderConfig | None = None,
     http_client_config: llm.HttpClientConfig | None = None,
+    disable_thinking: bool = True,
+    max_tokens: int = 1024,
 ) -> str:
     """Generate a conventional commit message from a prepared prompt."""
     return ask_llm_with_system_prompt(
@@ -402,6 +408,8 @@ def generate_commit_message_for_prompt(
         model=model,
         provider_config=provider_config,
         http_client_config=http_client_config,
+        disable_thinking=disable_thinking,
+        max_tokens=max_tokens,
     )
 
 
@@ -435,6 +443,8 @@ def generate_commit_message_for_status(
     context: str = "",
     provider_config: providers.ProviderConfig | None = None,
     http_client_config: llm.HttpClientConfig | None = None,
+    disable_thinking: bool = True,
+    max_tokens: int = 1024,
 ) -> str:
     """Generate a commit message for a staged status snapshot."""
     full_prompt = build_commit_message_prompt(status, context=context)
@@ -444,6 +454,8 @@ def generate_commit_message_for_status(
             model=model,
             provider_config=provider_config,
             http_client_config=http_client_config,
+            disable_thinking=disable_thinking,
+            max_tokens=max_tokens,
         )
     except llm.LLMError as exc:
         if not should_retry_with_compact_prompt(exc):
@@ -462,6 +474,8 @@ def generate_commit_message_for_status(
         model=model,
         provider_config=provider_config,
         http_client_config=http_client_config,
+        disable_thinking=disable_thinking,
+        max_tokens=max_tokens,
     )
 
 
@@ -549,6 +563,8 @@ def request_commit_message(
     context: str = "",
     provider_config: providers.ProviderConfig | None = None,
     http_client_config: llm.HttpClientConfig | None = None,
+    disable_thinking: bool = True,
+    max_tokens: int = 1024,
 ) -> str:
     """Request a commit message for the provided staged state."""
     try:
@@ -561,6 +577,8 @@ def request_commit_message(
                 context=context,
                 provider_config=provider_config,
                 http_client_config=http_client_config,
+                disable_thinking=disable_thinking,
+                max_tokens=max_tokens,
             )
     except llm.LLMError as exc:
         print_llm_error("Could not generate a commit message", exc)
@@ -576,6 +594,8 @@ def request_split_commit_plan(
     context: str = "",
     provider_config: providers.ProviderConfig | None = None,
     http_client_config: llm.HttpClientConfig | None = None,
+    disable_thinking: bool = True,
+    max_tokens: int = 1024,
 ) -> SplitCommitPlan:
     """Request and validate a split-commit plan for the staged patch units."""
     planner_system_prompt = load_named_prompt(SPLIT_COMMIT_PLANNER_PROMPT_FILENAME)
@@ -596,6 +616,8 @@ def request_split_commit_plan(
                 model=model,
                 provider_config=provider_config,
                 http_client_config=http_client_config,
+                disable_thinking=disable_thinking,
+                max_tokens=max_tokens,
             )
     except llm.LLMError as exc:
         if not should_retry_with_compact_prompt(exc):
@@ -629,6 +651,8 @@ def request_split_commit_plan(
                 model=model,
                 provider_config=provider_config,
                 http_client_config=http_client_config,
+                disable_thinking=disable_thinking,
+                max_tokens=max_tokens,
             )
     except llm.LLMError as exc:
         print_llm_error("Could not generate a split commit plan", exc)
@@ -648,6 +672,8 @@ def request_split_commit_messages(
     context: str = "",
     provider_config: providers.ProviderConfig | None = None,
     http_client_config: llm.HttpClientConfig | None = None,
+    disable_thinking: bool = True,
+    max_tokens: int = 1024,
 ) -> list[PreparedSplitCommit]:
     """Generate commit messages for each planned split-commit group."""
     try:
@@ -665,6 +691,8 @@ def request_split_commit_messages(
                     context=context,
                     provider_config=provider_config,
                     http_client_config=http_client_config,
+                    disable_thinking=disable_thinking,
+                    max_tokens=max_tokens,
                 )
 
             prepared_commits.append(
@@ -858,6 +886,8 @@ def handle_single_commit_flow(
     context: str = "",
     provider_config: providers.ProviderConfig | None = None,
     http_client_config: llm.HttpClientConfig | None = None,
+    disable_thinking: bool = True,
+    max_tokens: int = 1024,
 ) -> None:
     """Generate, display, and execute the single-commit flow."""
     commit_message = request_commit_message(
@@ -866,6 +896,8 @@ def handle_single_commit_flow(
         context=context,
         provider_config=provider_config,
         http_client_config=http_client_config,
+        disable_thinking=disable_thinking,
+        max_tokens=max_tokens,
     )
     display_commit_message(commit_message)
 
@@ -883,6 +915,8 @@ def handle_split_commit_flow(
     context: str = "",
     provider_config: providers.ProviderConfig | None = None,
     http_client_config: llm.HttpClientConfig | None = None,
+    disable_thinking: bool = True,
+    max_tokens: int = 1024,
 ) -> None:
     """Generate, display, and execute the split-commit flow."""
     patch_units = tuple(
@@ -901,6 +935,8 @@ def handle_split_commit_flow(
             context=context,
             provider_config=provider_config,
             http_client_config=http_client_config,
+            disable_thinking=disable_thinking,
+            max_tokens=max_tokens,
         )
         return
 
@@ -916,6 +952,8 @@ def handle_split_commit_flow(
             context=context,
             provider_config=provider_config,
             http_client_config=http_client_config,
+            disable_thinking=disable_thinking,
+            max_tokens=max_tokens,
         )
         return
 
@@ -938,6 +976,8 @@ def handle_split_commit_flow(
             context=context,
             provider_config=provider_config,
             http_client_config=http_client_config,
+            disable_thinking=disable_thinking,
+            max_tokens=max_tokens,
         )
     except SplitPlanningError as exc:
         console.print(
@@ -952,6 +992,8 @@ def handle_split_commit_flow(
             context=context,
             provider_config=provider_config,
             http_client_config=http_client_config,
+            disable_thinking=disable_thinking,
+            max_tokens=max_tokens,
         )
         return
 
@@ -969,6 +1011,8 @@ def handle_split_commit_flow(
         context=context,
         provider_config=provider_config,
         http_client_config=http_client_config,
+        disable_thinking=disable_thinking,
+        max_tokens=max_tokens,
     )
     prepared_commits = order_prepared_split_commits(prepared_commits)
 
@@ -1008,7 +1052,7 @@ def authenticate(
     ] = False,
     ca_bundle: CaBundleOption = None,
     insecure: InsecureOption = False,
-    native_tls: NativeTlsOption = False,
+    native_tls: NativeTlsOption = True,
 ):
     """Authenticate with GitHub Copilot and cache credentials locally."""
     print_cli_banner()
@@ -1035,7 +1079,7 @@ def summary(
     api_key: ApiKeyOption = None,
     ca_bundle: CaBundleOption = None,
     insecure: InsecureOption = False,
-    native_tls: NativeTlsOption = False,
+    native_tls: NativeTlsOption = True,
 ):
     """Show the configured LLM provider summary."""
     print_cli_banner()
@@ -1073,7 +1117,7 @@ def models_command(
     ] = None,
     ca_bundle: CaBundleOption = None,
     insecure: InsecureOption = False,
-    native_tls: NativeTlsOption = False,
+    native_tls: NativeTlsOption = True,
 ):
     """List available models for the configured LLM provider."""
     print_cli_banner()
@@ -1139,12 +1183,30 @@ def commit(
             help="Optional user-provided context to guide commit message",
         ),
     ] = "",
+    disable_thinking: Annotated[
+        bool,
+        cyclopts.Parameter(
+            name="--disable-thinking",
+            negative="--enable-thinking",
+            help=(
+                "Disable or minimize reasoning/thinking tokens for commit-message requests."
+            ),
+        ),
+    ] = True,
+    max_tokens: Annotated[
+        int,
+        cyclopts.Parameter(
+            name="--max-tokens",
+            help=("Maximum output tokens for LLM generation."),
+            validator=cyclopts.validators.Number(gte=1),
+        ),
+    ] = 1024,
     provider: ProviderOption = None,
     base_url: BaseUrlOption = None,
     api_key: ApiKeyOption = None,
     ca_bundle: CaBundleOption = None,
     insecure: InsecureOption = False,
-    native_tls: NativeTlsOption = False,
+    native_tls: NativeTlsOption = True,
 ):
     """
     Generate commit message based on changes in the current git repository and commit them.
@@ -1212,6 +1274,8 @@ def commit(
             context=context,
             provider_config=provider_config,
             http_client_config=http_client_config,
+            disable_thinking=disable_thinking,
+            max_tokens=max_tokens,
         )
         return
 
@@ -1223,6 +1287,8 @@ def commit(
         context=context,
         provider_config=provider_config,
         http_client_config=http_client_config,
+        disable_thinking=disable_thinking,
+        max_tokens=max_tokens,
     )
 
 

git_copilot_commit/llms/copilot.py

@@ -500,7 +500,13 @@ def list_models(client, credentials: CopilotCredentials) -> list[Model]:
 
 
 def complete_text_prompt(
-    client, credentials: CopilotCredentials, *, model: Model, prompt: str
+    client,
+    credentials: CopilotCredentials,
+    *,
+    model: Model,
+    prompt: str,
+    disable_thinking: bool = False,
+    max_tokens: int | None = None,
 ) -> str:
     api_surface = llm.infer_api_surface(model)
     if api_surface == "chat_completions":
@@ -512,6 +518,8 @@ def complete_text_prompt(
             ),
             model_id=model.id,
             prompt=prompt,
+            disable_thinking=disable_thinking,
+            max_tokens=max_tokens,
         )
     if api_surface == "responses":
         return llm.responses_completion_request(
@@ -524,6 +532,8 @@ def complete_text_prompt(
             ),
             model_id=model.id,
             prompt=prompt,
+            disable_thinking=disable_thinking,
+            max_tokens=max_tokens,
         )
 
     raise LLMError(
@@ -778,6 +788,8 @@ def ask(
     default_model: str | None = None,
     configured_default_model_path: Path | None = None,
     http_client_config: HttpClientConfig | None = None,
+    disable_thinking: bool = False,
+    max_tokens: int | None = None,
 ) -> str:
     def run(client) -> str:
         credentials = ensure_fresh_credentials(client)
@@ -794,6 +806,8 @@ def ask(
             credentials,
             model=selected_model,
             prompt=prompt,
+            disable_thinking=disable_thinking,
+            max_tokens=max_tokens,
         )
 
     return _with_reauthentication(run, http_client_config=http_client_config)

git_copilot_commit/llms/core.py

@@ -142,7 +142,7 @@ class Model:
 
 @dataclass(frozen=True, slots=True)
 class HttpClientConfig:
-    native_tls: bool = False
+    native_tls: bool = True
     insecure: bool = False
     ca_bundle: str | None = None
 
@@ -441,6 +441,80 @@ def filter_models_by_vendor(
     return filtered
 
 
+def _model_id_matches(model_id: str, keywords: tuple[str, ...]) -> bool:
+    normalized = model_id.lower()
+    return any(keyword in normalized for keyword in keywords)
+
+
+def _is_openai_reasoning_model(model_id: str) -> bool:
+    normalized = model_id.lower()
+    return normalized.startswith(("o1", "o3", "o4")) or "/o" in normalized
+
+
+def _uses_chat_template_thinking_controls(model_id: str) -> bool:
+    return _model_id_matches(
+        model_id,
+        (
+            "qwen",
+            "deepseek",
+            "granite",
+            "glm",
+            "hunyuan",
+            "magistral",
+            "mistral",
+            "nemotron",
+            "seed",
+            "step",
+        ),
+    )
+
+
+def disable_thinking_options(
+    *,
+    model_id: str,
+    api_surface: str,
+) -> dict[str, Any]:
+    normalized = model_id.lower()
+
+    if api_surface == "responses":
+        if "codex" in normalized:
+            return {"reasoning": {"effort": "none"}}
+        if "gpt-5" in normalized:
+            return {"reasoning": {"effort": "minimal"}}
+        if "gpt-oss" in normalized or _is_openai_reasoning_model(model_id):
+            return {"reasoning": {"effort": "low"}}
+        if _uses_chat_template_thinking_controls(model_id):
+            return {
+                "reasoning_effort": "none",
+                "chat_template_kwargs": {
+                    "enable_thinking": False,
+                    "thinking": False,
+                },
+            }
+        return {}
+
+    if "gemini" in normalized:
+        return {"reasoning_effort": "none"}
+    if "codex" in normalized:
+        return {"reasoning_effort": "none"}
+    if "gpt-5" in normalized:
+        return {"reasoning_effort": "minimal"}
+    if "gpt-oss" in normalized or _is_openai_reasoning_model(model_id):
+        return {"reasoning_effort": "low"}
+    if "claude" in normalized or "anthropic" in normalized:
+        return {"thinking": {"type": "disabled"}}
+    if _uses_chat_template_thinking_controls(model_id):
+        return {
+            "reasoning_effort": "none",
+            "chat_template_kwargs": {
+                "enable_thinking": False,
+                "thinking": False,
+            },
+        }
+
+    return {}
+
+
 def extract_completion_text(payload: Any) -> str:
     if not isinstance(payload, dict):
         raise LLMError("Chat completion returned an invalid payload.")
@@ -459,7 +533,9 @@ def extract_completion_text(payload: Any) -> str:
 
     content = message.get("content")
     if isinstance(content, str):
-        return content.strip()
+        stripped = content.strip()
+        if stripped:
+            return stripped
 
     if isinstance(content, list):
         text_parts: list[str] = []
@@ -477,6 +553,41 @@ def extract_completion_text(payload: Any) -> str:
         if joined:
             return joined
 
+    finish_reason = choice.get("finish_reason")
+    reasoning = message.get("reasoning")
+    has_reasoning = isinstance(reasoning, str) and reasoning.strip()
+    finish_reason_detail: str | None = None
+    if finish_reason is not None:
+        try:
+            finish_reason_detail = json.dumps(finish_reason)
+        except TypeError:
+            finish_reason_detail = repr(finish_reason)
+        finish_reason_detail = truncate_response_detail(finish_reason_detail)
+
+    if finish_reason == "length":
+        detail = (
+            " The response contained reasoning text but no final assistant content."
+            if has_reasoning
+            else ""
+        )
+        raise LLMError(
+            "Chat completion reached the completion token limit before returning "
+            'message content (finish_reason="length").'
+            f"{detail} Increase `max_tokens` or reduce the prompt."
+        )
+
+    if finish_reason_detail is not None:
+        raise LLMError(
+            "Chat completion message content was empty "
+            f"(finish_reason={finish_reason_detail})."
+        )
+
+    if has_reasoning:
+        raise LLMError(
+            "Chat completion message content was empty. The response contained "
+            "reasoning text but no final assistant content."
+        )
+
     raise LLMError("Chat completion message content was empty.")
 
 
@@ -487,24 +598,35 @@ def chat_completion_request(
     *,
     model_id: str,
     prompt: str,
+    disable_thinking: bool = False,
+    max_tokens: int | None = None,
 ) -> str:
+    request_body: dict[str, Any] = {
+        "model": model_id,
+        "messages": [
+            {
+                "role": "user",
+                "content": prompt,
+            }
+        ],
+        "temperature": 0,
+        "max_tokens": max_tokens if max_tokens is not None else 1024,
+        "stream": False,
+    }
+    if disable_thinking:
+        request_body.update(
+            disable_thinking_options(
+                model_id=model_id,
+                api_surface="chat_completions",
+            )
+        )
+
     payload = request_json(
         client,
         "POST",
         url,
         headers=headers,
-        json_body={
-            "model": model_id,
-            "messages": [
-                {
-                    "role": "user",
-                    "content": prompt,
-                }
-            ],
-            "temperature": 0,
-            "max_tokens": 1024,
-            "stream": False,
-        },
+        json_body=request_body,
     )
     return extract_completion_text(payload)
 
@@ -551,6 +673,46 @@ def extract_response_text(payload: Any) -> str:
     raise LLMError("Responses API output did not contain text.")
 
 
+def response_output_contains_reasoning(payload: dict[str, Any]) -> bool:
+    output = payload.get("output")
+    if not isinstance(output, list):
+        return False
+
+    for item in output:
+        if not isinstance(item, dict):
+            continue
+        if item.get("type") == "reasoning":
+            return True
+        content = item.get("content")
+        if not isinstance(content, list):
+            continue
+        for block in content:
+            if isinstance(block, dict) and block.get("type") in {
+                "reasoning_text",
+                "reasoning_summary",
+            }:
+                return True
+
+    return False
+
+
+def format_incomplete_response_error(
+    *,
+    reason: str,
+    final_response: dict[str, Any],
+) -> str:
+    message = f"Responses API response was incomplete: {reason}."
+    if reason == "max_output_tokens":
+        message += " Increase `--max-tokens` or reduce the prompt."
+    if response_output_contains_reasoning(final_response):
+        message += (
+            " The response contained reasoning output before final text; if this "
+            "provider cannot disable reasoning on `/responses`, use its "
+            "`/chat/completions` endpoint instead."
+        )
+    return message
+
+
 def responses_completion_request(
     client: httpx.Client,
     url: str,
@@ -558,8 +720,10 @@ def responses_completion_request(
     *,
     model_id: str,
     prompt: str,
+    disable_thinking: bool = False,
+    max_tokens: int | None = None,
 ) -> str:
-    request_body = {
+    request_body: dict[str, Any] = {
         "model": model_id,
         "input": [
             {
@@ -575,6 +739,15 @@ def responses_completion_request(
         "stream": True,
         "store": False,
     }
+    if max_tokens is not None:
+        request_body["max_output_tokens"] = max_tokens
+    if disable_thinking:
+        request_body.update(
+            disable_thinking_options(
+                model_id=model_id,
+                api_surface="responses",
+            )
+        )
 
     for attempt in range(HTTP_RETRY_ATTEMPTS):
         text_parts: list[str] = []
@@ -697,7 +870,12 @@ def responses_completion_request(
                     reason = raw_reason.strip()
             if text:
                 return f"{text}\n\n[Response incomplete: {reason}]"
-            raise LLMError(f"Responses API response was incomplete: {reason}.")
+            raise LLMError(
+                format_incomplete_response_error(
+                    reason=reason,
+                    final_response=final_response,
+                )
+            )
 
         if text:
             return text

git_copilot_commit/llms/openai_api.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 from pathlib import Path
+from urllib.parse import urlparse
 
 from rich.console import Console
 from rich.panel import Panel
@@ -9,6 +10,9 @@ from rich.table import Table
 from . import core as llm
 
 DEFAULT_SUPPORTED_ENDPOINTS = ("/chat/completions",)
+CHAT_COMPLETIONS_ENDPOINT = "/chat/completions"
+RESPONSES_ENDPOINT = "/responses"
+MODELS_ENDPOINT = "/models"
 
 console = Console()
 
@@ -32,16 +36,51 @@ def request_headers(
     return headers
 
 
+def endpoint_kind(url: str) -> str | None:
+    path = urlparse(url).path.rstrip("/")
+    if path.endswith(CHAT_COMPLETIONS_ENDPOINT):
+        return "chat_completions"
+    if path.endswith(RESPONSES_ENDPOINT):
+        return "responses"
+    if path.endswith(MODELS_ENDPOINT):
+        return "models"
+    return None
+
+
+def supported_endpoints_for_url(url: str) -> tuple[str, ...]:
+    kind = endpoint_kind(url)
+    if kind == "chat_completions":
+        return (CHAT_COMPLETIONS_ENDPOINT,)
+    if kind == "responses":
+        return (RESPONSES_ENDPOINT,)
+    return DEFAULT_SUPPORTED_ENDPOINTS
+
+
+def completion_api_surface_from_url(url: str) -> str:
+    kind = endpoint_kind(url)
+    if kind in {"chat_completions", "responses"}:
+        return kind
+    raise LLMError(
+        "OpenAI-compatible generation URL must end with "
+        f"`{CHAT_COMPLETIONS_ENDPOINT}` or `{RESPONSES_ENDPOINT}`."
+    )
+
+
 def list_models(
     client,
     *,
     base_url: str,
     api_key: str | None = None,
 ) -> list[Model]:
+    if endpoint_kind(base_url) != "models":
+        raise LLMError(
+            f"OpenAI-compatible models URL must end with `{MODELS_ENDPOINT}`."
+        )
+
     payload = llm.request_json(
         client,
         "GET",
-        f"{base_url}/models",
+        base_url,
         headers=request_headers(api_key),
     )
 
@@ -95,10 +134,22 @@ def ensure_model_ready(
     http_client_config: HttpClientConfig | None = None,
 ) -> Model:
     if model is not None:
-        return default_model(model)
+        return default_model(
+            model,
+            supported_endpoints=supported_endpoints_for_url(base_url),
+        )
 
     if default_model_id is not None:
-        return default_model(default_model_id)
+        return default_model(
+            default_model_id,
+            supported_endpoints=supported_endpoints_for_url(base_url),
+        )
+
+    if endpoint_kind(base_url) != "models":
+        raise LLMError(
+            "OpenAI-compatible provider cannot choose a model automatically from "
+            "a generation URL. Pass `--model` or configure a default model."
+        )
 
     with llm.make_http_client(http_client_config) as client:
         models = list_models(client, base_url=base_url, api_key=api_key)
@@ -120,7 +171,10 @@ def ask(
     configured_default_model_path: Path | None = None,
     provider_label: str = "OpenAI-compatible provider",
     http_client_config: HttpClientConfig | None = None,
+    disable_thinking: bool = False,
+    max_tokens: int | None = None,
 ) -> str:
+    api_surface = completion_api_surface_from_url(base_url)
     selected_model = ensure_model_ready(
         base_url=base_url,
         api_key=api_key,
@@ -131,23 +185,26 @@ def ask(
         http_client_config=http_client_config,
     )
 
-    api_surface = llm.infer_api_surface(selected_model)
     with llm.make_http_client(http_client_config) as client:
         if api_surface == "responses":
             return llm.responses_completion_request(
                 client,
-                f"{base_url}/responses",
+                base_url,
                 request_headers(api_key, accept="text/event-stream"),
                 model_id=selected_model.id,
                 prompt=prompt,
+                disable_thinking=disable_thinking,
+                max_tokens=max_tokens,
             )
 
         return llm.chat_completion_request(
             client,
-            f"{base_url}/chat/completions",
+            base_url,
             request_headers(api_key),
             model_id=selected_model.id,
             prompt=prompt,
+            disable_thinking=disable_thinking,
+            max_tokens=max_tokens,
         )
 
 
@@ -212,7 +269,10 @@ def show_summary(
                 f"{selected_model.id} ({llm.infer_api_surface(selected_model)})",
             )
     elif default_model_id is not None:
-        table.add_row("Default model", f"{default_model_id} (chat_completions)")
+        table.add_row(
+            "Default model",
+            f"{default_model_id} ({endpoint_kind(base_url) or 'unknown endpoint'})",
+        )
 
     console.print(Panel.fit(table, title="LLM Summary"))
     if warning is not None:

git_copilot_commit/llms/providers.py

@@ -79,16 +79,11 @@ def normalize_openai_base_url(value: str | None) -> str | None:
         return None
 
     normalized = normalized.rstrip("/")
-    for suffix in ("/chat/completions", "/responses", "/models"):
-        if normalized.endswith(suffix):
-            normalized = normalized[: -len(suffix)]
-            break
-
     parsed = urlparse(normalized)
     if not parsed.scheme or not parsed.netloc:
         raise llm.LLMError(
-            "OpenAI-compatible base URL must include a scheme and host, for example "
-            "`http://127.0.0.1:11434/v1`."
+            "OpenAI-compatible URL must include a scheme and host, for example "
+            "`http://127.0.0.1:11434/v1/chat/completions`."
         )
 
     return normalized.rstrip("/")
@@ -227,8 +222,8 @@ def resolve_provider_config(
         resolved_base_url = normalize_openai_base_url(os.getenv("OPENAI_BASE_URL"))
     if resolved_base_url is None:
         raise llm.LLMError(
-            "OpenAI-compatible provider requires a base URL. Pass "
-            "`--base-url http://127.0.0.1:11434/v1` or set "
+            "OpenAI-compatible provider requires an endpoint URL. Pass "
+            "`--base-url http://127.0.0.1:11434/v1/chat/completions` or set "
             "`GIT_COPILOT_COMMIT_BASE_URL`."
         )
 
@@ -316,6 +311,8 @@ def ask(
     provider_config: ProviderConfig | None = None,
     model: str | None = None,
     http_client_config: llm.HttpClientConfig | None = None,
+    disable_thinking: bool = False,
+    max_tokens: int | None = None,
 ) -> str:
     resolved_provider = provider_config or resolve_provider_config()
     default_model, config_file = load_default_model()
@@ -327,6 +324,8 @@ def ask(
             default_model=default_model,
             configured_default_model_path=config_file,
             http_client_config=http_client_config,
+            disable_thinking=disable_thinking,
+            max_tokens=max_tokens,
         )
 
     if resolved_provider.base_url is None:
@@ -341,6 +340,8 @@ def ask(
         configured_default_model_path=config_file,
         provider_label=resolved_provider.display_name,
         http_client_config=http_client_config,
+        disable_thinking=disable_thinking,
+        max_tokens=max_tokens,
     )
 
 

{git_copilot_commit-0.6.1.dist-info → git_copilot_commit-0.7.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: git-copilot-commit
-Version: 0.6.1
+Version: 0.7.0
 Summary: Automatically generate and commit changes using GitHub Copilot or OpenAI-compatible LLMs
 Author-email: Dheepak Krishnamurthy <1813121+kdheepak@users.noreply.github.com>
 License-File: LICENSE
@@ -18,14 +18,16 @@ Description-Content-Type: text/markdown
 [![PyPI](https://img.shields.io/pypi/v/git-copilot-commit)](https://pypi.org/project/git-copilot-commit/)
 [![License](https://img.shields.io/github/license/kdheepak/git-copilot-commit)](https://github.com/kdheepak/git-copilot-commit/blob/main/LICENSE)
 
-AI-powered Git commit assistant that generates conventional commit messages using GitHub Copilot or any OpenAI-compatible LLM.
+AI-powered Git commit assistant that generates conventional commit messages using GitHub Copilot or
+any OpenAI-compatible LLM.
 
 ![Screenshot of git-copilot-commit in action](https://github.com/user-attachments/assets/6a6d70a6-6060-44e6-8cf4-a6532e9e9142)
 
 ## Features
 
 - Generates commit messages based on your staged changes
-- Supports GitHub Copilot and OpenAI-compatible `/v1/models` + `/v1/chat/completions` APIs
+- Supports GitHub Copilot and OpenAI-compatible `/v1/chat/completions`, `/v1/responses`,
+  and `/v1/models` endpoints
 - Supports multiple LLM models: GPT, Claude, Gemini, local models, and more
 - Allows editing of generated messages before committing
 - Follows the [Conventional Commits](https://www.conventionalcommits.org/) standard
@@ -51,8 +53,8 @@ You can run the latest version of tool directly every time by invoking this one
 uvx git-copilot-commit --help
 ```
 
-Alternatively, you can install the tool once into a global isolated environment
-and run `git-copilot-commit` to invoke it:
+Alternatively, you can install the tool once into a global isolated environment and run
+`git-copilot-commit` to invoke it:
 
 ```bash
 # Install into global isolated environment
@@ -96,39 +98,27 @@ git-copilot-commit --help
 
 ### OpenAI-compatible provider
 
-1. Point the CLI at your server.
-
-   The base URL can be either the provider root such as `http://127.0.0.1:11434/v1`
-   or the full chat completions endpoint such as
-   `http://127.0.0.1:11434/v1/chat/completions`.
+1. List models by pointing the CLI at your server's `/models` endpoint.
 
    ```bash
    uvx git-copilot-commit models \
      --provider openai \
-     --base-url http://127.0.0.1:11434/v1
+     --base-url http://127.0.0.1:11434/v1/models
    ```
 
-2. Generate and commit.
+2. Generate and commit by pointing the CLI at the generation endpoint you want to use.
 
    ```bash
    uvx git-copilot-commit commit \
      --provider openai \
-     --base-url http://127.0.0.1:11434/v1 \
+     --base-url http://127.0.0.1:11434/v1/chat/completions \
      --model your-model-id
    ```
 
    If your server requires an API key, also pass `--api-key ...` or set `OPENAI_API_KEY`.
 
-3. Example: use a self-hosted GPT-OSS model:
-
-   ```bash
-   uvx git-copilot-commit commit \
-     --provider openai \
-     --base-url http://example.com:8001/v1/chat/completions \
-     --model openai/gpt-oss-120b
-   ```
-
-   Model ids with slashes such as `openai/gpt-oss-120b` are supported.
+   OpenAI-compatible generation URLs must end with `/chat/completions` or `/responses`.
+   Model listing URLs must end with `/models`.
 
 ## Usage
 
@@ -137,28 +127,39 @@ git-copilot-commit --help
 ```bash
 $ uvx git-copilot-commit commit --help
 
- Usage: git-copilot-commit commit [OPTIONS]
+ Usage: git-copilot-commit commit [ARGS]
 
  Generate commit message based on changes in the current git repository and commit them.
 
-╭─ Options ────────────────────────────────────────────────────────────────────────────────────────────────╮
-│ --all         -a                               Stage all files before committing                         │
-│ --split                                        Split staged hunks into multiple commits automatically.   │
-│                                                Pass `--split=N` to express a preference for N commits.   │
-│ --model       -m                     MODEL_ID  Model to use for generating commit message                │
-│ --yes         -y                               Automatically accept the generated commit message         │
-│ --context     -c                     TEXT      Optional user-provided context to guide commit message    │
-│ --provider                           TEXT      LLM provider to use: copilot or openai                   │
-│ --base-url                           URL       Base URL for an OpenAI-compatible provider                │
-│ --api-key                            TEXT      API key for an OpenAI-compatible provider                 │
-│ --ca-bundle                          PATH      Path to a custom CA bundle (PEM)                          │
-│ --insecure                                     Disable SSL certificate verification.                     │
-│ --native-tls      --no-native-tls              Use the OS's native certificate store via 'truststore'    │
-│                                                for httpx instead of the Python bundle. Ignored if        │
-│                                                --ca-bundle or --insecure is used.                        │
-│                                                [default: no-native-tls]                                  │
-│ --help                                         Show this message and exit.                               │
-╰──────────────────────────────────────────────────────────────────────────────────────────────────────────╯
+╭─ Parameters ─────────────────────────────────────────────────────────────────╮
+│ ALL --all -a --no-all     Stage all files before committing [default: False] │
+│ SPLIT --split --no-split  Split staged hunks into multiple commits           │
+│                           automatically. Pass --split=N to express a         │
+│                           preference for N commits. [default: False]         │
+│ MODEL --model -m          Model to use for generating commit message         │
+│ YES --yes -y --no-yes     Automatically accept the generated commit message  │
+│                           [default: False]                                   │
+│ CONTEXT --context -c      Optional user-provided context to guide commit     │
+│                           message [default: ""]                              │
+│ DISABLE-THINKING          Disable or minimize reasoning/thinking tokens for  │
+│   --disable-thinking      commit-message requests. [default: True]           │
+│   --enable-thinking                                                          │
+│ MAX-TOKENS --max-tokens   Maximum output tokens for LLM generation.          │
+│                           [default: 1024]                                    │
+│ PROVIDER --provider       LLM provider to use: copilot or openai.            │
+│ BASE-URL --base-url       Endpoint URL for an OpenAI-compatible provider,    │
+│                           for example                                        │
+│                           http://127.0.0.1:11434/v1/chat/completions.        │
+│ API-KEY --api-key         API key for an OpenAI-compatible provider. Omit    │
+│                           when the server does not require one.              │
+│ CA-BUNDLE --ca-bundle     Path to a custom CA bundle (PEM)                   │
+│ INSECURE --insecure       Disable SSL certificate verification. [default:    │
+│   --no-insecure           False]                                             │
+│ NATIVE-TLS --native-tls   Use the OS's native certificate store via          │
+│   --no-native-tls         'truststore' for httpx instead of the Python       │
+│                           bundle. Ignored if --ca-bundle or --insecure is    │
+│                           used. [default: True]                              │
+╰──────────────────────────────────────────────────────────────────────────────╯
 ```
 
 ## Examples
@@ -186,17 +187,51 @@ Use a local OpenAI-compatible server:
 ```bash
 uvx git-copilot-commit commit \
   --provider openai \
-  --base-url http://127.0.0.1:11434/v1 \
+  --base-url http://127.0.0.1:11434/v1/chat/completions \
   --model your-model-id
 ```
 
-Use a self-hosted GPT-OSS endpoint:
+Example with `openai/gpt-oss-120b` and `Qwen/Qwen3.6-35B-A3B`:
 
 ```bash
 uvx git-copilot-commit commit \
   --provider openai \
   --base-url http://example.com:8001/v1/chat/completions \
   --model openai/gpt-oss-120b
+
+uvx git-copilot-commit commit \
+  --provider openai \
+  --base-url http://example.com:8002/v1/chat/completions \
+  --model Qwen/Qwen3.6-35B-A3B
+```
+
+Use the Responses API endpoint:
+
+```bash
+uvx git-copilot-commit commit \
+  --provider openai \
+  --base-url http://example.com:8002/v1/responses \
+  --model your-model-id
+```
+
+Increase the output token budget:
+
+```bash
+uvx git-copilot-commit commit --max-tokens 4096
+```
+
+Thinking/reasoning is disabled or minimized by default for commit-message requests. To let the
+selected model use its default thinking behavior, pass:
+
+```bash
+uvx git-copilot-commit commit --enable-thinking
+```
+
+TLS uses the operating system's native certificate store by default. To use Python's default
+certificate bundle instead, pass:
+
+```bash
+uvx git-copilot-commit commit --no-native-tls
 ```
 
 Split staged hunks into separate commits:
@@ -246,8 +281,8 @@ Now you can run to review the message before committing:
 git ai-commit
 ```
 
-Alternatively, you can stage all files and auto accept the commit message and
-specify which model should be used to generate the commit in one CLI invocation.
+Alternatively, you can stage all files and auto accept the commit message and specify which model
+should be used to generate the commit in one CLI invocation.
 
 ```bash
 git ai-commit --all --yes --model claude-3.5-sonnet
@@ -257,7 +292,7 @@ You can also set provider defaults with environment variables:
 
 ```bash
 export GIT_COPILOT_COMMIT_PROVIDER=openai
-export GIT_COPILOT_COMMIT_BASE_URL=http://127.0.0.1:11434/v1
+export GIT_COPILOT_COMMIT_BASE_URL=http://127.0.0.1:11434/v1/chat/completions
 export GIT_COPILOT_COMMIT_API_KEY=...
 export OPENAI_API_KEY=...
 git ai-commit --provider openai --model your-model-id
@@ -267,7 +302,7 @@ For example:
 
 ```bash
 export GIT_COPILOT_COMMIT_PROVIDER=openai
-export GIT_COPILOT_COMMIT_BASE_URL=http://example.com:8001/v1
+export GIT_COPILOT_COMMIT_BASE_URL=http://example.com:8001/v1/chat/completions
 git ai-commit --model openai/gpt-oss-120b
 ```
 
@@ -279,5 +314,5 @@ git ai-commit --model openai/gpt-oss-120b
 > git config --global diff.context 3
 > ```
 >
-> This may be useful because this tool sends the diffs with surrounding context
-> to the LLM for generating a commit message
+> This may be useful because this tool sends the diffs with surrounding context to the LLM for
+> generating a commit message

{git_copilot_commit-0.6.1.dist-info → git_copilot_commit-0.7.0.dist-info}/RECORD

@@ -1,19 +1,19 @@
 git_copilot_commit/__init__.py,sha256=v3x5oBkxwKJEZLv62QqSmP3iqNKLtZgrWZfH8eFzlQg,60
-git_copilot_commit/cli.py,sha256=x5p_f71DhnYEED8-9rvJKqkjTbcwfXlN3zIvfxmYQfU,38011
+git_copilot_commit/cli.py,sha256=ZlrXY6c4JUxMqLEkykhnNTgoegOfxLq9Pp7Yqy1PuOk,40391
 git_copilot_commit/git.py,sha256=EbXiicWygSlMM-F6rY4LCkchwCvsFTziJcdUZM-1vnw,21059
 git_copilot_commit/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 git_copilot_commit/settings.py,sha256=WrM10_J3F7QBfOVmPDWpNZrNHhmZSeN-9FqQZxgdWvQ,3730
 git_copilot_commit/split_commits.py,sha256=rHyuVJggjmYjbva7BVqsM3aZRxUgOKkuZtxxvFRcu6Q,15060
 git_copilot_commit/version.py,sha256=AieHOUX52g6N67HL0iLWtDKrgOYyulxwHWViu26Jrd4,105
 git_copilot_commit/llms/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-git_copilot_commit/llms/copilot.py,sha256=O2jhdhWhsLdjx0LgU30_JhReyfR9yimfhs_KIfddNSY,25010
-git_copilot_commit/llms/core.py,sha256=PGYa2Znsu27juK5lCKcrB_GPY1jpLXfVgLrDnR15JR4,25586
-git_copilot_commit/llms/openai_api.py,sha256=wkadrdSDadbLRaLWEpOhsYYjrbtEYg14CFbkXKYQxM0,6429
-git_copilot_commit/llms/providers.py,sha256=rA2mdCQR8pfVDhwV5mqpdHlT1nxkWtowQ1Smt0zXCa0,11565
+git_copilot_commit/llms/copilot.py,sha256=_RK4jpjziXeTrpcSA-Kaj1AT-18y4Nztu1Kqdkh95Bs,25415
+git_copilot_commit/llms/core.py,sha256=4JseguAA17RIPi59nXcCYn1DP3duw2CZpjKrZKsLMsk,31182
+git_copilot_commit/llms/openai_api.py,sha256=--6RVwxV2aGsJT2paJzv86CihoPczpfEVFA4P5_kSSg,8325
+git_copilot_commit/llms/providers.py,sha256=N0DLkfRP-Z85kcYYoXWWHXr5nIwXUXDV88skS-NXLpU,11649
 git_copilot_commit/prompts/commit-message-generator-prompt.md,sha256=3Dz8GCdumFNAtXOdTlpRtgBnmX0WyrPL6tdfMgNyYiE,2411
 git_copilot_commit/prompts/split-commit-planner-prompt.md,sha256=tDI0v1udOhkRQM31M892FMzcPMYHExnU0fjTGia1V2k,1510
-git_copilot_commit-0.6.1.dist-info/METADATA,sha256=3pBOX2m_9i-UKx-dQUu5fwr2gA638RePMzEBe8rf4E4,8986
-git_copilot_commit-0.6.1.dist-info/WHEEL,sha256=QccIxa26bgl1E6uMy58deGWi-0aeIkkangHcxk2kWfw,87
-git_copilot_commit-0.6.1.dist-info/entry_points.txt,sha256=-D4bQqiuSPwQJG2zx--vJbZD1iqB5coUfoJ_gmC3rSg,66
-git_copilot_commit-0.6.1.dist-info/licenses/LICENSE,sha256=14lNZAoKJPI1U7eGpletjN_PFm1JwP1vT_0jFKY6eWg,1065
-git_copilot_commit-0.6.1.dist-info/RECORD,,
+git_copilot_commit-0.7.0.dist-info/METADATA,sha256=lI1E3-S5S1HEsTFOaqPuXwPE0WCF2kNx_mrWZFNReyg,10006
+git_copilot_commit-0.7.0.dist-info/WHEEL,sha256=QccIxa26bgl1E6uMy58deGWi-0aeIkkangHcxk2kWfw,87
+git_copilot_commit-0.7.0.dist-info/entry_points.txt,sha256=-D4bQqiuSPwQJG2zx--vJbZD1iqB5coUfoJ_gmC3rSg,66
+git_copilot_commit-0.7.0.dist-info/licenses/LICENSE,sha256=14lNZAoKJPI1U7eGpletjN_PFm1JwP1vT_0jFKY6eWg,1065
+git_copilot_commit-0.7.0.dist-info/RECORD,,