PyPI - python-codex - Versions diffs - 0.2.4__tar.gz → 0.2.6__tar.gz - Mend

python-codex 0.2.4tar.gz → 0.2.6tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Files changed (128) hide show

{python_codex-0.2.4 → python_codex-0.2.6}/.github/workflows/publish.yml RENAMED Viewed

@@ -107,8 +107,10 @@ jobs:
       - name: Build distributions
         run: |
-          python -m pip install --upgrade build
-          python -m build
+          python -m pip install --upgrade build twine
+          python -m build --outdir dist/python-codex .
+          python -m build --outdir dist/pycodex-ws packages/pycodex-ws
+          python -m twine check dist/python-codex/* dist/pycodex-ws/*
       - name: Upload distributions
         uses: actions/upload-artifact@v4
@@ -124,7 +126,7 @@ jobs:
       id-token: write
     environment:
       name: testpypi
-      url: https://test.pypi.org/project/pycodex/
+      url: https://test.pypi.org/project/python-codex/
     steps:
       - name: Download distributions
         uses: actions/download-artifact@v4
@@ -135,6 +137,7 @@ jobs:
       - name: Publish to TestPyPI
         uses: pypa/gh-action-pypi-publish@release/v1
         with:
+          packages-dir: dist/python-codex/
           repository-url: https://test.pypi.org/legacy/
   publish-pypi:
@@ -145,7 +148,7 @@ jobs:
       id-token: write
     environment:
       name: pypi
-      url: https://pypi.org/project/pycodex/
+      url: https://pypi.org/project/python-codex/
     steps:
       - name: Download distributions
         uses: actions/download-artifact@v4
@@ -155,3 +158,28 @@ jobs:
       - name: Publish to PyPI
         uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          packages-dir: dist/python-codex/
+  publish-pycodex-ws:
+    if: github.event_name == 'release' || github.event_name == 'push' || (github.event_name == 'workflow_dispatch' && inputs.repository == 'pypi')
+    needs:
+      - build
+      - publish-pypi
+    runs-on: ubuntu-latest
+    permissions:
+      id-token: write
+    environment:
+      name: pypi
+      url: https://pypi.org/project/pycodex-ws/
+    steps:
+      - name: Download distributions
+        uses: actions/download-artifact@v4
+        with:
+          name: python-package-distributions
+          path: dist/
+      - name: Publish pycodex-ws to PyPI
+        uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          packages-dir: dist/pycodex-ws/

{python_codex-0.2.4 → python_codex-0.2.6}/AGENTS.md RENAMED Viewed

@@ -3,7 +3,9 @@
 ## Repo Priors
 - 这个仓库的目标是尽可能准确地复现上游 Codex（`https://github.com/openai/codex`），不是发明一个“类似 Codex”的简化替代品；凡是行为差异都应优先朝原版 Codex 收敛。
-- 发布到 PyPI 时，distribution name 用 `python-codex`；代码包目录、`import` 路径和 CLI 命令仍保持 `pycodex`。
+- PyPI 主 distribution name 用 `python-codex`；代码包目录、`import` 路径和 CLI 命令仍保持 `pycodex`。
+- `pycodex-ws` 是与 `python-codex` 锁步同版本的薄 metapackage，只依赖精确同版本的 `python-codex`；不要在两个 wheel 中重复打包 `pycodex/`、`workspace_server/` 或重复声明 `pycodex-ws` console script。
+- 双包发布时先发布 `python-codex`，成功后再发布 `pycodex-ws`，避免新 metapackage 短暂指向尚不存在的核心版本。
 - 包结构直接放在 repo 根目录下的 `pycodex/`，不使用 `src/` layout。
 - Rust 到 Python 的主映射：`submission_loop` -> `pycodex/runtime.py`，`run_turn` / `run_sampling_request` -> `pycodex/agent.py`，`ToolRouter` -> `pycodex/tools/base_tool.py` + `pycodex/tools/`.
 - 会话协议先收敛到 4 类 item：`UserMessage`、`AssistantMessage`、`ToolCall`、`ToolResult`；只有这层稳定后再扩 richer event model。
@@ -19,6 +21,7 @@
 - 对 vLLM chat-completions 打开 `return_token_ids=true` 时，streaming `prompt_token_ids` 只出现在首个 chunk，后续每个 chunk 的 `choices[*].token_ids` 都是 decode delta；要在 `responses_server` 侧导出 trajectory 时，按“首个 `prompt_token_ids` + 按序拼接所有 chunk 的 `token_ids`”重建即可。
 - `pycodex` 默认是最小交互 CLI；无 prompt 时进入 REPL，并通过 `AgentRuntime` 跑外层提交循环。当前会显示最小事件流、assistant 流式输出、简单 title/history（`/title`, `/history`），并默认注册一组与原版一一对应的本地工具子集。
 - Web workspace lives in the standalone `workspace_server/` package and is launched with `pycodex-ws --listen <host:port> --board <html>`, not through `pycodex` CLI dispatch. CLI and web share `pycodex.interactive_session.run_interactive_session`; slash-command semantics such as `/resume`, `/compact`, `/model`, and `/link` belong to that shared interactive shell loop, while workspace only supplies a web view/input adapter and tab/session lifecycle.
+- Relative images in board HTML resolve to HTTP siblings such as `/w/<workspace-id>/<image>` (or `/<image>` in single-workspace mode), not to `/board/...`. Keep the image fallback routes after concrete workspace/API routes, resolve files from the board's directory, and reject non-`image/*` files or paths that escape that directory.
 - 交互 CLI 的事件流展示优先表达用户可感知的阶段（例如工具开始/完成、模型回看工具结果），不要直接把内部 `iteration` 计数暴露成主要状态文案；`iterations` 应继续保留在 `TurnResult` 等程序化结果里。
 - 在交互 CLI 里，`stream_error` 表示当前 Responses stream attempt 失败且模型客户端可能马上自动重试；不要在这个事件上 `finish_stream()` 输出当前 assistant delta buffer，否则第一次失败 attempt 的文本和重试成功后的最终回复会重复显示。真正 fatal 的失败仍由 `turn_failed` 走通用 flush，保留 partial 输出。
 - prompt/context 相关逻辑统一放在 `pycodex/context.py`：`AgentLoop` 只维护真实会话历史；每轮请求前由 `ContextManager` 注入 base instructions、developer message、`AGENTS.md` 指令和 `<environment_context>`，且这些注入项不写回 history。
@@ -61,6 +64,7 @@
 - auto-compact 对齐上游配置名 `model_auto_compact_token_limit`；为空时关闭，触发依据是最近一次模型上报的 `usage.total_tokens`，pre-turn 压缩上一轮历史，mid-turn 压缩工具 follow-up 前的当前历史，并继续复用现有 compacted rollout 记录。
 - Responses streaming 里的 `response.incomplete` 不是连接断开：不要让 `ResponsesModelClient` 把它当 retryable incomplete stream 反复重连。普通 turn 应明确报 `response.incomplete`；compact 请求如果已经收到 assistant partial summary，可以用这个 partial summary 完成 replacement history，避免 midturn auto-compact 卡在 5 次 retry。
 - 上游 Codex Responses 请求当前不传模型级 `max_output_tokens`，也没有读取 `model_max_output_tokens` 这个 config key；这个名字在上游主要用于工具输出截断，不要为了上游对齐把它加进模型请求。
+- `service_tier = "fast"` 在 config 中是 Fast mode 的持久写法，但 Responses wire 值是 `priority`；只对模型 metadata 的 `service_tiers` 明确包含该 id 的模型发送，`default` 和不支持的 tier 都省略。
 - 普通 turn 遇到 `ResponsesIncompleteError(reason="max_output_tokens")` 时，上游语义是保留异常前已经收到的 `response.output_item.done`；pycodex 因为模型客户端按整轮返回，需要在异常路径把这些 done assistant/reasoning items 写入 history 和 rollout，才能让用户下一句 `continue` 接上。不要把纯 `response.output_text.delta` 合成 history，也不要持久化没有 tool result 的 tool call。
 - Feishu card tests read `~/.codex/.feishu_refresh_token` through production code; when running `tests/test_feishu_card.py` locally, isolate HOME (for example `HOME=/tmp/pycodex-empty-home env -u VIRTUAL_ENV uv run pytest tests/test_feishu_card.py tests/test_feishu_link.py`) unless the test itself controls `HOME`.
 - `lark_oapi.ws.client` creates a module-level asyncio `loop` at import time and `Client.start()` always uses that global. For `/link` long-connection listeners, bind that SDK global to a listener-thread-owned loop before constructing the client, and stop it through private `_disconnect()` plus `loop.stop()` on `/unlink`; otherwise unlink/link can reuse a still-running SDK loop and fail with `RuntimeError: This event loop is already running`.

{python_codex-0.2.4 → python_codex-0.2.6}/PKG-INFO RENAMED Viewed

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: python-codex
-Version: 0.2.4
+Version: 0.2.6
 Summary: A minimal Python extraction of Codex's main agent loop
 License-File: LICENSE
 Requires-Python: >=3.6.2
@@ -23,8 +23,17 @@ Description-Content-Type: text/markdown
 English README. Chinese version: `README_ZH.md`
-PyPI distribution name: `python-codex`
-Import path and CLI command remain `pycodex`.
+PyPI distributions:
+- Primary package: `python-codex`
+- Workspace install alias: `pycodex-ws`
+The import path remains `pycodex`; the CLI commands are `pycodex` and
+`pycodex-ws`.
+`pycodex-ws` is a thin metapackage that depends on the exact matching
+`python-codex` version. The implementation and console script remain owned by
+`python-codex`, so the two distributions never install duplicate modules.
 This repository extracts the core Codex agent loop from upstream Codex
 (`https://github.com/openai/codex`) into a deliberately small Python version,
@@ -44,6 +53,13 @@ Relevant Rust reference points:
 ## Quick Start
+Install the full package or the workspace-oriented alias:
+```bash
+pip install python-codex
+pip install pycodex-ws
+```
 Install dependencies first:
 ```bash
@@ -179,7 +195,8 @@ Current behavior:
 - interactive mode shows a compact event stream for user-visible phases such as
   tool execution and model follow-up after tool results
 - assistant text is printed from streaming deltas directly
-- interactive mode supports `/history`, `/title`, `/model`, `/resume`, and `/compact`
+- interactive mode supports `/history`, `/title`, `/model`, `/resume`, `/compact`,
+  and `/fork`
 - `/model <name>` switches the model used by later turns in the current
   interactive session; `/model` shows the current model and available choices
 - `/resume` with no argument lists the currently resumable sessions by their
@@ -189,9 +206,15 @@ Current behavior:
 - `/compact` synthesizes a local handoff summary, replaces the in-memory
   conversation history with the compacted view, and appends a compacted-history
   entry to the rollout so later `/resume` sees the same state
+- `/fork` generates a new model session id while preserving the current history,
+  rollout, and workspace tab
 - `model_auto_compact_token_limit = <tokens>` in `config.toml` enables the same
   compaction path automatically when the latest reported usage reaches that
   threshold before a follow-up sampling request or the next user turn
+- `service_tier = "fast"` enables Fast mode for models whose vendored metadata
+  advertises the `priority` service tier; pycodex follows upstream Codex by
+  sending `service_tier = "priority"` on the Responses request, while
+  `service_tier = "default"` or unsupported tiers are omitted
 - if a model request fails with `context_length_exceeded`, pycodex now treats
   the provider-reported requested token count as a failed-request usage sample,
   triggers the same compact path immediately, and retries the request once; if
@@ -219,8 +242,13 @@ Current behavior:
   `delete(name)` controls; omitted names become `workspace-1`, `workspace-2`,
   etc. If `board` is omitted when adding a workspace, pycodex assigns a random
   writable `/tmp/pcws-*.html` board path. Add/delete actions and later
-  session-state saves refresh the JSON file. `--password <value>` enables a
-  password-only login page for workspace pages, APIs, and websocket connections.
+  session-state saves refresh the JSON file. Board HTML can reference local
+  images beside the board (including nested paths) with relative URLs; only
+  `image/*` files contained by the board directory are served.
+  Assistant Markdown supports KaTeX formulas with `$...$`, `$$...$$`,
+  `\(...\)`, and `\[...\]` delimiters.
+  `--password <value>` enables a password-only login page for workspace pages,
+  APIs, and websocket connections.
 - steer is enabled by default in interactive mode: normal input goes into the
   runtime steer path, the current request stops at the next safe boundary, and
   later steer text is appended to the next model request's `input` in order;

{python_codex-0.2.4 → python_codex-0.2.6}/README.md RENAMED Viewed

@@ -2,8 +2,17 @@
 English README. Chinese version: `README_ZH.md`
-PyPI distribution name: `python-codex`
-Import path and CLI command remain `pycodex`.
+PyPI distributions:
+- Primary package: `python-codex`
+- Workspace install alias: `pycodex-ws`
+The import path remains `pycodex`; the CLI commands are `pycodex` and
+`pycodex-ws`.
+`pycodex-ws` is a thin metapackage that depends on the exact matching
+`python-codex` version. The implementation and console script remain owned by
+`python-codex`, so the two distributions never install duplicate modules.
 This repository extracts the core Codex agent loop from upstream Codex
 (`https://github.com/openai/codex`) into a deliberately small Python version,
@@ -23,6 +32,13 @@ Relevant Rust reference points:
 ## Quick Start
+Install the full package or the workspace-oriented alias:
+```bash
+pip install python-codex
+pip install pycodex-ws
+```
 Install dependencies first:
 ```bash
@@ -158,7 +174,8 @@ Current behavior:
 - interactive mode shows a compact event stream for user-visible phases such as
   tool execution and model follow-up after tool results
 - assistant text is printed from streaming deltas directly
-- interactive mode supports `/history`, `/title`, `/model`, `/resume`, and `/compact`
+- interactive mode supports `/history`, `/title`, `/model`, `/resume`, `/compact`,
+  and `/fork`
 - `/model <name>` switches the model used by later turns in the current
   interactive session; `/model` shows the current model and available choices
 - `/resume` with no argument lists the currently resumable sessions by their
@@ -168,9 +185,15 @@ Current behavior:
 - `/compact` synthesizes a local handoff summary, replaces the in-memory
   conversation history with the compacted view, and appends a compacted-history
   entry to the rollout so later `/resume` sees the same state
+- `/fork` generates a new model session id while preserving the current history,
+  rollout, and workspace tab
 - `model_auto_compact_token_limit = <tokens>` in `config.toml` enables the same
   compaction path automatically when the latest reported usage reaches that
   threshold before a follow-up sampling request or the next user turn
+- `service_tier = "fast"` enables Fast mode for models whose vendored metadata
+  advertises the `priority` service tier; pycodex follows upstream Codex by
+  sending `service_tier = "priority"` on the Responses request, while
+  `service_tier = "default"` or unsupported tiers are omitted
 - if a model request fails with `context_length_exceeded`, pycodex now treats
   the provider-reported requested token count as a failed-request usage sample,
   triggers the same compact path immediately, and retries the request once; if
@@ -198,8 +221,13 @@ Current behavior:
   `delete(name)` controls; omitted names become `workspace-1`, `workspace-2`,
   etc. If `board` is omitted when adding a workspace, pycodex assigns a random
   writable `/tmp/pcws-*.html` board path. Add/delete actions and later
-  session-state saves refresh the JSON file. `--password <value>` enables a
-  password-only login page for workspace pages, APIs, and websocket connections.
+  session-state saves refresh the JSON file. Board HTML can reference local
+  images beside the board (including nested paths) with relative URLs; only
+  `image/*` files contained by the board directory are served.
+  Assistant Markdown supports KaTeX formulas with `$...$`, `$$...$$`,
+  `\(...\)`, and `\[...\]` delimiters.
+  `--password <value>` enables a password-only login page for workspace pages,
+  APIs, and websocket connections.
 - steer is enabled by default in interactive mode: normal input goes into the
   runtime steer path, the current request stops at the next safe boundary, and
   later steer text is appended to the next model request's `input` in order;

{python_codex-0.2.4 → python_codex-0.2.6}/README_ZH.md RENAMED Viewed

@@ -2,8 +2,16 @@
 中文 README。English version: `README.md`
-PyPI distribution name: `python-codex`
-Import path and CLI command remain `pycodex`.
+PyPI distributions：
+- 主包：`python-codex`
+- Workspace 安装别名：`pycodex-ws`
+Import 路径仍保持 `pycodex`；CLI 命令为 `pycodex` 和 `pycodex-ws`。
+`pycodex-ws` 是一个薄 metapackage，精确依赖同版本的 `python-codex`。
+实际代码和 console script 继续由 `python-codex` 提供，两个 distribution
+不会重复安装同名模块。
 这个仓库把上游 Codex（`https://github.com/openai/codex`）里最核心的 agent
 闭环先抽成一个尽量小的 Python 版本，重点保留两层结构：
@@ -21,6 +29,13 @@ Import path and CLI command remain `pycodex`.
 ## 快速开始
+安装完整包或 workspace 安装别名：
+```bash
+pip install python-codex
+pip install pycodex-ws
+```
 先安装开发依赖：
 ```bash

python_codex-0.2.6/packages/pycodex-ws/README.md ADDED Viewed

@@ -0,0 +1,13 @@
+# pycodex-ws
+`pycodex-ws` is the workspace-oriented installation name for
+[`python-codex`](https://pypi.org/project/python-codex/).
+```bash
+pip install pycodex-ws
+pycodex-ws --help
+```
+This distribution intentionally contains only package metadata. It depends on
+the exact matching `python-codex` version, which provides the `workspace_server`
+implementation and the `pycodex-ws` console script.

python_codex-0.2.6/packages/pycodex-ws/pycodex_ws_meta/__init__.py ADDED Viewed

	@@ -0,0 +1 @@
1	+ """Metadata-only package for the pycodex-ws distribution."""

python_codex-0.2.6/packages/pycodex-ws/pyproject.toml ADDED Viewed

@@ -0,0 +1,16 @@
+[build-system]
+requires = ["hatchling>=1.17.1,<1.18"]
+build-backend = "hatchling.build"
+[project]
+name = "pycodex-ws"
+version = "0.2.6"
+description = "Browser workspace install alias for python-codex"
+readme = "README.md"
+requires-python = ">=3.6.2"
+dependencies = [
+    "python-codex==0.2.6",
+]
+[tool.hatch.build.targets.wheel]
+packages = ["pycodex_ws_meta"]

{python_codex-0.2.4 → python_codex-0.2.6}/pycodex/context.py RENAMED Viewed

@@ -1,8 +1,6 @@
 from dataclasses import dataclass
 from datetime import datetime
-from functools import lru_cache
-import json
 from pathlib import Path
 import typing
@@ -12,6 +10,7 @@ except ModuleNotFoundError:  # pragma: no cover - Python 3.10 path
     import tomli as tomllib
 from .collaboration import DEFAULT_COLLABORATION_MODE, CollaborationMode
+from .model_metadata import load_models_by_slug
 from .protocol import ContextMessage, ConversationItem, JSONDict, Prompt, ToolSpec
 from .utils.get_env import (
     get_sandbox_tag,
@@ -23,7 +22,6 @@ from .utils.get_env import (
 DEFAULT_BASE_INSTRUCTIONS_PATH = (
     Path(__file__).resolve().parent / "prompts" / "default_base_instructions.md"
 )
-DEFAULT_MODELS_PATH = Path(__file__).resolve().parent / "prompts" / "models.json"
 DEFAULT_COLLABORATION_INSTRUCTIONS_PATH = (
     Path(__file__).resolve().parent / "prompts" / "collaboration_default.md"
 )
@@ -270,7 +268,7 @@ class ContextManager:
         model_metadata = None
         model_slug = self._config.model
         if model_slug is not None:
-            model_metadata = _load_models_by_slug().get(model_slug)
+            model_metadata = load_models_by_slug().get(model_slug)
         context_window = self._config.model_context_window
         if context_window is None and model_metadata is not None:
@@ -293,7 +291,7 @@ class ContextManager:
         model_slug = self._config.model
         if model_slug is None:
             return None
-        model_metadata = _load_models_by_slug().get(model_slug)
+        model_metadata = load_models_by_slug().get(model_slug)
         if model_metadata is None:
             return None
         return _normalize_int(model_metadata.get("auto_compact_token_limit"))
@@ -302,7 +300,7 @@ class ContextManager:
         model_slug = self._config.model
         if model_slug is None:
             return None
-        model_metadata = _load_models_by_slug().get(model_slug)
+        model_metadata = load_models_by_slug().get(model_slug)
         if model_metadata is None:
             return None
@@ -560,18 +558,6 @@ def _read_first_instruction_file(base: 'Path') -> 'typing.Union[str, None]':
     return None
-@lru_cache(maxsize=1)
-def _load_models_by_slug() -> 'typing.Dict[str, JSONDict]':
-    payload = json.loads(DEFAULT_MODELS_PATH.read_text(encoding="utf-8"))
-    models = payload.get("models", [])
-    by_slug: 'typing.Dict[str, JSONDict]' = {}
-    for model in models:
-        slug = model.get("slug")
-        if isinstance(slug, str):
-            by_slug[slug] = model
-    return by_slug
 def _resolve_personality_message(variables, personality: 'typing.Union[str, None]') -> 'str':
     if not isinstance(variables, dict):
         return ""

{python_codex-0.2.4 → python_codex-0.2.6}/pycodex/interactive_session.py RENAMED Viewed

@@ -25,10 +25,11 @@ MODEL_COMMAND = "/model"
 QUEUE_COMMAND = "/queue"
 RESUME_COMMAND = "/resume"
 COMPACT_COMMAND = "/compact"
+FORK_COMMAND = "/fork"
 LINK_COMMAND = "/link"
 UNLINK_COMMAND = "/unlink"
 EXTRA_COMMANDS_LINE = (
-    "Extra commands: /help, /history, /title, /model, /resume, /compact, /link, /unlink"
+    "Extra commands: /help, /history, /title, /model, /resume, /compact, /fork, /link, /unlink"
 )
@@ -301,6 +302,19 @@ async def run_interactive_session(
                 except Exception as exc:  # pragma: no cover - defensive surface
                     view.show_error(str(exc))
                 continue
+            if prompt_text == FORK_COMMAND:
+                if has_pending_turn_tasks():
+                    view.write_line(
+                        "Cannot fork while work is running or queued."
+                    )
+                    continue
+                if not hasattr(model_client, "_session_id"):
+                    view.write_line("Current model does not support session IDs.")
+                    continue
+                new_session_id = uuid7_string()
+                model_client._session_id = new_session_id
+                view.write_line(f"Forked session: {new_session_id}")
+                continue
             if prompt_text.startswith(f"{LINK_COMMAND} "):
                 link_target = prompt_text[len(LINK_COMMAND) :].strip()
                 if not link_target:

{python_codex-0.2.4 → python_codex-0.2.6}/pycodex/model.py RENAMED Viewed

@@ -19,16 +19,19 @@ except ModuleNotFoundError:  # pragma: no cover - Python 3.10 path
 from .protocol import (
     AssistantMessage,
+    JSONDict,
     ModelResponse,
     ModelStreamEvent,
     Prompt,
     ReasoningItem,
     ToolCall,
 )
+from .model_metadata import model_metadata
 from .utils import build_user_agent, uuid7_string
 DEFAULT_CODEX_CONFIG_PATH = Path.home() / ".codex" / "config.toml"
 DEFAULT_ORIGINATOR = "pycodex"
+RESPONSES_LITE_HEADER = "x-openai-internal-codex-responses-lite"
 ModelStreamEventHandler = Callable[[ModelStreamEvent], None]
 NOOP_MODEL_STREAM_EVENT_HANDLER: 'ModelStreamEventHandler' = lambda _event: None
 DEFAULT_STREAM_MAX_RETRIES = 5
@@ -65,6 +68,7 @@ class ResponsesProviderConfig:
     beta_features_header: 'typing.Union[str, None]' = None
     stream_max_retries: 'typing.Union[int, None]' = None
     stream_idle_timeout_ms: 'typing.Union[int, None]' = None
+    service_tier: 'typing.Union[str, None]' = None
     @classmethod
     def from_codex_config(
@@ -116,6 +120,7 @@ class ResponsesProviderConfig:
             reasoning_effort=selected.get("model_reasoning_effort"),
             reasoning_summary=selected.get("model_reasoning_summary"),
             verbosity=selected.get("model_verbosity"),
+            service_tier=selected.get("service_tier"),
             sandbox_mode=selected.get("sandbox_mode"),
             beta_features_header=",".join(beta_features) or None,
             stream_max_retries=_optional_int(provider.get("stream_max_retries")),
@@ -157,6 +162,65 @@ class ResponsesProviderConfig:
             return DEFAULT_STREAM_IDLE_TIMEOUT_MS / 1000.0
         return max(int(self.stream_idle_timeout_ms), 1) / 1000.0
+    def metadata(self) -> 'typing.Union[JSONDict, None]':
+        return model_metadata(self.model)
+    def use_responses_lite(self) -> 'bool':
+        metadata = self.metadata()
+        if metadata is None:
+            return False
+        return metadata.get("use_responses_lite") is True
+    def effective_reasoning_effort(self) -> 'typing.Union[str, None]':
+        if self.reasoning_effort is not None:
+            return str(self.reasoning_effort)
+        metadata = self.metadata()
+        if not _metadata_supports_reasoning(metadata):
+            return None
+        return _optional_metadata_string(metadata, "default_reasoning_level")
+    def effective_reasoning_summary(self) -> 'typing.Union[str, None]':
+        summary = self.reasoning_summary
+        if summary is None:
+            metadata = self.metadata()
+            if not _metadata_supports_reasoning(metadata):
+                return None
+            summary = _optional_metadata_string(metadata, "default_reasoning_summary")
+        if summary is None:
+            return None
+        if str(summary).strip().lower() == "none":
+            return None
+        return str(summary)
+    def effective_verbosity(self) -> 'typing.Union[str, None]':
+        if self.verbosity is not None:
+            return str(self.verbosity)
+        metadata = self.metadata()
+        if metadata is None or metadata.get("support_verbosity") is not True:
+            return None
+        return _optional_metadata_string(metadata, "default_verbosity")
+    def effective_service_tier(self) -> 'typing.Union[str, None]':
+        service_tier = self.service_tier
+        if service_tier is None:
+            return None
+        service_tier = str(service_tier).strip()
+        if service_tier == "fast":
+            service_tier = "priority"
+        if not service_tier or service_tier == "default":
+            return None
+        metadata = self.metadata()
+        if metadata is None:
+            return None
+        supported = metadata.get("service_tiers")
+        if not isinstance(supported, list):
+            return None
+        for tier in supported:
+            if isinstance(tier, dict) and tier.get("id") == service_tier:
+                return service_tier
+        return None
 def _optional_bool(value: 'typing.Union[bool, str, int, None]') -> 'typing.Union[bool, None]':
     if value is None:
@@ -171,6 +235,43 @@ def _optional_bool(value: 'typing.Union[bool, str, int, None]') -> 'typing.Union
     raise ValueError(f"invalid boolean config value: {value!r}")
+def _metadata_supports_reasoning(
+    metadata: 'typing.Union[JSONDict, None]',
+) -> 'bool':
+    return metadata is not None and metadata.get("supports_reasoning_summaries") is True
+def _optional_metadata_string(
+    metadata: 'typing.Union[JSONDict, None]',
+    key: 'str',
+) -> 'typing.Union[str, None]':
+    if metadata is None:
+        return None
+    value = metadata.get(key)
+    if value is None:
+        return None
+    text = str(value).strip()
+    return text or None
+def _strip_image_details(items: 'typing.Iterable[object]') -> 'None':
+    for item in items:
+        if not isinstance(item, dict):
+            continue
+        content = item.get("content")
+        if isinstance(content, list):
+            _strip_image_detail_from_content_items(content)
+        output = item.get("output")
+        if isinstance(output, list):
+            _strip_image_detail_from_content_items(output)
+def _strip_image_detail_from_content_items(items: 'typing.Iterable[object]') -> 'None':
+    for item in items:
+        if isinstance(item, dict) and item.get("type") == "input_image":
+            item.pop("detail", None)
 class ResponsesApiError(RuntimeError):
     pass
@@ -392,34 +493,73 @@ class ResponsesModelClient:
             ) from exc
     def _build_payload(self, prompt: 'Prompt') -> 'typing.Dict[str, object]':
+        use_responses_lite = self._config.use_responses_lite()
+        input_items = [item.serialize() for item in prompt.input]
+        if use_responses_lite:
+            _strip_image_details(input_items)
+        tools = [tool.serialize() for tool in prompt.tools]
         payload: 'typing.Dict[str, object]' = {
             "model": self.model,
-            "instructions": prompt.base_instructions or "",
-            "input": [item.serialize() for item in prompt.input],
-            "tools": [tool.serialize() for tool in prompt.tools],
-            "parallel_tool_calls": prompt.parallel_tool_calls,
+            "input": input_items,
+            "parallel_tool_calls": prompt.parallel_tool_calls and not use_responses_lite,
             "store": False,
             "stream": True,
             "include": ["reasoning.encrypted_content"],
             "prompt_cache_key": self._session_id,
         }
-        if prompt.tools:
+        if use_responses_lite:
+            prefix: 'typing.List[typing.Dict[str, object]]' = [
+                {
+                    "type": "additional_tools",
+                    "role": "developer",
+                    "tools": tools,
+                }
+            ]
+            if prompt.base_instructions:
+                prefix.append(
+                    {
+                        "type": "message",
+                        "role": "developer",
+                        "content": [
+                            {
+                                "type": "input_text",
+                                "text": prompt.base_instructions,
+                            }
+                        ],
+                    }
+                )
+            payload["input"] = prefix + input_items
+        else:
+            payload["instructions"] = prompt.base_instructions or ""
+            payload["tools"] = tools
+        if prompt.tools or use_responses_lite:
             payload["tool_choice"] = "auto"
         reasoning: 'typing.Dict[str, str]' = {}
-        if self._config.reasoning_effort is not None:
-            reasoning["effort"] = self._config.reasoning_effort
-        if self._config.reasoning_summary is not None:
-            reasoning["summary"] = self._config.reasoning_summary
+        reasoning_effort = self._config.effective_reasoning_effort()
+        reasoning_summary = self._config.effective_reasoning_summary()
+        if reasoning_effort is not None:
+            reasoning["effort"] = reasoning_effort
+        if reasoning_summary is not None:
+            reasoning["summary"] = reasoning_summary
+        if use_responses_lite and reasoning:
+            reasoning["context"] = "all_turns"
         if reasoning:
             payload["reasoning"] = reasoning
         text = None
-        if self._config.verbosity is not None:
-            text = {"verbosity": self._config.verbosity}
+        verbosity = self._config.effective_verbosity()
+        if verbosity is not None:
+            text = {"verbosity": verbosity}
         if text is not None:
             payload["text"] = text
+        service_tier = self._config.effective_service_tier()
+        if service_tier is not None:
+            payload["service_tier"] = service_tier
         return payload
     def _list_models_sync(self) -> 'typing.List[str]':
@@ -483,6 +623,8 @@ class ResponsesModelClient:
             headers["authorization"] = f"Bearer {api_key}"
         if self._config.beta_features_header is not None:
             headers["x-codex-beta-features"] = self._config.beta_features_header
+        if self._config.use_responses_lite():
+            headers[RESPONSES_LITE_HEADER] = "true"
         if self._openai_subagent is not None:
             headers["x-openai-subagent"] = self._openai_subagent
         if prompt.turn_metadata is not None:

python-codex 0.2.4__tar.gz → 0.2.6__tar.gz

python-codex 0.2.4tar.gz → 0.2.6tar.gz