codex-agent-framework 0.1.10__tar.gz → 0.1.12__tar.gz

{codex_agent_framework-0.1.10 → codex_agent_framework-0.1.12}/CHANGELOG.md

@@ -3,6 +3,37 @@
 All notable changes to this project will be documented in this file.
 
 This project loosely follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/) and uses semantic versioning where practical.
+
+## [0.1.12] - 2026-05-05
+### Added
+- Add a persistent Playwright/Chromium browser controller with tools to open/close the browser, select tabs, navigate, click, fill, select options, press keys, inspect action snapshots, and capture screenshots.
+- Add `browser_goto(url)` to navigate the active tab without opening or switching tabs.
+- Add screenshot-backed visual fallback for pages where DOM snapshots are insufficient, including PDF viewers and image-heavy pages.
+
+### Changed
+- Replace the Selenium/Gecko rendered extraction path with Playwright-managed Chromium.
+- Update web extraction dependencies and documentation to install Chromium with Playwright.
+
+### Tests
+- Add browser controller and rendered extraction regression coverage.
+
+## [0.1.11] - 2026-05-05
+### Added
+- Add a separate `view` tool for folders, URLs, rich document extraction, and Python object inspection.
+- Add context status counts for persistent and temporary session messages, surfaced in the TUI status bar.
+
+### Changed
+- Make `read` a strict local UTF-8 file reader with reliable forced line numbers and regex pattern search by default.
+- Keep extracted `view` output unnumbered, including search snippets, because transformed document line structure is not source-of-truth reliable.
+- Update the system prompt to direct agents to use `read` before targeted edits and `view` for broad extraction.
+
+### Fixed
+- Preserve blank lines and real source line numbers in `read` pattern snippets.
+- Apply line-based pagination before token truncation.
+
+### Tests
+- Add regression coverage for strict `read`, unnumbered `view`, blank-line preservation, regex/literal search, and persistent/temporary message status counts.
+
 ## [0.1.10] - 2026-05-05
 ### Fixed
 - Persist only backend `compaction_summary` items in compaction messages, avoiding bulky compacted conversation payloads while preserving summary replay.

{codex_agent_framework-0.1.10/codex_agent_framework.egg-info → codex_agent_framework-0.1.12}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codex-agent-framework
-Version: 0.1.10
+Version: 0.1.12
 Summary: A lightweight event-driven Codex agent runtime.
 Author: Baptiste
 License-Expression: MIT
@@ -20,7 +20,6 @@ Requires-Dist: beautifulsoup4
 Requires-Dist: codex-backend-sdk
 Requires-Dist: fastapi
 Requires-Dist: filetype
-Requires-Dist: get-gecko-driver
 Requires-Dist: modict
 Requires-Dist: numpy
 Requires-Dist: odfpy
@@ -28,6 +27,7 @@ Requires-Dist: openai
 Requires-Dist: openpyxl
 Requires-Dist: pathspec
 Requires-Dist: pillow
+Requires-Dist: playwright
 Requires-Dist: pydub
 Requires-Dist: pypdf
 Requires-Dist: pynteract
@@ -36,7 +36,6 @@ Requires-Dist: PyYAML
 Requires-Dist: regex
 Requires-Dist: requests
 Requires-Dist: rich
-Requires-Dist: selenium
 Requires-Dist: textual
 Requires-Dist: tiktoken
 Requires-Dist: trafilatura
@@ -73,7 +72,7 @@ It provides a reusable `Agent` abstraction with persistent sessions, local tools
 - Python 3.10 or newer.
 - A working model backend configuration compatible with `codex-backend-sdk` / OpenAI usage.
 - GTK 3 plus Ayatana AppIndicator or AppIndicator bindings are needed for the tray controller on Linux.
-- Firefox / GeckoDriver may be needed for Selenium-backed web extraction paths.
+- Playwright's managed Chromium browser is used for rendered web extraction when static fetching is insufficient. Install it with `python -m playwright install chromium` if needed.
 
 ## Installation
 
@@ -400,6 +399,8 @@ python -m build
 The distribution includes prompt text files and `codex_agent/get_text/default_gitignore` through package data and `MANIFEST.in`.
 
 ## Recent changes
+- `0.1.12`: add a persistent Playwright/Chromium browser controller with tab navigation, DOM/action snapshots, screenshots, form/click/key tools, and `browser_goto(url)` for active-tab navigation.
+- `0.1.11`: split strict line-numbered UTF-8 `read` from unnumbered extracted `view`, preserve blank lines in read snippets, and show persistent+temporary message counts in the TUI status bar.
 - `0.1.10`: persist only backend compaction summaries, drop bulky compacted conversation payloads, and refresh context status after compaction.
 - `0.1.9`: maintenance packaging release after validating the local execution environment and deploy workflow.
 - `0.1.8`: scope TUI replay/SSE catch-up to the active session and make bash/python subprocesses inherit the project Python environment, including service-launched agents.

{codex_agent_framework-0.1.10 → codex_agent_framework-0.1.12}/README.md

@@ -25,7 +25,7 @@ It provides a reusable `Agent` abstraction with persistent sessions, local tools
 - Python 3.10 or newer.
 - A working model backend configuration compatible with `codex-backend-sdk` / OpenAI usage.
 - GTK 3 plus Ayatana AppIndicator or AppIndicator bindings are needed for the tray controller on Linux.
-- Firefox / GeckoDriver may be needed for Selenium-backed web extraction paths.
+- Playwright's managed Chromium browser is used for rendered web extraction when static fetching is insufficient. Install it with `python -m playwright install chromium` if needed.
 
 ## Installation
 
@@ -352,6 +352,8 @@ python -m build
 The distribution includes prompt text files and `codex_agent/get_text/default_gitignore` through package data and `MANIFEST.in`.
 
 ## Recent changes
+- `0.1.12`: add a persistent Playwright/Chromium browser controller with tab navigation, DOM/action snapshots, screenshots, form/click/key tools, and `browser_goto(url)` for active-tab navigation.
+- `0.1.11`: split strict line-numbered UTF-8 `read` from unnumbered extracted `view`, preserve blank lines in read snippets, and show persistent+temporary message counts in the TUI status bar.
 - `0.1.10`: persist only backend compaction summaries, drop bulky compacted conversation payloads, and refresh context status after compaction.
 - `0.1.9`: maintenance packaging release after validating the local execution environment and deploy workflow.
 - `0.1.8`: scope TUI replay/SSE catch-up to the active session and make bash/python subprocesses inherit the project Python environment, including service-launched agents.

{codex_agent_framework-0.1.10 → codex_agent_framework-0.1.12}/codex_agent/agent.py

@@ -952,9 +952,10 @@ class Agent:
     def add_provider(self, func=None, name=None):
         """Register an ephemeral context provider.
 
-        Providers are called each time context is built. They must return a
-        string or None. Returned strings are wrapped in ProviderMessage and are
-        not persisted to the session.
+        Providers are called each time context is built. They may return None,
+        a string, a Message instance, or a list/tuple mixing strings and Message
+        instances. Strings are wrapped in ProviderMessage; Message instances are
+        passed through as already formatted ephemeral context.
         """
         if func is None:
             def decorator(f):
@@ -1090,11 +1091,52 @@ class Agent:
             return [msg for msg in self.session.messages if filter(msg)]
         return sort(self.session.messages)
 
+    def provider_context_wrapper(self, provider_name, content):
+        return f'<context_provider name="{provider_name}">\n{content}\n</context_provider>'
+
+    def provider_message_from_output(self, provider_name, output):
+        if isinstance(output, str):
+            return ProviderMessage(name=provider_name, content=output)
+        if not isinstance(output, Message):
+            raise TypeError(
+                f"Provider {provider_name!r} items must be str or Message, got {type(output).__name__}."
+            )
+
+        message = output.copy()
+        if isinstance(message, ProviderMessage):
+            if not message.get("name"):
+                message.name = provider_name
+            return message
+
+        if isinstance(message, ImageMessage):
+            description = message.get("description") or "Image provided by an ephemeral context provider."
+            message.description = self.provider_context_wrapper(provider_name, description)
+            return message
+
+        content = message.get("content")
+        if isinstance(content, str):
+            message.content = self.provider_context_wrapper(provider_name, content)
+        elif isinstance(content, list):
+            wrapped_parts = []
+            for part in content:
+                if isinstance(part, str):
+                    wrapped_parts.append(self.provider_context_wrapper(provider_name, part))
+                elif isinstance(part, dict) and part.get("type") == "text":
+                    part = dict(part)
+                    part["text"] = self.provider_context_wrapper(provider_name, part.get("text", ""))
+                    wrapped_parts.append(part)
+                else:
+                    wrapped_parts.append(part)
+            message.content = wrapped_parts
+        return message
+
     def get_providers_messages(self):
-        """Retrieves messages from all registered providers.
+        """Retrieves ephemeral messages from all registered providers.
 
-        Returns:
-            List of ProviderMessage objects generated by providers.
+        Providers may return None, a string, a Message, or a list/tuple of
+        strings and Messages. Each output item becomes its own message; strings
+        are wrapped in ProviderMessage, while existing Messages keep their type
+        and only their textual content is marked as provider context.
         """
         provider_msgs=[]
         for provider_name, provider in self.providers.items():
@@ -1105,11 +1147,11 @@ class Agent:
                 reset_current_agent(agent_context)
             if provider_output is None:
                 continue
-            if not isinstance(provider_output, str):
-                raise TypeError(
-                    f"Provider {provider_name!r} must return str or None, got {type(provider_output).__name__}."
-                )
-            provider_msgs.append(ProviderMessage(name=provider_name, content=provider_output))
+            outputs = provider_output if isinstance(provider_output, (list, tuple)) else [provider_output]
+            for output in outputs:
+                if output is None:
+                    continue
+                provider_msgs.append(self.provider_message_from_output(provider_name, output))
         return provider_msgs
 
     def context_status(self):
@@ -1154,6 +1196,14 @@ class Agent:
         visible_history_count = len(truncated_others[max(0, history_start):])
         sent_session_messages = visible_history_count + len(images)
         total_session_messages = len(self.session.messages) if self.session else 0
+        persistent_session_messages = sum(
+            1 for msg in session_messages
+            if msg.get("turns_left", -1) == -1
+        )
+        temporary_session_messages = sum(
+            1 for msg in session_messages
+            if isinstance(msg.get("turns_left", -1), int) and msg.get("turns_left", -1) > 0
+        )
         percent = (current_count / input_limit * 100) if input_limit else 0
         status = modict(
             used_tokens=current_count,
@@ -1167,6 +1217,8 @@ class Agent:
             sent_session_messages=sent_session_messages,
             total_session_messages=total_session_messages,
             sent_session_message_percent=(sent_session_messages / total_session_messages * 100) if total_session_messages else 0,
+            persistent_session_messages=persistent_session_messages,
+            temporary_session_messages=temporary_session_messages,
             context_session_messages=len(session_messages),
             image_messages=len(images),
             pending_messages=len(self.pending),
@@ -1199,6 +1251,8 @@ class Agent:
             sent_session_messages=estimated.sent_session_messages,
             total_session_messages=estimated.total_session_messages,
             sent_session_message_percent=estimated.sent_session_message_percent,
+            persistent_session_messages=estimated.get("persistent_session_messages", 0),
+            temporary_session_messages=estimated.get("temporary_session_messages", 0),
             quota_5h_percent=None,
             quota_7d_percent=None,
             model=self.config.get("model") or "gpt-5.4",
@@ -1223,6 +1277,8 @@ class Agent:
                 sent_session_messages=estimated.sent_session_messages,
                 total_session_messages=estimated.total_session_messages,
                 sent_session_message_percent=estimated.sent_session_message_percent,
+                persistent_session_messages=estimated.get("persistent_session_messages", 0),
+                temporary_session_messages=estimated.get("temporary_session_messages", 0),
                 model=self.config.get("model") or "gpt-5.4",
                 busy=self.busy,
                 current_turn_id=self.current_turn_id,
@@ -1244,6 +1300,8 @@ class Agent:
             sent_session_messages=estimated.sent_session_messages,
             total_session_messages=estimated.total_session_messages,
             sent_session_message_percent=estimated.sent_session_message_percent,
+            persistent_session_messages=estimated.get("persistent_session_messages", 0),
+            temporary_session_messages=estimated.get("temporary_session_messages", 0),
             model=self.config.get("model") or "gpt-5.4",
             busy=self.busy,
             current_turn_id=self.current_turn_id,
@@ -1267,6 +1325,8 @@ class Agent:
             sent_session_messages=context_status.get("sent_session_messages", 0),
             total_session_messages=context_status.get("total_session_messages", 0),
             sent_session_message_percent=context_status.get("sent_session_message_percent", 0),
+            persistent_session_messages=context_status.get("persistent_session_messages", 0),
+            temporary_session_messages=context_status.get("temporary_session_messages", 0),
             quota_5h_percent=wham.quota_5h_percent,
             quota_7d_percent=wham.quota_7d_percent,
             model=model or self.config.get("model") or "gpt-5.4",
@@ -1383,6 +1443,14 @@ class Agent:
         formatted_context = list(msg.format(context=self.shell.namespace) for msg in context)
         total_session_messages = len(self.session.messages) if self.session else 0
         sent_session_messages = len(history) + len(images)
+        persistent_session_messages = sum(
+            1 for msg in session_messages
+            if msg.get("turns_left", -1) == -1
+        )
+        temporary_session_messages = sum(
+            1 for msg in session_messages
+            if isinstance(msg.get("turns_left", -1), int) and msg.get("turns_left", -1) > 0
+        )
         token_count_source = "local_estimate"
         try:
             current_count = self.ai.count_input_tokens(
@@ -1398,6 +1466,8 @@ class Agent:
             sent_session_messages=sent_session_messages,
             total_session_messages=total_session_messages,
             sent_session_message_percent=(sent_session_messages / total_session_messages * 100) if total_session_messages else 0,
+            persistent_session_messages=persistent_session_messages,
+            temporary_session_messages=temporary_session_messages,
             token_count_source=token_count_source,
         )
         return formatted_context