codex-agent-framework 0.1.18__tar.gz → 0.1.19__tar.gz

{codex_agent_framework-0.1.18 → codex_agent_framework-0.1.19}/CHANGELOG.md

@@ -4,6 +4,15 @@ 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.
 
+## [0.1.19] - 2026-05-08
+### Changed
+- Compacting a session now purges compacted raw history from the active session file, keeping only the previous compaction anchor, the new compaction summary, and any unfinished remainder turn.
+- Drop compaction summaries older than the latest anchor from active session files because each new compaction already incorporates the previous anchor.
+
+### Tests
+- Add regression coverage for compaction prefix pruning, latest-anchor replay, older compaction removal, persisted session shape, and status/event message counts.
+- Validate the full suite at 450 passing tests.
+
 ## [0.1.18] - 2026-05-08
 ### Added
 - Add generic plugin event and hook handler discovery so plugins can react to agent lifecycle events without hard-coded core coupling.

{codex_agent_framework-0.1.18/codex_agent_framework.egg-info → codex_agent_framework-0.1.19}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codex-agent-framework
-Version: 0.1.18
+Version: 0.1.19
 Summary: A lightweight event-driven Codex agent runtime.
 Author: Baptiste
 License-Expression: MIT
@@ -638,10 +638,10 @@ python -m pyflakes codex_agent tests
 
 The tests isolate `AGENT_RUNTIME_DIR` automatically, so they should not create or resume sessions from your real `~/.agent_runtime`.
 
-The published `0.1.18` release was validated at:
+The published `0.1.19` release was validated at:
 
 ```text
-449 passed
+450 passed
 ```
 
 ## Packaging
@@ -659,6 +659,7 @@ The distribution includes prompt text files, `codex_agent/get_text/default_gitig
 
 ## Recent changes
 
+- `0.1.19`: prune compacted raw session history after compaction, keep only the latest compaction anchor plus the new compaction and any unfinished remainder, and drop older redundant compaction summaries from active session files.
 - `0.1.18`: make `start server/tray` detach by default, move durable memory fully into the memory plugin, keep wakeups as a core runtime primitive, expose RAG memory timestamps/sources, and move browser/desktop backend controllers into core modules with plugins as facades.
 - `0.1.17`: reorganize the TUI and FastAPI server into dedicated `codex_agent.tui` and `codex_agent.server` packages, remove obsolete message payload fields, and clean unused imports.
 - `0.1.16`: tolerate missing OpenAI API keys at startup by disabling OpenAI-dependent voice and memory features, and skip memory archiving when the memory plugin is unavailable.

{codex_agent_framework-0.1.18 → codex_agent_framework-0.1.19}/README.md

@@ -589,10 +589,10 @@ python -m pyflakes codex_agent tests
 
 The tests isolate `AGENT_RUNTIME_DIR` automatically, so they should not create or resume sessions from your real `~/.agent_runtime`.
 
-The published `0.1.18` release was validated at:
+The published `0.1.19` release was validated at:
 
 ```text
-449 passed
+450 passed
 ```
 
 ## Packaging
@@ -610,6 +610,7 @@ The distribution includes prompt text files, `codex_agent/get_text/default_gitig
 
 ## Recent changes
 
+- `0.1.19`: prune compacted raw session history after compaction, keep only the latest compaction anchor plus the new compaction and any unfinished remainder, and drop older redundant compaction summaries from active session files.
 - `0.1.18`: make `start server/tray` detach by default, move durable memory fully into the memory plugin, keep wakeups as a core runtime primitive, expose RAG memory timestamps/sources, and move browser/desktop backend controllers into core modules with plugins as facades.
 - `0.1.17`: reorganize the TUI and FastAPI server into dedicated `codex_agent.tui` and `codex_agent.server` packages, remove obsolete message payload fields, and clean unused imports.
 - `0.1.16`: tolerate missing OpenAI API keys at startup by disabling OpenAI-dependent voice and memory features, and skip memory archiving when the memory plugin is unavailable.

{codex_agent_framework-0.1.18 → codex_agent_framework-0.1.19}/codex_agent/sessions.py

@@ -73,6 +73,8 @@ class AgentSession(modict):
 
         insert_at = plan.start + len(compacted_messages)
         self.messages.insert(insert_at, message)
+        previous_anchor = [plan.anchor] if plan.anchor is not None else []
+        self.messages = previous_anchor + self.messages[insert_at:]
         self.save()
         return message
 

{codex_agent_framework-0.1.18 → codex_agent_framework-0.1.19/codex_agent_framework.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codex-agent-framework
-Version: 0.1.18
+Version: 0.1.19
 Summary: A lightweight event-driven Codex agent runtime.
 Author: Baptiste
 License-Expression: MIT
@@ -638,10 +638,10 @@ python -m pyflakes codex_agent tests
 
 The tests isolate `AGENT_RUNTIME_DIR` automatically, so they should not create or resume sessions from your real `~/.agent_runtime`.
 
-The published `0.1.18` release was validated at:
+The published `0.1.19` release was validated at:
 
 ```text
-449 passed
+450 passed
 ```
 
 ## Packaging
@@ -659,6 +659,7 @@ The distribution includes prompt text files, `codex_agent/get_text/default_gitig
 
 ## Recent changes
 
+- `0.1.19`: prune compacted raw session history after compaction, keep only the latest compaction anchor plus the new compaction and any unfinished remainder, and drop older redundant compaction summaries from active session files.
 - `0.1.18`: make `start server/tray` detach by default, move durable memory fully into the memory plugin, keep wakeups as a core runtime primitive, expose RAG memory timestamps/sources, and move browser/desktop backend controllers into core modules with plugins as facades.
 - `0.1.17`: reorganize the TUI and FastAPI server into dedicated `codex_agent.tui` and `codex_agent.server` packages, remove obsolete message payload fields, and clean unused imports.
 - `0.1.16`: tolerate missing OpenAI API keys at startup by disabling OpenAI-dependent voice and memory features, and skip memory archiving when the memory plugin is unavailable.

{codex_agent_framework-0.1.18 → codex_agent_framework-0.1.19}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "codex-agent-framework"
-version = "0.1.18"
+version = "0.1.19"
 description = "A lightweight event-driven Codex agent runtime."
 readme = "README.md"
 requires-python = ">=3.10"

{codex_agent_framework-0.1.18 → codex_agent_framework-0.1.19}/tests/test_agent.py

@@ -341,7 +341,7 @@ def test_agent_interrupt_emits_request_event():
     assert agent.interrupt("api") is True
     assert seen == [("agent_interrupt_requested_event", "api", True)]
 
-def test_agent_session_compact_inserts_compaction_before_remainder(tmp_path, monkeypatch):
+def test_agent_session_compact_purges_compacted_prefix_before_remainder(tmp_path, monkeypatch):
     monkeypatch.setattr(utils, "RUNTIME_DIR", str(tmp_path))
     user = DeveloperMessage(content="old session marker")
     done = AssistantMessage(content="done")
@@ -352,7 +352,8 @@ def test_agent_session_compact_inserts_compaction_before_remainder(tmp_path, mon
     message = session.compact(client=client)
 
     assert isinstance(message, CompactionMessage)
-    assert session.messages == [user, done, message, current]
+    assert session.messages == [message, current]
+    assert [msg.id for msg in AgentSession.load(session.session_id).messages] == [message.id, current.id]
     assert "output_items" not in message
     assert message.summaries == [FakeCompactionResult.output_items[1]]
     assert message.compacted_message_count == 2
@@ -382,12 +383,31 @@ def test_agent_session_compact_starts_at_latest_compaction(tmp_path, monkeypatch
 
     message = session.compact(client=client)
 
-    assert session.messages == [old_user, old_done, previous, new_user, new_done, message]
+    assert session.messages == [previous, message]
+    assert [msg.id for msg in AgentSession.load(session.session_id).messages] == [previous.id, message.id]
     assert client.history[0] == {"type": "compaction_summary", "id": "old"}
     assert "very old" not in json.dumps(client.history)
     assert "new done" in json.dumps(client.history)
 
 
+def test_agent_session_compact_includes_only_latest_compaction_anchor(tmp_path, monkeypatch):
+    monkeypatch.setattr(utils, "RUNTIME_DIR", str(tmp_path))
+    older = CompactionMessage(output_items=[{"type": "compaction_summary", "id": "older"}])
+    latest = CompactionMessage(output_items=[{"type": "compaction_summary", "id": "latest"}])
+    new_user = DeveloperMessage(content="new facts")
+    new_done = AssistantMessage(content="new facts done")
+    session = AgentSession(messages=[older, latest, new_user, new_done])
+    client = FakeCompactionClient()
+
+    message = session.compact(client=client)
+
+    assert session.messages == [latest, message]
+    assert [msg.id for msg in AgentSession.load(session.session_id).messages] == [latest.id, message.id]
+    assert client.history[0] == {"type": "compaction_summary", "id": "latest"}
+    assert {item.get("id") for item in client.history if item.get("type") == "compaction_summary"} == {"latest"}
+    assert "new facts done" in json.dumps(client.history)
+
+
 def test_agent_compact_session_emits_volume_events(tmp_path, monkeypatch):
     monkeypatch.setattr(utils, "RUNTIME_DIR", str(tmp_path))
     agent = Agent(session="new", voice_enabled=False)
@@ -425,11 +445,11 @@ def test_agent_compact_session_emits_volume_events(tmp_path, monkeypatch):
     assert seen[1].status.context_current_tokens != 5000
     assert seen[1].compacted_message_count == 2
     assert seen[1].post_compaction_message_count == 2
-    assert seen[1].total_session_messages == 4
+    assert seen[1].total_session_messages == 2
     status = agent.get_status()
     assert status.source == "local_estimate"
     assert status.sent_session_messages == 2
-    assert status.total_session_messages == 4
+    assert status.total_session_messages == 2
     assert status.context_current_tokens != 5000
     assert status.quota_5h_percent == 12
     assert status.quota_7d_percent == 34