feat(bub): capture long-running agent events - #1230
Open
PsiACE wants to merge 2 commits into
Open
Conversation
This was referenced Aug 13, 2026
PsiACE
marked this pull request as ready for review
August 13, 2026 11:10
Contributor
There was a problem hiding this comment.
Pull request overview
This PR adds an opt-in Bub-side “capture layer” that records long-running agent trajectory events (user prompt, LLM results, tool results) as bounded PowerContext Content Sources, periodically checkpointing them through the Memory pipeline and redacting sensitive values at the client boundary.
Changes:
- Switch Bub integration configuration to Bub’s Pydantic settings (
PowerContextSettings) and wire capture/checkpoint options into the plugin. - Capture and flush Bub events as Content Sources with JSONL optional capture logging and secret redaction.
- Add an e2e regression test for credential redaction and update integration dependencies/docs.
Reviewed changes
Copilot reviewed 5 out of 6 changed files in this pull request and generated 5 comments.
Show a summary per file
| File | Description |
|---|---|
| integrations/bub/src/powercontext_bub/plugin.py | Implements Pydantic-backed settings, event capture/checkpointing, and redaction utilities. |
| integrations/bub/src/powercontext_bub/init.py | Exposes PowerContextSettings alongside the plugin. |
| integrations/bub/README.md | Documents opt-in capture behavior and new configuration options. |
| integrations/bub/pyproject.toml | Adds pydantic-settings dependency for Bub settings integration. |
| e2e/bub/uv.lock | Locks the new dependency for the Bub e2e environment. |
| e2e/bub/tests/test_bub_capture.py | Adds regression coverage for credential redaction in captured tool results. |
Suppressed comments (1)
integrations/bub/src/powercontext_bub/plugin.py:393
- Same
isinstance(value, list | tuple)issue here: it will raiseTypeErrorwhen walking Codex auth JSON arrays, which can break secret redaction (and therefore capture) at runtime. Use(list, tuple)instead.
def _sensitive_values(value: Any, *, sensitive: bool = False) -> set[str]:
if isinstance(value, dict):
secrets: set[str] = set()
for key, item in value.items():
secrets.update(_sensitive_values(item, sensitive=sensitive or _is_sensitive_key(str(key))))
return secrets
if isinstance(value, list | tuple):
secrets = set()
for item in value:
secrets.update(_sensitive_values(item, sensitive=sensitive))
return secrets
if sensitive and isinstance(value, str) and len(value) >= 8:
return {value}
return set()
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
| timeout=float(os.getenv("POWERCONTEXT_BUB_TIMEOUT", "10")), | ||
| max_bytes=max_bytes, | ||
| ) | ||
| base_url: HttpUrl = HttpUrl("http://127.0.0.1:8000") |
Comment on lines
+344
to
+351
| def _sanitize(value: Any) -> Any: | ||
| if isinstance(value, dict): | ||
| return { | ||
| str(key): "[REDACTED]" if _is_sensitive_key(str(key)) else _sanitize(item) for key, item in value.items() | ||
| } | ||
| if isinstance(value, list | tuple): | ||
| return [_sanitize(item) for item in value] | ||
| return value |
Comment on lines
+371
to
+377
| def _codex_auth_secrets() -> set[str]: | ||
| codex_home = Path(os.getenv("CODEX_HOME", str(Path.home() / ".codex"))).expanduser() | ||
| try: | ||
| auth = json.loads((codex_home / "auth.json").read_text(encoding="utf-8")) | ||
| except (OSError, json.JSONDecodeError): | ||
| return set() | ||
| return _sensitive_values(auth) |
Comment on lines
+60
to
+64
| assert len(captured_requests) == 1 | ||
| request = captured_requests[0] | ||
| assert request.metadata["event"] == "tool_result" | ||
| assert sensitive_value not in request.content | ||
| assert "[REDACTED]" in request.content |
Comment on lines
66
to
+70
| def load_state(self, message: Any, session_id: str) -> TurnState: | ||
| del message, session_id | ||
| return { | ||
| STATE_KEY: { | ||
| "base_url": self.settings.base_url, | ||
| "scope_id": self.settings.scope_id, | ||
| "base_url": self.base_url, |
Contributor
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 5 out of 6 changed files in this pull request and generated 1 comment.
Suppressed comments (3)
integrations/bub/src/powercontext_bub/plugin.py:70
load_statediscards thesession_idBub provides, but event capture later falls back tostate.get("session_id", "unknown"). If Bub doesn’t injectsession_idintostate, captured events will all be tagged with "unknown" (andsource_iduniqueness degrades across sessions). Persist thesession_idin the returned state so capture metadata is stable without callers patching state manually.
def load_state(self, message: Any, session_id: str) -> TurnState:
del message, session_id
return {
STATE_KEY: {
"base_url": self.base_url,
integrations/bub/src/powercontext_bub/plugin.py:155
save_stateperforms only a single flush. Sinceflush_memoryis explicitly bounded, one call may not advancecurrent_cursorto the latest capturedposition, leaving some captured Sources unprocessed when the session ends. Consider retrying flush until the cursor reaches the captured high-watermark (or no further progress is possible), with a small upper bound to avoid infinite loops.
if not self.settings.capture_events:
return
async with self._capture_lock:
await self._flush_captured_sources(state, final=True)
e2e/bub/tests/test_bub_capture.py:58
- This test sets
BUB_API_KEYto the samesensitive_valuethat is passed in tool arguments and echoed inToolCallResult.result, so it can still pass even if key-based redaction ofargumentsregresses (because env-value redaction would remove the sentinel from the serialized content). Use a different env value and avoid embedding the argument sentinel inresultso the test specifically verifies argument-key redaction at the client boundary.
settings = PowerContextSettings(
base_url="http://127.0.0.1:8000",
scope_id="test:scope",
capture_events=True,
capture_checkpoint_every=100,
Comment on lines
+344
to
+351
| def _sanitize(value: Any) -> Any: | ||
| if isinstance(value, dict): | ||
| return { | ||
| str(key): "[REDACTED]" if _is_sensitive_key(str(key)) else _sanitize(item) for key, item in value.items() | ||
| } | ||
| if isinstance(value, list | tuple): | ||
| return [_sanitize(item) for item in value] | ||
| return value |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stack
Merge in this order. This PR depends on #1229.
Which issue or RFC does this PR close?
Implements the Bub capture layer of #1229.
Rationale for this change
Long-running acceptance workloads need durable evidence without persisting credentials or adding harness-specific provider configuration.
What changes are included in this PR?
Are there any user-facing changes?
Capture is opt-in. Existing Bub behavior and configuration remain unchanged.
How was this change tested?
make harness-syncmake harness-checkAI usage statement
OpenAI Codex (GPT-5) assisted with implementation and tests. The author is responsible for the submitted changes.