feat: user-facing activity logs with 19 curated events

[abhizipstack]

What

Introduces a user-facing activity log system that replaces developer-internal log dumps with plain-language messages for all major user actions. The Logs panel now defaults to a "User activity" view showing curated, human-readable messages.

Backend — Event infrastructure:

• New UserLevel base class in base_types.py with audience()="user" (existing events default to "developer")
• LogHelper.log() extended with audience param, threaded through eventmgr.write_line to the socket payload
• 19 proto types + event classes (U001–U019) across 5 priority tiers

19 events across all user-facing operations:

Code	Event	Trigger location
P1 — Core model operations
U001	ModelRunStarted	visitran.py:execute_graph
U002	ModelRunSucceeded	visitran.py:execute_graph
U003	ModelRunFailed	visitran.py:execute_graph
U004	TransformationApplied	transformation/views.py:set_model_transformation
U005	TransformationDeleted	transformation/views.py:delete_model_transformation
U006	ModelConfigured	transformation/views.py:set_model_config_and_reference
U007	SeedCompleted	visitran.py:validate_and_run_seed
P2 — Job scheduler
U008	JobCreated	scheduler/views.py:create_periodic_task
U009	JobUpdated	scheduler/views.py:update_periodic_task
U010	JobDeleted	scheduler/views.py:delete_periodic_task
U011	JobTriggered	scheduler/views.py:_dispatch_task_run
P3 — Model/file CRUD
U012	ModelCreated	explorer/views.py:create_model_explorer
U013	FileDeleted	explorer/views.py:delete_a_file_or_folder
U014	FileRenamed	explorer/views.py:rename_a_file_or_folder
P4 — Connections
U015	ConnectionCreated	connection/views.py:create_connection
U016	ConnectionTested	connection/views.py:test_connection
U017	ConnectionDeletedEvt	connection/views.py:delete_connection
P5 — Environments
U018	EnvironmentCreated	environment/views.py:create_environment
U019	EnvironmentDeleted	environment/views.py:delete_environment

Frontend:

• Socket handler reads data?.data?.audience alongside level/message
• New "User activity" option (default) in the log-level dropdown — filters to audience === "user"
• User-activity rows styled with primary-colored left border + subtle background + bolder font
• Developer logs unchanged when viewing "All logs" / "Info & above" / etc.

Bug fix (bundled):

• CreateConnection.jsx: hasDetailsChanged useMemo was declared after the useCallback that depends on it — TDZ crash. Moved above. (Introduced by PR FIX: Connection Name and Description Losing Last Character When Editing #57)

Why

The bottom Logs panel showed raw developer-internal messages like Executing Model Node: Database: globe_testing Database Type: postgres .... Users couldn't tell what happened after an action. Now the default view shows plain-language messages like Applied sort transformation on "mdoela" and Model "mdoela" built successfully in 0.42s.

How

• UserLevel inherits from BaseEvent, overrides audience() → "user".
• BaseEvent gets a default audience() → "developer" for backward compat.
• eventmgr.py:write_line reads audience from the event via getattr and passes to LogHelper.log().
• Socket payload shape becomes {level, message, audience} — additive, no breaking change.
• Frontend filters: logsLevel === "user" → show only audience === "user" entries.

Can this PR break any existing features?

Low risk:

• BaseEvent.audience() default is "developer" — all existing events behave identically.
• LogHelper.log() audience param defaults to "developer" — existing callers unaffected.
• Socket payload adds audience field — frontend falls back to "developer" if absent.
• fire_event calls are added AFTER the operation succeeds — if the event itself fails, it's caught by the existing try/except in write_line and logged, never interrupts the operation.
• CreateConnection TDZ fix is a pure declaration-order change — no logic change.

Database Migrations

None.

Env Config

None.

Related Issues or PRs

• PR fix: CreateConnection crash — hasDetailsChanged temporal dead zone #65 — CreateConnection TDZ fix (also bundled here for the same branch)
• PR FIX: Reduce logs panel noise and make execution messages readable (OR-1457) #59 — Log-level filter infrastructure (this PR builds on it)

Dependencies Versions

No changes.

Notes on Testing

Backend verified:

• Smoke test via Django shell: all 19 events create correctly, messages render in plain language, audience() returns "user", LogHelper.log() includes audience in payload.
• execute_graph confirmed to run with 2 loggers active (file_log + stdout_log) during model execution.
• Events fire through LogHelper → Celery queue → socket emission pipeline.

Frontend verified:

• "User activity" is the default dropdown option.
• User-audience rows render with left border + background styling.
• Developer logs remain visible in "All logs" mode.
• WebSocket delivery requires gunicorn (not runserver) or production nginx — tested pipeline end-to-end, socket proxy limitation in React dev server documented.

Checklist

I have read and understood the Contribution Guidelines.

🤖 Generated with Claude Code

[greptile-apps]

Greptile Summary

This PR introduces a user-facing activity log system with 19 curated events (U001–U019) across the backend and frontend. The event infrastructure is well-structured: UserLevel extends BaseEvent, developer logs are unchanged, and the additive socket payload change is backward-compatible. Several prior-round issues (name-lookup safety, seed schema, field naming, false-failure on connection delete, JobTriggered premature firing) have been addressed in follow-up commits.

Two minor inaccuracies remain in the new event messages worth noting before merge.

Confidence Score: 4/5

Safe to merge with one known pre-existing runtime bug in the seed path that prior review rounds flagged; all new view-function events are exception-safe.

All prior P0/P1 findings (false-failure on connection delete, premature JobTriggered, seed schema empty string, field naming) are addressed in the follow-up commits. The two new findings are both P2. Score stays at 4 rather than 5 because the SeedCompleted status() method-shadowing bug — flagged and acknowledged in the previous round — appears unresolved in HEAD, and would cause a successful seed run to be reported as a failure through the activity log path.

backend/visitran/events/types.py (SeedCompleted.status() still shadows proto field); backend/visitran/visitran.py (repr vs str); backend/backend/core/scheduler/views.py (multi-model scope label)

Important Files Changed

Filename	Overview
backend/visitran/events/base_types.py	Adds UserLevel base class with audience()="user", title(), subtitle(), status() defaults; BaseEvent.audience() defaults to "developer" for backward compat — clean design, but status() naming collides with proto field of the same name in SeedCompleted.
backend/visitran/events/eventmgr.py	Branches write_line on audience: user events emit a rich structured payload (title/subtitle/status/timestamp); developer events unchanged. getattr(msg.data, "status", lambda: "info")() is safe for all new events except SeedCompleted (proto field shadows method).
backend/visitran/events/types.py	Adds 20 UserLevel event classes (U001–U019 + ConnectionDeleteFailedEvt); messages and titles are well-written; SeedCompleted.status() method shadows its status: str proto field at runtime.
backend/visitran/visitran.py	Fires ModelRunStarted/Succeeded/Failed per node and SeedCompleted per seed; start_time correctly scoped per iteration; generic except path uses repr(err) instead of str(err) for the user-facing error field.
backend/backend/core/scheduler/views.py	JobCreated/Updated/Deleted/Triggered events fire after successful operations; scope falls back to "job" for multi-model models_override, producing misleading "running all models" log message for partial runs.
backend/backend/core/routers/connection/views.py	ConnectionCreated/Tested/Deleted events added; VisitranBackendBaseException re-raised cleanly; ConnectionDeleteFailedEvt fires before re-raising the structured exception.
backend/backend/core/routers/environment/views.py	EnvironmentCreated/Deleted events fire correctly; ProtectedError handler refactored to use EnvironmentInUse exception with a cleaner job-name list comprehension.
frontend/src/ide/editor/no-code-model/no-code-model.jsx	Adds "User activity" default log filter, rich card rendering for user events (status icon, title, subtitle, timestamp), and guards against messages with only a title field; developer log rendering unchanged.
backend/backend/errors/validation_exceptions.py	Adds EnvironmentInUse and ConnectionDeleteFailed structured exceptions with HTTP 400 and severity="Warning"; wired to the new error-code templates.
docker/docker-compose.yaml	Adds celery_log_task_queue to the Celery worker's -Q list so log-event tasks are consumed in the docker-compose environment.

Sequence Diagram

sequenceDiagram
    participant View as Django View / visitran.py
    participant FE as fire_event()
    participant EM as EventManager write_line
    participant LH as LogHelper publish_log
    participant CQ as Celery log queue
    participant UI as Frontend jsx

    View->>FE: fire_event(UserLevelEvent)
    FE->>EM: write_line(EventMsg)
    EM->>EM: audience → "user"
    EM->>EM: build payload title/subtitle/status/timestamp
    EM->>LH: publish_log(session_id, payload)
    LH->>CQ: kombu producer publish
    Note over LH: exceptions caught internally
    CQ-->>UI: socket logs event
    UI->>UI: audience user → store rich entry
    UI->>UI: logsLevel user → render card

Loading

Prompt To Fix All With AI

This is a comment left during a code review.
Path: backend/backend/core/scheduler/views.py
Line: 639-641

Comment:
**"Running all models" shown for multi-model partial runs**

`scope` falls back to `"job"` whenever `models_override` has more than one element, so `JobTriggered.message()` renders `"running all models"` even when the user triggered only a specific subset. The one-model case is correct, but the multi-model case is misleading.

```suggestion
    if not models_override:
        scope = "job"
    elif len(models_override) == 1:
        scope = models_override[0]
    else:
        scope = f"{len(models_override)} models"
```

You'd also want to update `JobTriggered.message()` / `subtitle()` in `types.py` to handle the "N models" scope string alongside `"job"` and a single model name.

How can I resolve this? If you propose a fix, please make it concise.

---

This is a comment left during a code review.
Path: backend/visitran/visitran.py
Line: 388-397

Comment:
**`repr(err)` leaks Python noise into the user-facing message**

The domain-exception catch (line 374) passes `str(visitran_err)` — clean, human-readable. The generic catch passes `repr(err)`, which produces output like `KeyError('column_name')` or `AttributeError("'NoneType' object has no attribute 'foo'")`. Since this string surfaces directly in the `ModelRunFailed` subtitle shown to users, it should use `str(err)` for consistency.

```suggestion
                fire_event(
                    ModelRunFailed(
                        model_name=dest_table or str(node_name),
                        error=str(err),
                    )
                )
```

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (9): Last reviewed commit: "fix: rename status() to event_status() t..." | Re-trigger Greptile}

[abhizipstack]

Additional fix pushed: Environment delete now shows a proper error when the environment is used by a job.

• Added EnvironmentInUse exception class + BackendErrorMessages.ENVIRONMENT_IN_USE error template following the same pattern as SourceTableDoesNotExist, ConnectionDependency, etc.
• Raises through handle_http_request decorator → frontend gets {error_message, is_markdown: true, severity: 'Warning'} → renders markdown with bold job names.
• Before: generic 'Something went wrong'. After: formatted warning showing which job(s) reference the environment.

[wicky-zipstack]

@abhizipstack
1. fire_event() is NOT wrapped in try/except — eventmgr.py:fire_event

PR description says "if the event itself fails, it's caught by the existing try/except in write_line" — I traced the code, no such guard exists. Most fire_event calls fire after the operation succeeds (e.g. delete_connection, _dispatch_task_run). If Redis hiccups or StateStore raises, the view returns 500 even though the connection/job was actually deleted.

Fix in fire_event (covers all 19 sites):

def fire_event(e: BaseEvent, level=None) -> None:
    try:
        event_manager.fire_event(e, level=level)
    except Exception:
        logger.exception("fire_event failed for %s", type(e).__name__)

2. JobTriggered fires BEFORE dispatch — scheduler/views.py:_dispatch_task_run

fire_event(JobTriggered(...)) happens before the Celery dispatch + sync fallback. If both fail, log shows "Job triggered" but nothing ran. Move it AFTER successful dispatch.

3. Bundled with merged PR #65 — CreateConnection.jsx

The TDZ fix is already in main via #65. Please rebase so the diff doesn't show duplicate.

4. ConnectionDeletedEvt naming — types.py:984

No collision with existing ConnectionDeleted — drop the Evt suffix to match EnvironmentDeleted/JobDeleted. Also: proto field is connection_id but you're passing the connection name (connection/views.py:32). Rename the field to connection_name.

5. Multi-model scope detection — scheduler/views.py:_dispatch_task_run

scope = models_override[0] if ... len(models_override) == 1 else "job" — if 2+ models in override, falls through to "job" which is wrong (it's a partial run). Use ", ".join(models_override) or a "selection" bucket.

6. Extra DB roundtrip in delete_connection — connection/views.py:29-31

Fetches full connection (including masked credentials) just for the name. Either return name from delete_connection, or fetch only the name field.

7. FileDeleted for batch — explorer/views.py

", ".join(deleted_files) becomes a 1000-char string when wiping many models. Truncate: f"{first_3} and {N-3} more".

8. localStorage migration — no-code-model.jsx:233

Existing users have "info"/"debug" saved → new "user" default only applies to first-time users. Returning users never see the new view unless they manually pick it. Consider one-time migration from "info" → "user".

9. Bundled delete_environment refactor

The new EnvironmentInUse exception + ENVIRONMENT_IN_USE error code are good improvements but unrelated to user activity logs. Either split out or call out clearly in the description.

10. No tests — at minimum add: LogHelper.log(audience=...) payload shape + regression test that fire_event doesn't bubble exceptions when StateStore is unavailable (covers #1).

[tahierhussain]

LGTM