cleveragents/cleveragents-core

Fork 3

UAT: `CorrectionService` stores correction attempts in-memory only — no persistence to `correction_attempts` DB table #4909

New issue

Open

opened 2026-04-08 20:18:32 +00:00 by HAL9000 · 2 comments

HAL9000 commented

2026-04-08 20:18:32 +00:00

Owner

Bug Report

Feature Area: Correction Model — Correction Attempt ULID Tracking / Persistence
Severity: High (correction history is lost on process restart; ULID tracking is ephemeral)
Found by: UAT tester, code analysis
Spec reference: docs/specification.md §correction_attempts DDL (lines ~18894, ~45822); §Correction Safety Guarantees

What Was Tested

The CorrectionService was analyzed to verify that correction attempts are persisted to the correction_attempts database table as required by the spec.

Expected Behavior (from spec)

The spec defines a correction_attempts table with ULID primary key (correction_attempt_id) for tracking every correction attempt:

CREATE TABLE correction_attempts (
    correction_attempt_id TEXT PRIMARY KEY,  -- ULID
    plan_id TEXT NOT NULL REFERENCES plans(plan_id),
    original_decision_id TEXT NOT NULL REFERENCES decisions(decision_id),
    new_decision_id TEXT REFERENCES decisions(decision_id),
    mode TEXT NOT NULL,                      -- revert|append
    guidance TEXT NOT NULL,
    archived_artifacts_path TEXT,
    state TEXT NOT NULL DEFAULT 'pending',   -- pending|executing|complete|failed
    created_at TEXT NOT NULL DEFAULT ...,
    completed_at TEXT
);

The spec's safety guarantees state:

Corrections always:

Create a new attempt revision (increment plan.attempt)

Preserve old artifacts for diff/compare

The CorrectionAttemptRecord domain model and CorrectionAttemptModel DB model are both implemented. The CorrectionAttemptRepository in src/cleveragents/infrastructure/database/repositories.py provides create, get, list_by_plan, update, and delete methods.

Actual Behavior

CorrectionService.__init__ stores all state in plain Python dicts:

def __init__(self, ...):
    self._corrections: dict[str, CorrectionRequest] = {}
    self._impacts: dict[str, CorrectionImpact] = {}
    self._attempts: dict[str, list[CorrectionAttempt]] = {}
    self._results: dict[str, CorrectionResult] = {}

The service never calls CorrectionAttemptRepository.create() or any other repository method. The CorrectionAttemptRecord domain model (which maps to the DB table) is never instantiated by CorrectionService. The CorrectionAttempt objects tracked in self._attempts are a different, simpler model that is never persisted.

The CorrectionService is wired in application/container.py with only checkpoint_service and event_bus — no repository is injected.

Code Location

src/cleveragents/application/services/correction_service.py, lines 101–111 (constructor — no repository)
src/cleveragents/application/container.py, lines 884–888 (container wiring — no repository injected)
src/cleveragents/infrastructure/database/repositories.py — CorrectionAttemptRepository exists but is never used by CorrectionService

Impact

Correction history is lost on process restart — users cannot query past corrections
agents plan diff --correction <CORRECTION_ATTEMPT_ID> cannot look up correction attempts from the DB
The spec's ULID tracking for correction attempts is non-functional (ULIDs are generated but never stored)
Audit trail for corrections is absent
The correction_attempts table is never written to during normal operation

Automated by CleverAgents Bot
Supervisor: UAT Testing | Agent: uat-tester

## Bug Report **Feature Area:** Correction Model — Correction Attempt ULID Tracking / Persistence **Severity:** High (correction history is lost on process restart; ULID tracking is ephemeral) **Found by:** UAT tester, code analysis **Spec reference:** `docs/specification.md` §`correction_attempts` DDL (lines ~18894, ~45822); §Correction Safety Guarantees --- ### What Was Tested The `CorrectionService` was analyzed to verify that correction attempts are persisted to the `correction_attempts` database table as required by the spec. ### Expected Behavior (from spec) The spec defines a `correction_attempts` table with ULID primary key (`correction_attempt_id`) for tracking every correction attempt: ```sql CREATE TABLE correction_attempts ( correction_attempt_id TEXT PRIMARY KEY, -- ULID plan_id TEXT NOT NULL REFERENCES plans(plan_id), original_decision_id TEXT NOT NULL REFERENCES decisions(decision_id), new_decision_id TEXT REFERENCES decisions(decision_id), mode TEXT NOT NULL, -- revert|append guidance TEXT NOT NULL, archived_artifacts_path TEXT, state TEXT NOT NULL DEFAULT 'pending', -- pending|executing|complete|failed created_at TEXT NOT NULL DEFAULT ..., completed_at TEXT ); ``` The spec's safety guarantees state: > Corrections always: > - Create a new attempt revision (increment `plan.attempt`) > - Preserve old artifacts for diff/compare The `CorrectionAttemptRecord` domain model and `CorrectionAttemptModel` DB model are both implemented. The `CorrectionAttemptRepository` in `src/cleveragents/infrastructure/database/repositories.py` provides `create`, `get`, `list_by_plan`, `update`, and `delete` methods. ### Actual Behavior `CorrectionService.__init__` stores all state in plain Python dicts: ```python def __init__(self, ...): self._corrections: dict[str, CorrectionRequest] = {} self._impacts: dict[str, CorrectionImpact] = {} self._attempts: dict[str, list[CorrectionAttempt]] = {} self._results: dict[str, CorrectionResult] = {} ``` The service **never calls** `CorrectionAttemptRepository.create()` or any other repository method. The `CorrectionAttemptRecord` domain model (which maps to the DB table) is never instantiated by `CorrectionService`. The `CorrectionAttempt` objects tracked in `self._attempts` are a different, simpler model that is never persisted. The `CorrectionService` is wired in `application/container.py` with only `checkpoint_service` and `event_bus` — no repository is injected. ### Code Location - `src/cleveragents/application/services/correction_service.py`, lines 101–111 (constructor — no repository) - `src/cleveragents/application/container.py`, lines 884–888 (container wiring — no repository injected) - `src/cleveragents/infrastructure/database/repositories.py` — `CorrectionAttemptRepository` exists but is never used by `CorrectionService` ### Impact - Correction history is lost on process restart — users cannot query past corrections - `agents plan diff --correction <CORRECTION_ATTEMPT_ID>` cannot look up correction attempts from the DB - The spec's ULID tracking for correction attempts is non-functional (ULIDs are generated but never stored) - Audit trail for corrections is absent - The `correction_attempts` table is never written to during normal operation --- **Automated by CleverAgents Bot** Supervisor: UAT Testing | Agent: uat-tester

HAL9000 referenced this issue

2026-04-08 20:20:23 +00:00

[Automated] CleverAgents Build Session - 2026-04-08 #4799

HAL9000 referenced this issue

2026-04-08 20:22:25 +00:00

[Automated] CleverAgents Build Session - 2026-04-08 #4799

HAL9000 added the

labels

2026-04-08 22:59:16 +00:00

HAL9000 added this to the v3.3.0 milestone

2026-04-08 23:01:01 +00:00

HAL9000 commented

2026-04-08 23:01:12 +00:00

Author

Owner

Issue triaged by project owner:

State: Verified
Priority: High — correction history is lost on process restart; correction_attempts table is never written to; agents plan diff --correction cannot function; audit trail is absent
Milestone: v3.3.0 — Core correction persistence is required for the Corrections milestone acceptance criteria
Story Points: 5 — L — requires injecting CorrectionAttemptRepository into CorrectionService, wiring in container.py, and updating all correction flows to persist attempts
MoSCoW: Should Have — the repository and DB model already exist; this is a wiring gap that makes the correction audit trail non-functional; important for v3.3.0 completion
Parent Epic: Correction Model epic (v3.3.0)

Automated by CleverAgents Bot
Supervisor: Project Owner | Agent: project-owner

Issue triaged by project owner: - **State**: Verified - **Priority**: High — correction history is lost on process restart; `correction_attempts` table is never written to; `agents plan diff --correction` cannot function; audit trail is absent - **Milestone**: v3.3.0 — Core correction persistence is required for the Corrections milestone acceptance criteria - **Story Points**: 5 — L — requires injecting `CorrectionAttemptRepository` into `CorrectionService`, wiring in `container.py`, and updating all correction flows to persist attempts - **MoSCoW**: Should Have — the repository and DB model already exist; this is a wiring gap that makes the correction audit trail non-functional; important for v3.3.0 completion - **Parent Epic**: Correction Model epic (v3.3.0) --- **Automated by CleverAgents Bot** Supervisor: Project Owner | Agent: project-owner

freemo added a new dependency

2026-04-08 23:41:59 +00:00

#4959 EPIC: Correction Engine — Revert & Append Modes with plan.attempt Tracking

freemo referenced this issue

2026-04-08 23:43:03 +00:00

LEGENDARY: Decision Intelligence — Plan Decisions, Corrections, Subplans & Checkpoints #4942

HAL9000 added the

Points

label

2026-04-09 05:42:41 +00:00

HAL9000 commented

2026-04-09 05:43:05 +00:00

Author

Owner

Label compliance fix applied:

Added missing label: Points/3 (M — medium complexity)
Reason: Issue is in State/Verified but was missing a story points estimate. Estimated as Points/3 (M) based on single-area bug fix with moderate complexity.

Automated by CleverAgents Bot
Supervisor: Backlog Grooming | Agent: backlog-groomer

Label compliance fix applied: - Added missing label: `Points/3` (M — medium complexity) - Reason: Issue is in `State/Verified` but was missing a story points estimate. Estimated as Points/3 (M) based on single-area bug fix with moderate complexity. --- **Automated by CleverAgents Bot** Supervisor: Backlog Grooming | Agent: backlog-groomer