QuickLab-MCP 0.1.0__py3-none-any.whl

quicklab_mcp-0.1.0.dist-info/METADATA

@@ -0,0 +1,540 @@
+Metadata-Version: 2.4
+Name: QuickLab-MCP
+Version: 0.1.0
+Summary: QuickLab MCP Server - Production Tabular ML Experimentation Control Plane for AI Agents
+Author: QuickLab Team
+License-Expression: Apache-2.0
+Project-URL: Homepage, https://github.com/Aldrin-Joan/QuickLab-MCP
+Project-URL: Repository, https://github.com/Aldrin-Joan/QuickLab-MCP.git
+Project-URL: Issues, https://github.com/Aldrin-Joan/QuickLab-MCP/issues
+Project-URL: Documentation, https://github.com/Aldrin-Joan/QuickLab-MCP#readme
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: Intended Audience :: Science/Research
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.13
+Description-Content-Type: text/markdown
+Requires-Dist: fastmcp==4.0.5
+Requires-Dist: pydantic<3.0.0,>=2.7.0
+Requires-Dist: pydantic-settings<3.0.0,>=2.2.0
+Requires-Dist: sqlalchemy[asyncio]<3.0.0,>=2.0.30
+Requires-Dist: asyncpg>=0.29.0
+Requires-Dist: redis>=5.0.4
+Requires-Dist: boto3>=1.34.0
+Requires-Dist: aioboto3>=13.0.0
+Requires-Dist: botocore>=1.34.0
+Requires-Dist: pyarrow>=16.0.0
+Requires-Dist: pandas>=2.2.0
+Requires-Dist: numpy>=1.26.0
+Requires-Dist: scikit-learn>=1.5.0
+Requires-Dist: xgboost>=2.0.3
+Requires-Dist: lightgbm>=4.3.0
+Requires-Dist: catboost>=1.2.5
+Requires-Dist: joblib>=1.4.0
+Requires-Dist: httpx>=0.27.0
+Requires-Dist: pyjwt[crypto]>=2.8.0
+Requires-Dist: cryptography>=42.0.0
+Requires-Dist: alembic>=1.13.0
+Requires-Dist: opentelemetry-api>=1.24.0
+Requires-Dist: opentelemetry-sdk>=1.24.0
+Requires-Dist: opentelemetry-exporter-otlp-proto-grpc>=1.24.0
+Requires-Dist: uvicorn>=0.30.0
+Provides-Extra: dev
+Requires-Dist: pytest>=8.2.0; extra == "dev"
+Requires-Dist: pytest-asyncio>=0.23.0; extra == "dev"
+Requires-Dist: pytest-cov>=5.0.0; extra == "dev"
+Requires-Dist: aiosqlite>=0.20.0; extra == "dev"
+Requires-Dist: ruff>=0.4.0; extra == "dev"
+Requires-Dist: mypy>=1.10.0; extra == "dev"
+Requires-Dist: types-redis>=4.6.0; extra == "dev"
+Requires-Dist: types-boto3>=1.34.0; extra == "dev"
+Requires-Dist: types-PyYAML>=6.0.0; extra == "dev"
+
+# QuickLab MCP
+
+[![Python Version](https://img.shields.io/badge/python-3.13%2B-blue.svg)](https://www.python.org/downloads/)
+[![PyPI](https://img.shields.io/badge/pypi-QuickLab--MCP-blue.svg)](https://pypi.org/project/QuickLab-MCP/)
+[![MCP Protocol](https://img.shields.io/badge/MCP%20Protocol-2026--07--28-orange.svg)](https://modelcontextprotocol.io/)
+[![FastMCP](https://img.shields.io/badge/FastMCP-4.0.5%20GA-green.svg)](https://github.com/jlowin/fastmcp)
+[![Tests](https://img.shields.io/badge/tests-104%20passed%20%7C%20100%25-brightgreen.svg)]()
+[![Code Style](https://img.shields.io/badge/code%20style-ruff-black.svg)](https://github.com/astral-sh/ruff)
+[![License](https://img.shields.io/badge/license-Apache%202.0-blue.svg)](LICENSE)
+[![Security](https://img.shields.io/badge/security-OAuth%202.1%20%2F%20OIDC-purple.svg)]()
+
+**QuickLab MCP** (formerly ModelLab) is an enterprise-grade Machine Learning Experimentation Control Plane exposed via the **Model Context Protocol (MCP)**. Engineered with FastMCP 4 and conforming to the standard MCP baseline (`2026-07-28`), QuickLab enables autonomous AI agents and human data scientists to discover approved architectures, validate tabular datasets, submit reproducible training experiments, inspect predictions/metrics, and perform deep diagnostic analysis—**with zero arbitrary code execution**.
+
+---
+
+## Key Highlights
+
+- **Zero Placeholders, Real Tabular ML**: Fully implemented, sandboxed training algorithms for all 7 approved tabular architectures: **Logistic Regression**, **Random Forest**, **XGBoost**, **LightGBM**, **CatBoost**, **Linear SVM**, and **Multi-Layer Perceptron (MLP)**.
+- **23 MCP Tools, 3 Resources, 2 Prompts**: Comprehensive tool catalog across Projects, Discovery, Ingestion, Experimentation, Artifacts, and Diagnostics.
+- **Strict Leak-Free Preprocessing**: Preprocessing transforms (median/mode imputation, standard scaling, one-hot encoding) are fitted strictly on training folds and serialized alongside models.
+- **Dual Production Transports**:
+  - **STDIO Transport**: Pure JSON-RPC on standard output (`sys.stdout`) with all structured logging isolated strictly to `sys.stderr`.
+  - **Streamable HTTP Transport**: Modern HTTP transport mounted at `/mcp` with Bearer auth middleware and operational probes (`/health/live`, `/health/ready`).
+- **OAuth 2.1 & Multi-Tenant Security**: Tenant boundary enforcement, RBAC matrix (`Viewer`, `Researcher`, `Operator`, `Admin`), token scope attenuation, and anti-enumeration protections (HTTP 404 instead of 403 on cross-tenant probes).
+- **Transactional Outbox Architecture**: Guarantees zero lost experiment submissions using PostgreSQL transactional Outbox events dispatched to Redis queues.
+- **S3 / MinIO Object Storage**: Immutable tenant-isolated artifact storage with SHA-256 integrity verification, presigned download URLs, and offline memory fallback.
+- **Full Test Coverage**: **104 tests passing (100% green)** across unit, integration, security matrix, MCP contract, and full end-to-end experiment lifecycle.
+
+---
+
+## Architectural Topology
+
+```mermaid
+flowchart TD
+    subgraph Clients["AI Agents & IDE Interfaces"]
+        Claude["Claude Desktop / VS Code / Cursor"]
+        Autonomous["Autonomous Agent Mesh"]
+    end
+
+    subgraph ControlPlane["ModelLab Control Plane (FastMCP 4)"]
+        HTTPGateway["Streamable HTTP Gateway (/mcp)"]
+        STDIOGateway["STDIO Transport (Local IPC)"]
+        AuthMiddleware["OAuth 2.1 / OIDC Bearer Validator"]
+        RateLimiter["Sliding-Window Rate Limiter"]
+        ToolRouter["21 MCP Tools & Resource Engine"]
+    end
+
+    subgraph Persistence["Persistence & Message Broker"]
+        Postgres[(PostgreSQL 16\nSchema & Outbox Events)]
+        Redis[(Redis 7\nQueue & Rate Limit Window)]
+        MinIO[(MinIO / S3\nDatasets & Model Artifacts)]
+    end
+
+    subgraph WorkerLayer["Sandboxed Worker Execution"]
+        WorkerEngine["Tabular ML Execution Runner"]
+        Preproc["Leak-Free Preprocessing Pipeline"]
+        Trainers["7 Approved Tabular Trainers\n(RF, XGB, LGBM, CatB, LR, SVM, MLP)"]
+    end
+
+    subgraph Telemetry["Observability & Diagnostics"]
+        OTEL["OpenTelemetry Collector"]
+        JSONStderr["Structured JSON Logs (stderr)"]
+    end
+
+    Claude -->|STDIO| STDIOGateway
+    Autonomous -->|HTTP Bearer| HTTPGateway
+    HTTPGateway --> AuthMiddleware
+    AuthMiddleware --> RateLimiter
+    RateLimiter --> ToolRouter
+    STDIOGateway --> ToolRouter
+
+    ToolRouter --> Postgres
+    ToolRouter --> Redis
+    ToolRouter --> MinIO
+
+    Postgres -->|Transactional Outbox| Redis
+    Redis --> WorkerEngine
+    WorkerEngine --> Preproc
+    Preproc --> Trainers
+    Trainers --> MinIO
+    Trainers --> Postgres
+
+    ControlPlane --> JSONStderr
+    ControlPlane --> OTEL
+    WorkerEngine --> OTEL
+```
+
+---
+
+## Complete MCP Tool Catalog
+
+ModelLab exposes 23 granular tools across 6 functional categories:
+
+### 1. Project Management Tools
+
+| Tool Name | Parameters | Required Scope | Description |
+| :--- | :--- | :--- | :--- |
+| `create_project` | `name`, `description` *(opt)* | `ml:projects:write` | Create a new project workspace under the caller's tenant boundary. |
+| `list_projects` | `limit` *(opt)*, `offset` *(opt)* | `ml:projects:read` | List active projects accessible within the caller's tenant boundary. |
+
+### 2. Discovery Tools
+
+| Tool Name | Parameters | Required Scope | Description |
+| :--- | :--- | :--- | :--- |
+| `list_models` | *none* | `ml:models:read` | List all approved tabular ML models in the catalog. |
+| `get_model` | `model_id` | `ml:models:read` | Inspect hyperparameter schemas, bounds, and container digests for a model. |
+| `list_model_versions` | `model_id` | `ml:models:read` | List registered immutable versions for a specific model family. |
+| `list_datasets` | `project_id` *(opt)* | `ml:datasets:read` | List datasets available within the caller's tenant/project boundary. |
+| `get_dataset` | `dataset_id` | `ml:datasets:read` | Retrieve metadata, format, and version history for a specific dataset. |
+| `list_dataset_versions` | `dataset_id` | `ml:datasets:read` | List immutable versions, row counts, and SHA-256 hashes for a dataset. |
+
+### 3. Dataset Management & Validation Tools
+
+| Tool Name | Parameters | Required Scope | Description |
+| :--- | :--- | :--- | :--- |
+| `register_dataset` | `project_id`, `name`, `description`, `format`, `data_base64`, `version` | `ml:datasets:write` | Ingest and validate a tabular dataset (CSV/Parquet/JSONL), store as Parquet, and record content hash. |
+| `validate_dataset` | `format`, `data_base64` | `ml:datasets:read` | Validate raw dataset bytes against schema rules and row limits without persisting. |
+| `inspect_dataset` | `dataset_id`, `version`, `sample_rows` | `ml:datasets:read` | Inspect schema column types, missingness, and head preview rows. |
+
+### 4. Experimentation Tools
+
+| Tool Name | Parameters | Required Scope | Description |
+| :--- | :--- | :--- | :--- |
+| `create_experiment` | `project_id`, `dataset_version_id`, `model_version_id`, `task_type`, `target_column`, `feature_columns`, `primary_metric`, `additional_metrics`, `hyperparameters`, `random_seed`, `idempotency_key` | `ml:experiments:create` | Formulate, validate (target/feature columns, model bounds), and queue an ML experiment with transactional Outbox event. |
+| `get_experiment` | `experiment_id` | `ml:experiments:read` | Retrieve current lifecycle state (`QUEUED`, `RUNNING`, `SUCCEEDED`, `FAILED`, `CANCELLED`), run counts, timestamps, and error details (`error_message`, `error_type`) if failed. |
+| `cancel_experiment` | `experiment_id` | `ml:experiments:cancel` | Gracefully cancel an active or queued experiment run. |
+| `list_experiments` | `project_id`, `status`, `limit`, `offset` | `ml:experiments:read` | Paginated listing of experiments within a project with optional status filter. |
+
+### 5. Results & Artifacts Tools
+
+| Tool Name | Parameters | Required Scope | Description |
+| :--- | :--- | :--- | :--- |
+| `get_experiment_metrics` | `experiment_id` | `ml:experiments:read` | Retrieve computed metrics across `train` and `validation` splits. |
+| `get_experiment_predictions` | `experiment_id`, `limit` | `ml:artifacts:read` | Retrieve sample prediction records (`y_true`, `y_pred`, `y_proba_*`) and presigned download URL. |
+| `list_experiment_artifacts` | `experiment_id` | `ml:artifacts:read` | List all persisted artifacts (`model.joblib`, `preprocessor.joblib`, `predictions.parquet`, `metrics.json`). |
+| `read_experiment_artifact` | `artifact_id` | `ml:artifacts:read` | Retrieve artifact metadata and secure presigned S3 download URL. |
+| `compare_experiments` | `experiment_ids` | `ml:experiments:read` | Side-by-side comparison of validation metrics and hyperparameters across runs. |
+
+### 6. Structured Diagnostic Analysis Tools
+
+| Tool Name | Parameters | Required Scope | Description |
+| :--- | :--- | :--- | :--- |
+| `analyze_experiment` | `experiment_id` | `ml:analysis:create` | Automated diagnostic report evaluating generalization gap, overfitting severity, and data leakage risks. |
+| `analyze_model_errors` | `experiment_id` | `ml:analysis:create` | Deep error slice analysis including residual skew, false positives vs false negatives, and anomaly counts. |
+| `check_experiment_validity` | `experiment_id` | `ml:analysis:create` | Automated verification of experiment validity, dataset lineage, and metric reproducibility. |
+
+---
+
+## MCP Resources & Prompts
+
+### Resources
+- `modellab://models`: Static resource cataloging all approved model architectures.
+- `modellab://models/{model_id}`: Parameterized template for full model hyperparameter schemas.
+- `modellab://experiments/{experiment_id}/metrics`: Parameterized template for experiment metrics.
+
+### Prompts
+- `experiment_design`: Guided architectural workflow prompting agents to formulate sound hypotheses, choose candidate models, and select non-leaking split strategies.
+- `error_diagnosis`: Structured diagnostic workflow assisting agents in diagnosing training regressions, severe class imbalances, or generalization gaps.
+
+---
+
+## Approved Model Families & Hyperparameter Bounds
+
+| Model Family | `model_id` | Supported Task Types | Key Hyperparameters (Strictly Validated) |
+| :--- | :--- | :--- | :--- |
+| **Random Forest** | `random_forest` | Binary / Multiclass Classification, Regression | `n_estimators` [10–1000], `max_depth` [1–50], `min_samples_split` [2–20] |
+| **XGBoost** | `xgboost` | Binary / Multiclass Classification, Regression | `n_estimators` [10–2000], `max_depth` [1–16], `learning_rate` [0.001–1.0], `subsample` [0.1–1.0] |
+| **LightGBM** | `lightgbm` | Binary / Multiclass Classification, Regression | `n_estimators` [10–2000], `num_leaves` [7–256], `learning_rate` [0.001–1.0] |
+| **CatBoost** | `catboost` | Binary / Multiclass Classification, Regression | `iterations` [10–2000], `depth` [1–12], `learning_rate` [0.001–1.0] |
+| **Logistic Regression** | `logistic_regression` | Binary / Multiclass Classification | `C` [0.001–1000.0], `penalty` (`l1`, `l2`, `elasticnet`, `none`), `solver` |
+| **Linear SVM** | `linear_svm` | Binary / Multiclass Classification, Regression | `C` [0.001–100.0], `loss` (`hinge`, `squared_hinge`), `max_iter` [100–10000] |
+| **Multi-Layer Perceptron** | `mlp` | Binary / Multiclass Classification, Regression | `hidden_layer_sizes`, `activation` (`relu`, `tanh`), `alpha` [1e-6–1.0], `learning_rate_init` |
+
+---
+
+## Quickstart
+
+### Prerequisites
+- Python 3.13+ (or Conda environment `ML_LLM`)
+- Docker & Docker Compose (optional, for local services)
+
+### 1. Installation
+
+#### Instant Execution via `uvx` (Recommended)
+No cloning required — run QuickLab directly with Astral `uv`:
+```bash
+uvx quicklab-mcp
+```
+
+#### Local Development Installation
+```bash
+# Clone the repository
+git clone https://github.com/Aldrin-Joan/QuickLab-MCP.git
+cd QuickLab-MCP
+
+# Install dependencies using uv
+uv sync
+```
+
+### 2. Run the Full Test Suite
+
+Verify all 75 automated tests across unit, integration, contract, security, and e2e suites:
+
+```bash
+pytest tests/ -v
+```
+
+Output:
+```text
+================= 75 passed, 7 warnings in 187.73s (0:03:07) ==================
+```
+
+### 3. Launch Local Infrastructure (Optional)
+
+Spin up PostgreSQL 16, Redis 7, MinIO S3, and OpenTelemetry Collector:
+
+```bash
+docker compose up -d
+```
+
+Apply database migrations:
+
+```bash
+alembic upgrade head
+```
+
+### 4. Running ModelLab
+
+#### Option A: Streamable HTTP Mode (Production & Remote Agents)
+
+```bash
+python -m ml_mcp.server.http
+```
+
+Access operational endpoints:
+- Liveness Probe: `http://127.0.0.1:8000/health/live`
+- Readiness Probe: `http://127.0.0.1:8000/health/ready`
+- MCP Endpoint: `http://127.0.0.1:8000/mcp`
+
+#### Option B: STDIO Transport Mode (IDE & Desktop Agents)
+
+```bash
+python -m ml_mcp.server.stdio
+```
+
+---
+
+## MCP Client Configuration
+
+ModelLab supports both **STDIO** (local subprocess IPC) and **Streamable HTTP** (networked SSE + POST) transports across both **Docker** and **UV / local Python**.
+
+Select your preferred runtime configuration below:
+
+### 1. Docker + STDIO
+Runs ModelLab inside an isolated Docker container with standard I/O (`sys.stdin` / `sys.stdout`) piped directly to the MCP client.
+
+```json
+{
+  "mcpServers": {
+    "modellab": {
+      "command": "docker",
+      "args": [
+        "run",
+        "-i",
+        "--rm",
+        "--entrypoint", "python",
+        "--network", "host",
+        "-e", "APP_ENV=production",
+        "-e", "DATABASE_URL=postgresql+asyncpg://postgres:postgres@localhost:5433/modellab",
+        "-e", "REDIS_URL=redis://localhost:6379/0",
+        "-e", "OBJECT_STORE_ENDPOINT_URL=http://localhost:9000",
+        "-e", "OBJECT_STORE_ACCESS_KEY_ID=minioadmin",
+        "-e", "OBJECT_STORE_SECRET_ACCESS_KEY=minioadmin",
+        "-e", "OBJECT_STORE_BUCKET_NAME=modellab-artifacts",
+        "modellab-api",
+        "-m", "ml_mcp.server.stdio"
+      ]
+    }
+  }
+}
+```
+
+> **Note for Docker Desktop (macOS / Windows)**: If `--network host` is unavailable in your Docker Desktop environment, connect via the shared Docker network:
+> ```json
+> "args": [
+>   "run", "-i", "--rm",
+>   "--entrypoint", "python",
+>   "--network", "modellab_default",
+>   "-e", "DATABASE_URL=postgresql+asyncpg://postgres:postgres@postgres:5432/modellab",
+>   "-e", "REDIS_URL=redis://redis:6379/0",
+>   "-e", "OBJECT_STORE_ENDPOINT_URL=http://minio:9000",
+>   "-e", "OBJECT_STORE_ACCESS_KEY_ID=minioadmin",
+>   "-e", "OBJECT_STORE_SECRET_ACCESS_KEY=minioadmin",
+>   "-e", "OBJECT_STORE_BUCKET_NAME=modellab-artifacts",
+>   "modellab-api",
+>   "-m", "ml_mcp.server.stdio"
+> ]
+> ```
+
+---
+
+### 2. Docker + HTTP
+Connects to the ModelLab API service running inside Docker Compose (`docker compose -f docker-compose.http.yml up -d`).
+
+```json
+{
+  "mcpServers": {
+    "modellab": {
+      "url": "http://localhost:8000/mcp",
+      "headers": {
+        "Authorization": "Bearer <YOUR_JWT_BEARER_TOKEN>"
+      }
+    }
+  }
+}
+```
+
+---
+
+### 3. UVX + STDIO (Recommended Turnkey Setup)
+Runs QuickLab globally via `uvx` directly from PyPI:
+
+```json
+{
+  "mcpServers": {
+    "quicklab": {
+      "command": "uvx",
+      "args": ["quicklab-mcp"],
+      "env": {
+        "APP_ENV": "production",
+        "DATABASE_URL": "postgresql+asyncpg://postgres:postgres@localhost:5433/modellab",
+        "REDIS_URL": "redis://localhost:6379/0",
+        "OBJECT_STORE_ENDPOINT_URL": "http://localhost:9000",
+        "OBJECT_STORE_ACCESS_KEY_ID": "minioadmin",
+        "OBJECT_STORE_SECRET_ACCESS_KEY": "minioadmin",
+        "OBJECT_STORE_BUCKET_NAME": "modellab-artifacts"
+      }
+    }
+  }
+}
+```
+
+Or for local source development with `uv`:
+```json
+{
+  "mcpServers": {
+    "quicklab": {
+      "command": "uv",
+      "args": [
+        "run",
+        "--directory", "/absolute/path/to/QuickLab-MCP",
+        "python", "-m", "ml_mcp.server.stdio"
+      ],
+      "env": {
+        "APP_ENV": "production",
+        "DATABASE_URL": "postgresql+asyncpg://postgres:postgres@localhost:5433/modellab",
+        "REDIS_URL": "redis://localhost:6379/0",
+        "OBJECT_STORE_ENDPOINT_URL": "http://localhost:9000",
+        "OBJECT_STORE_ACCESS_KEY_ID": "minioadmin",
+        "OBJECT_STORE_SECRET_ACCESS_KEY": "minioadmin",
+        "OBJECT_STORE_BUCKET_NAME": "modellab-artifacts"
+      }
+    }
+  }
+}
+```
+
+---
+
+### 4. UV + HTTP
+Starts the ModelLab HTTP service locally with `uv` and connects client agents via Streamable HTTP.
+
+**1. Launch the server:**
+```bash
+uv run python -m ml_mcp.server.http
+```
+
+**2. Client Configuration (`mcp.json` / `claude_desktop_config.json`):**
+```json
+{
+  "mcpServers": {
+    "modellab": {
+      "url": "http://localhost:8000/mcp",
+      "headers": {
+        "Authorization": "Bearer <YOUR_JWT_BEARER_TOKEN>"
+      }
+    }
+  }
+}
+```
+
+---
+
+### Generating an OAuth 2.1 Bearer Token (for HTTP Modes)
+
+Generate a cryptographically signed JWT token for authentication in HTTP mode:
+
+```bash
+uv run python -c "
+from ml_mcp.server.auth import TokenValidator
+from ml_mcp.domain.policies import Role, Scope
+
+validator = TokenValidator()
+token = validator.create_access_token(
+    principal_id='agent-user',
+    tenant_id='default-tenant',
+    role=Role.ADMIN,
+    scopes=[s.value for s in Scope],
+)
+print('Bearer Token:\n' + token)
+"
+```
+
+---
+
+## Security & Access Control (RBAC)
+
+ModelLab implements a zero-trust security architecture adhering to OAuth 2.1:
+
+| Scope | Viewer | Researcher | Operator | Admin |
+| :--- | :---: | :---: | :---: | :---: |
+| `ml:models:read` | ✅ | ✅ | ✅ | ✅ |
+| `ml:datasets:read` | ✅ | ✅ | ✅ | ✅ |
+| `ml:datasets:write` | ❌ | ✅ | ✅ | ✅ |
+| `ml:experiments:read` | ✅ | ✅ | ✅ | ✅ |
+| `ml:experiments:create` | ❌ | ✅ | ✅ | ✅ |
+| `ml:experiments:cancel` | ❌ | ✅ | ✅ | ✅ |
+| `ml:analysis:create` | ❌ | ✅ | ✅ | ✅ |
+| `ml:artifacts:read` | ✅ | ✅ | ✅ | ✅ |
+| `ml:worker:operate` | ❌ | ❌ | ✅ | ✅ |
+| `ml:experiment:override` | ❌ | ❌ | ❌ | ✅ |
+
+- **Scope Attenuation**: Access tokens carrying specific scopes attenuate privileges below role defaults.
+- **Anti-Enumeration Defense**: Requests attempting cross-tenant resource access consistently return HTTP 404 (`ResourceNotFoundError`) rather than 403, preventing attackers from probing valid resource IDs.
+- **Sensitive Key Redaction**: Telemetry and logs automatically scrub sensitive credentials (`password`, `secret`, `token`, `key`, `authorization`).
+
+---
+
+## Production Containerization
+
+Multi-stage [Dockerfile](Dockerfile) provides hardened images running as unprivileged users:
+
+```bash
+# Build API Server container
+docker build --target ml-mcp-api -t modellab/ml-mcp-api:latest .
+
+# Build Sandboxed Worker container
+docker build --target ml-worker-tabular-cpu -t modellab/ml-worker-tabular-cpu:latest .
+```
+
+Both container images run with read-only root filesystems and non-root users (`uid 10001` / `uid 10002`).
+
+---
+
+## Project Repository Structure
+
+```text
+ModelLab/
+├── .github/workflows/ci.yml       # Automated CI/CD pipeline (lint, test, build)
+├── alembic/versions/              # Database migration scripts
+├── deploy/docker/                 # OTEL Collector and container deployment configs
+├── Docs/                          # Architecture, Tech-Stack & Operational Guide
+├── src/ml_mcp/                    # Core ModelLab application source
+│   ├── application/               # Application services (models, datasets, analysis)
+│   ├── config/                    # Pydantic Settings & environment parsing
+│   ├── domain/                    # Domain models, policies, RBAC & error hierarchy
+│   ├── infrastructure/            # Postgres ORM, Redis queue, S3 storage, OTEL
+│   ├── mcp/                       # FastMCP 4 tools, schemas, resources & prompts
+│   ├── server/                    # Server factory, auth validator, HTTP & STDIO
+│   └── workers/                   # Preprocessing, ML trainers & execution runner
+├── tests/                         # 75 Automated Test Cases
+│   ├── contract/                  # FastMCP tool contract compliance tests
+│   ├── e2e/                       # Full ML experiment lifecycle end-to-end tests
+│   ├── integration/               # Database, Outbox, Redis & Worker integration
+│   ├── security/                  # Multi-tenant isolation & RBAC security matrix
+│   └── unit/                      # Fast unit tests across all domain layers
+├── docker-compose.yml             # Local PostgreSQL, Redis, MinIO & OTEL stack
+├── Dockerfile                     # Multi-stage production container build
+├── pyproject.toml                 # Package definition and locked dependencies
+└── README.md                      # Project documentation
+```
+
+---
+
+## License
+
+ModelLab is licensed under the Apache License, Version 2.0. See [LICENSE](LICENSE) for details.

quicklab_mcp-0.1.0.dist-info/RECORD

@@ -0,0 +1,92 @@
+ml_mcp/__init__.py,sha256=5NKCu2Ln5IJ4GP5to3YpPcDUVGLDpVgU4xDUbCstj_M,63
+ml_mcp/application/__init__.py,sha256=G-ZEVx1uiabj6vGQGdMvxqPmkudfqGbb9olEWN5YcNM,33
+ml_mcp/application/analysis/__init__.py,sha256=aFHqMi3rFE-O5GZUCKSzdJuVUpq8u91IILgdhyEe1UA,352
+ml_mcp/application/analysis/diagnostics.py,sha256=8mG6nWlhtV5zx-T95Dq736EuB9hC-yO_VGN1iTH3IeQ,4745
+ml_mcp/application/analysis/service.py,sha256=XlEWSuZdbRZfbda1QqqKtArhQkcd9hJbMXK4Na9xzF8,6691
+ml_mcp/application/artifacts/__init__.py,sha256=85Ge7LiOVi3X6e1DineTas-djVLpEuLaCMENqS5HsEM,133
+ml_mcp/application/artifacts/service.py,sha256=VArkaMzdQc4LTew7nzlMoAWUr8qzQfyPIghTYC5hIpM,5438
+ml_mcp/application/datasets/__init__.py,sha256=FOig5S_BxChC1j5alYm2JtHVSdsS_SiogipPxW0lTOU,229
+ml_mcp/application/datasets/service.py,sha256=7lV1Ulu2YmTVCouWQ3AuCeL6aWmaal__ArQ0gJWM5vs,6413
+ml_mcp/application/datasets/validator.py,sha256=Lzmpg8LMZk8GRtbKlZNoTPUZmmXNyuNOhLcO3c08J0g,6744
+ml_mcp/application/experiments/__init__.py,sha256=OCTOflG16HnaCE7WP2dZOWPGUPg7fJNDccc09Q_Tx8A,141
+ml_mcp/application/experiments/service.py,sha256=1wDTk0vrgAUx405SK9NQ8IXKEnFH5LGzSwODLOgCAK0,9779
+ml_mcp/application/models/__init__.py,sha256=lOf3lQgHkF430D6I4M3qddvnUoc1BFXLdV20JK5kSL0,203
+ml_mcp/application/models/catalog_seed.py,sha256=fbMi5dXwCxyQS5mY8Is0Pc9rVuY-MTG4YnuLkIbtAGY,7556
+ml_mcp/application/models/service.py,sha256=ql3LGQK2TIw_2AokE8IcF1JqmB_XQRRGQWCYgZL3cDg,7705
+ml_mcp/application/projects/__init__.py,sha256=DiYSga6TqMv1XJxjmCATNQ49mJF__TYAMeu-lb9iJGY,130
+ml_mcp/application/projects/service.py,sha256=V2HprpqG_q_sZdU7swF54qIyxVG563lCfK6njdJcDGQ,3333
+ml_mcp/config/__init__.py,sha256=Y2vwEx9LaWwnxVgDdWi4droTS_i9kYhsVOLmlwcCAdU,145
+ml_mcp/config/settings.py,sha256=hHQRz1m0guOLq7ZFNmDDYjHUPzpoF5hIK5myOgO6geI,5700
+ml_mcp/domain/__init__.py,sha256=bS6yxbqgrdZchhheO-d3crM232Bh68v2CIGsJVLsojM,48
+ml_mcp/domain/errors/__init__.py,sha256=pTs3gJriPqaAFI7MvVyRGJ23_PpzLnDREtpDbwfpZhI,1055
+ml_mcp/domain/errors/base.py,sha256=CshLpNOU7aMMbtGeK942nN4L2NfXVyX8hX65fJHxc2A,5491
+ml_mcp/domain/errors/codes.py,sha256=mbxQagM4qBt4G1iA__2s8Ty5-SZrIP2zGSoGIQYIcnM,943
+ml_mcp/domain/policies/__init__.py,sha256=9k1QWtiX_-YW441mlh3Qs3G3cb-K_ECkDHL3dabyMGc,439
+ml_mcp/domain/policies/rbac.py,sha256=Ek0HQMqjDrpD9oGxU58bpcWpmyD-muFtdVxV4EqXg2E,4198
+ml_mcp/domain/policies/tenant_isolation.py,sha256=PSD3zRAxPO7Z0ywErQQ_LAEVEv1nzLQjON7cTNq-9KU,1493
+ml_mcp/domain/policies/tool_policy.py,sha256=xw0ZWCBgzfNS3zcLnXeI3Ihyurb-J6RqQ4zRLJyIcJk,5806
+ml_mcp/domain/value_objects/__init__.py,sha256=6kWLexj_bLrhymNfHGfl2VGe3fd7rRI8VyDWQ2-ruQA,837
+ml_mcp/domain/value_objects/dataset.py,sha256=0Vzgnl7J_McpaJwUHM9yt1HvaPDTj1Rm6vBB9t4SQSU,1795
+ml_mcp/domain/value_objects/experiment.py,sha256=6VDefA_jADAWFjiBckFyZSV25o33INx3jgtkkzSFjkQ,4276
+ml_mcp/domain/value_objects/model.py,sha256=dozqWALnxZq3iB9JeOWX0b-hoOouFh7vnEZiSu_Ms9o,1447
+ml_mcp/infrastructure/__init__.py,sha256=D_StCHKRqyvxeKpP-_lrddfbBbJe75K86cT5biZvmcw,36
+ml_mcp/infrastructure/object_storage/__init__.py,sha256=u4lX7ltzW6xqD7-VEL8hSE8mhYp7FTm32fPVYM0qXbc,176
+ml_mcp/infrastructure/object_storage/s3.py,sha256=ZAS1DVqOrs-zdGBrUOUDBVQ7SQ1tb3L7LuJT0JUpZ3s,11518
+ml_mcp/infrastructure/postgres/__init__.py,sha256=Eh9Vp9-pjlBI33g_vXD29fOcXr_IvkPL-J73C7f9opg,395
+ml_mcp/infrastructure/postgres/base.py,sha256=UaBIzrNt4PCbMpllhVl_kU42lehH0OaKPS9Sc3Mcmaw,1060
+ml_mcp/infrastructure/postgres/models.py,sha256=Y2thMyPpIapWerEXh83rJj6GWjN5EGis75KC6rsTGcE,16206
+ml_mcp/infrastructure/postgres/session.py,sha256=8rrT23q0dfEN659IxJJLhe0wc5BP6zMmhWACKQX4KCc,3908
+ml_mcp/infrastructure/postgres/repositories/__init__.py,sha256=cqMMMLyDG7XuvWt0OG3gxgjUsMXXdEwbxiGePxWrQ7g,911
+ml_mcp/infrastructure/postgres/repositories/artifacts.py,sha256=P8CpQl-T2EEgiXJ5A5hdF4SNFkVoWcOQ9O5kLBjPJgY,1273
+ml_mcp/infrastructure/postgres/repositories/audit.py,sha256=R2FFOOC4wN6j1tejDxvTu3NutA_ClAERs9_3z9LoGW8,981
+ml_mcp/infrastructure/postgres/repositories/datasets.py,sha256=0znA8JSyb3ycxc44jaehixvlv85aZZY5MrBshb2pif0,3236
+ml_mcp/infrastructure/postgres/repositories/experiments.py,sha256=TWn8lZ4wzll7Vf4hfa2VY5dVWchK0qneT1bd1h4sZMA,4792
+ml_mcp/infrastructure/postgres/repositories/metrics.py,sha256=xeEPeevxyUzGDD6S3ZxTkQnpPiR6T5fhhjv6schuUdQ,2139
+ml_mcp/infrastructure/postgres/repositories/models.py,sha256=xRzlRqrbdZ9Zcj8Wq7EZLepldu0caccYJ00xdiQl6tw,2306
+ml_mcp/infrastructure/postgres/repositories/outbox.py,sha256=emKjSHiPqQ46ohWd48D5qD7NiU9pD_FSSo75LbxttdU,1733
+ml_mcp/infrastructure/postgres/repositories/projects.py,sha256=8L1D1x4ze2rNYthI-l2Rvtr5d2auU8PqUbTia54VeL0,3066
+ml_mcp/infrastructure/queue/__init__.py,sha256=n2LacJamnQuge-1uQOb422x6OraFWOjVGKjTkYgyavI,253
+ml_mcp/infrastructure/queue/outbox_processor.py,sha256=KlorHH9gAsg88Y0EtPZQleU4tvZWUICifV2MlKm3KNg,2342
+ml_mcp/infrastructure/queue/task_queue.py,sha256=PGDXZhBWSnrotSokA_0e1u5p3VEzMYLmRChLHoWQkUQ,3249
+ml_mcp/infrastructure/redis/__init__.py,sha256=I6krHFx5_D1ToYV7lBsF4gQE2OObHIrZuaG3f2lil30,343
+ml_mcp/infrastructure/redis/client.py,sha256=q8eKG1BMggwhgSQyDHlWAYCtle7ECR3Dh_RecPP8Lc4,1926
+ml_mcp/infrastructure/redis/rate_limiter.py,sha256=ahLPYceisKhRqDnxlGzXQiW8gZ9ZV685YhWLcpdVXf0,4280
+ml_mcp/infrastructure/telemetry/__init__.py,sha256=eGtXhRaYpBCUdxnaGhjqf6jYqVfayq-VTLUYtGyYGwo,465
+ml_mcp/infrastructure/telemetry/logging.py,sha256=pf32yniOGeuVe8Xu9sTeZJn5ZCTUSRbZEm7sLQ5K6fc,4718
+ml_mcp/infrastructure/telemetry/otel.py,sha256=K3ChWsaaS4KOheuwJZudXLCiWVoBtHlEGIOMcKtEv1I,5565
+ml_mcp/mcp/__init__.py,sha256=6VQq-epTZC9ttcgn8xp4WsV7Id7tzByWs8phCNrVEOI,38
+ml_mcp/mcp/prompts.py,sha256=dWoEC3GzJmtAmF30C_667pfv20BJqB9CIdNEcfvDVY4,1723
+ml_mcp/mcp/resources.py,sha256=1eJLWjPrPzYw3nFnQkOD6Ww6eRLGIFzVaUWGIlM1xzg,1389
+ml_mcp/mcp/schemas/__init__.py,sha256=wCjDRgWW8J9lIyY2rdoI1n-Uwn1N8QgeX5u6Z--fJ2c,37
+ml_mcp/mcp/schemas/tools.py,sha256=0fFMD32t2dsIFIFa5rjS9Drj4shOV2A3ayOWGYVUyoo,4378
+ml_mcp/mcp/tools/__init__.py,sha256=chI70okepM0zhsxl49zdgW9k-J8Y6Oewyz3eTbLjAEU,1741
+ml_mcp/mcp/tools/analysis.py,sha256=7Zk61-sJEgkAsB319yfOAg4blj9yiq7phKQ0iFSDN-g,1282
+ml_mcp/mcp/tools/datasets.py,sha256=ZgoArz9GK-BuLbAPuel8rb1ZQlroYla9QvG4H7BvOHg,2724
+ml_mcp/mcp/tools/discovery.py,sha256=EBcMDSHFl2OwZfkBiNBDXIHDztmpDBuUnn3XCwavwos,3199
+ml_mcp/mcp/tools/experiments.py,sha256=gGfUtxEtlTiEV6cTa6ltqhu3grjiEMqL5_5m_kGGd8I,3657
+ml_mcp/mcp/tools/projects.py,sha256=a4WM3cwAE-53yEHjU5pHXMScGZ-RK0pe1jst6BKP66Y,1173
+ml_mcp/mcp/tools/results.py,sha256=1HikgbP5kBqjy7E33d_oQT0hNJx6o_uIjnXlUUJp_us,2038
+ml_mcp/server/__init__.py,sha256=IKeTjAkSQmw1FHUytnB2C6Jkn92z4dGeiDbnuDjbiug,33
+ml_mcp/server/app.py,sha256=4J7SsW1saJGqa5qew7e0MWWhmZ3ouAU-ueIGS218lcA,18825
+ml_mcp/server/auth.py,sha256=PlR4eQXbJBDX9waC0xIg-HdbeH9RSTCvskfL5TEZ39o,5293
+ml_mcp/server/context.py,sha256=VhNiDkYJulORb3TmKl5v231RNiK_s4hMadKJlcNsMYc,2015
+ml_mcp/server/http.py,sha256=TiIMQQIXlgC-wD4X4IG1CCrveVtTRGUjyGfMmk7nErs,5977
+ml_mcp/server/middleware.py,sha256=2qOkGjAV4hpqVMwe6eyviKAFAbENxraSMupyaVVhgkc,4251
+ml_mcp/server/stdio.py,sha256=4dL5OD2zi9ftZ5quOe8q0RPpmGN1exu9_577y1NK3QA,1246
+ml_mcp/workers/__init__.py,sha256=TxJ2Un6ky24-Nslw68gM87mpcFvL7T0lV2dGpvGS-w0,34
+ml_mcp/workers/daemon.py,sha256=jVAOC7ac9svDp2yL-ie2oZCz1fbDjCzuEZIAILyq7pQ,5022
+ml_mcp/workers/metrics.py,sha256=HTCZ2PtWHtq4COZUNZZLffzJWU2Zuq7FhetxQcebGIM,3858
+ml_mcp/workers/preprocessing.py,sha256=VmqDUg-9_JneR6VnSp9PFXkIyJ6jRsJdDKNlhirhV28,6430
+ml_mcp/workers/runner.py,sha256=cs8rqs086-mjVnebO41y4USIGScp3mIyOxhRBBxceKA,9156
+ml_mcp/workers/sandbox.py,sha256=FzucmltcgB9Ns0N9q8WE3_hFFLgnWHUErEA8amwkNdg,1176
+ml_mcp/workers/trainers/__init__.py,sha256=db6zP8APwfYh0ePFeBxPYyXnw_giB673oSuHVG0hv1Q,1828
+ml_mcp/workers/trainers/base.py,sha256=td30S54OPhdgFn0yGvzfXdJk9wUZrKue-HVt0si1bgc,1407
+ml_mcp/workers/trainers/catboost_model.py,sha256=q57rS5ciZxwVkD8szpcGSYVV88Hsi8mh8-JZIdQ1pGI,3006
+ml_mcp/workers/trainers/lightgbm_model.py,sha256=HM6OQM1e5tRwWtaR7H8XKmI-qvOkqmrCD-Wn2ZwwBzE,2854
+ml_mcp/workers/trainers/sklearn_models.py,sha256=4lKAd0KVMHethg0j8AxJFie-pHTO1KUQ0HRWaq7Rl6c,9399
+ml_mcp/workers/trainers/xgboost_model.py,sha256=dTmonoqo4wiUwwroYbnZMmFsa8pORfwboRyRTuCnoBs,2940
+quicklab_mcp-0.1.0.dist-info/METADATA,sha256=0ke7J8ZSGngQ0yh35O8GKutBBqUCjaf_2CA9aGwwX0A,23667
+quicklab_mcp-0.1.0.dist-info/WHEEL,sha256=a-udL2m4V_6Yn8b-uKzWb12RjRxwwMo6ohYFM24tpaY,91
+quicklab_mcp-0.1.0.dist-info/entry_points.txt,sha256=7f1Tqr8i6tUIR36fCr0t9cUb41OlRcmlt_xUrwRAA9M,145
+quicklab_mcp-0.1.0.dist-info/top_level.txt,sha256=tSWoPBcibglwAU_c0ODk8ewNwk13EuPSIgzayLv_pPc,7
+quicklab_mcp-0.1.0.dist-info/RECORD,,

quicklab_mcp-0.1.0.dist-info/WHEEL

@@ -0,0 +1,5 @@
+Wheel-Version: 1.0
+Generator: setuptools (84.2.0)
+Root-Is-Purelib: true
+Tag: py3-none-any
+

quicklab_mcp-0.1.0.dist-info/entry_points.txt

@@ -0,0 +1,4 @@
+[console_scripts]
+modellab = ml_mcp.server.stdio:run_stdio
+quicklab = ml_mcp.server.stdio:run_stdio
+quicklab-mcp = ml_mcp.server.stdio:run_stdio

quicklab_mcp-0.1.0.dist-info/top_level.txt

@@ -0,0 +1 @@
+ml_mcp