Metadata-Version: 2.4
Name: flows-put-lines
Version: 1.0.0
Summary: Provision PTV Flows KPIs from GTFS static feeds for entire transit lines
Author-email: PTV Group <luca.paone@ptvgroup.com>
License: MIT
Project-URL: Homepage, https://github.com/lucapaone-ptvgroup/Flows-put-lines
Project-URL: Repository, https://github.com/lucapaone-ptvgroup/Flows-put-lines
Keywords: gtfs,transit,flows,kpi,ptv
Classifier: Development Status :: 4 - Beta
Classifier: Intended Audience :: Developers
Classifier: Topic :: Software Development :: Libraries :: Python Modules
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: pandas>=2.2.0
Requires-Dist: requests>=2.31.0
Requires-Dist: PyYAML>=6.0
Provides-Extra: dev
Requires-Dist: pytest>=7.0; extra == "dev"
Requires-Dist: black>=23.0; extra == "dev"
Requires-Dist: mypy>=1.0; extra == "dev"
Dynamic: license-file

# Flows-put-lines
A Python service that downloads a static GTFS feed and provisions Expected Travel Time KPIs on PTV FLOWS per entire line — one KPI per GTFS direction (`direction_id` 0/1). Lines are selected via `route_short_name`. Alerts use UNUSUAL thresholds with offset `timetostart: 0`.

## Installation

### As a Standalone CLI Tool

Install dependencies and run directly:

```bash
python -m pip install -r requirements.txt
```

### As a Package (for use in other projects)

Install the package in editable mode:

```bash
pip install -e .
```

Or install from the repository:

```bash
pip install git+https://github.com/lucapaone-ptvgroup/Flows-put-lines.git
```

After installation, you can use it programmatically:

```python
from src.gtfs_extractor import download_gtfs, extract_lines_with_shapes
from src.flows_kpi import create_kpi, get_all_kpis
from src.config import load_config

# Load configuration
config = load_config("config.yaml")

# Use the package functions
# ... (see Claude Skill section for examples)
```

Or use the command-line tool:

```bash
flows-put-lines --config config.yaml --apply
```

## Configure
- Edit [config.yaml](config.yaml):
	- `gtfs.url`: your GTFS feed URL (http/https)
	- `gtfs.route_short_names`: list of route short names or leave empty for all
	- `gtfs.max_routes`: limit processing to a small set (e.g., 5)
	- `kpi.thresholds.warning_minutes` and `critical_minutes`
	- Set environment variable `PTV_API_KEY` for PTV Flows access
	- Set `kpi.apply: true` to call the API (default is dry-run)

## Run
- **Dry-run** (prints payloads without calling API):

```bash
python -m src.main --config config.yaml
```

- **Apply** (create KPI instances):

```bash
PTV_API_KEY=your_key_here \
python -m src.main --config config.yaml --apply
```

- **Delete all existing KPIs** (cleanup before re-running):

```bash
PTV_API_KEY=your_key_here \
python -m src.main --config config.yaml --delete-all-kpis
```

### CLI Options
- `--config <path>`: Path to YAML configuration file (default: `config.yaml`)
- `--apply`: Force API calls to create KPIs (overrides `kpi.apply: false` in config)
- `--delete-all-kpis`: Fetch all running KPIs and delete them; exits after completion (useful for cleanup before re-provisioning)

## Notes
- One KPI is created per GTFS direction (`direction_id` 0/1) using the canonical shape direction (first → last stop).
- Endpoints per direction are derived via majority first/last stops across trips (`direction_id` 0/1).
- Very long corridors may need segmentation if the API rejects large PATHs; start with dry-run and adjust as needed.

### Corridor Entities Requirement (Apply mode)
- Creating KPIs with locationType `PATH` requires Flows corridor entities (`location.entities`) — OpenLR-encoded street segments along the path. The API does not persist shape-only locations.
- If your API key does not include Flows Map/Network access, the server cannot map-match your WKT to entities and KPI creation fails with `GENERAL_VALIDATION_ERROR` and `Entity list is empty`.
- To proceed:
	- Request Flows Map/Network API access to enable server-side map-matching of paths to entities, or
	- Provide an existing `locationId` referencing a precomputed corridor, or
	- Supply OpenLR codes per corridor segment in the `entities` field.
- References: OpenLR concept and KPI API samples include `entities` on creation; see the official docs.

## Caching & Stats
- The GTFS download uses basic HTTP caching (`ETag`/`Last-Modified`) to avoid re-downloading when the feed hasn’t changed. A metadata file is stored in `data/gtfs_meta.json`.
- For this use case, you can restrict processing to a few lines via `gtfs.max_routes` and `gtfs.route_short_names`. Updating the GTFS file is only needed when lines or their shapes change; otherwise the cached file is reused.
- The CLI prints a summary with counts of routes processed, directions with shape, and KPI payloads built.
## Map Visualization
After creating KPIs in apply mode, the tool automatically generates an interactive HTML map showing all KPI shapes using **original GTFS shapes** (complete, not trimmed or modified by Route Matcher).

**Map location:** `data/shapes/kpi_map.html`

**Features:**
- 🗺️ Interactive OpenStreetMap base layer with pan and zoom
- 🎨 Color-coded lines by direction:
  - **Blue** (#3388ff) = Direction 0
  - **Orange** (#ff6633) = Direction 1
- 👆 Click any line to see KPI details popup (name, route, direction, KPI ID, status)
- 📊 Info box displays selected KPI information
- 📍 Legend showing direction color coding
- 🔍 Auto-zoom to fit all KPI shapes in view
- ✅ Full GTFS shapes preserved (not trimmed to Route Matcher entities)

**Metadata export:** `data/shapes/shapes_metadata.json` contains all KPI details in JSON format for programmatic access.

**Usage:** Simply open `data/shapes/kpi_map.html` in your browser after running `--apply` to visualize all provisioned KPIs on the map.

**Example test results:**
- Successfully tested with Rome GTFS feed (routes 211, C2, 62)
- 6 KPIs created (3 routes × 2 directions)
- All shapes render completely with no trimming

## Claude Skill / Programmatic Usage

This package includes a **Claude Skill** that allows AI assistants and other Python projects to use the flows-put-lines functionality programmatically.

### Quick Start with Claude Skill

1. **Install the package:**
   ```bash
   pip install -e .
   ```

2. **Set your API key:**
   ```bash
   export PTV_API_KEY="your_api_key_here"
   ```

3. **Use the skill in Python** (prompts for missing `PTV_API_KEY` and GTFS URL):
   ```python
   from claude_skill.flows_put_lines_skill import FlowsPutLinesSkill

   # Initialize the skill
   skill = FlowsPutLinesSkill("config.yaml")

   # List existing KPIs
   kpis = skill.list_kpis()
   print(f"Found {len(kpis)} KPIs")

     # Provision new KPIs from GTFS (if gtfs_url is missing, will prompt)
     results = skill.provision_kpis_from_gtfs(
       gtfs_url="",
       route_short_names=["1", "2", "3"],
       max_routes=3,
       apply=True  # Set False for dry-run
   )

   print(f"Created {results['kpis_created']} KPIs")
   print(f"Map: {results['map_path']}")

   # Delete all KPIs
   deleted = skill.delete_all_kpis()
   print(f"Deleted {deleted} KPIs")
   ```

### Claude Skill API

The `FlowsPutLinesSkill` class provides these methods:

- **`list_kpis()`** - Returns list of all existing KPIs
- **`delete_all_kpis()`** - Deletes all KPIs, returns count deleted
- **`provision_kpis_from_gtfs(gtfs_url, route_short_names, max_routes, apply)`** - Downloads GTFS and creates KPIs
  - Returns: `{routes_processed, directions_with_shape, kpis_created, kpi_details, map_path}`
- **`get_kpi_details(kpi_id)`** - Gets detailed status for a specific KPI

### Complete Example

```python
from claude_skill.flows_put_lines_skill import FlowsPutLinesSkill

# Initialize
skill = FlowsPutLinesSkill("config.yaml")

# Clean up old KPIs
deleted = skill.delete_all_kpis()
print(f"✓ Cleaned up {deleted} old KPIs")

# Create new KPIs from Rome GTFS
results = skill.provision_kpis_from_gtfs(
    gtfs_url="https://dati.comune.roma.it/.../rome_static_gtfs.zip",
    route_short_names=["211", "C2", "62"],
    max_routes=3,
    apply=True
)

print(f"✓ Processed {results['routes_processed']} routes")
print(f"✓ Created {results['kpis_created']} KPIs")
print(f"✓ Map: {results['map_path']}")

# Check individual KPI status
for kpi in results['kpi_details']:
    details = skill.get_kpi_details(kpi['kpi_id'])
    print(f"  {kpi['name']}: {details.get('statusCode')}")
```

**Full documentation:** See [claude_skill/README.md](claude_skill/README.md) for complete API reference and examples.

## Package Structure

```
Flows-put-lines/
├── skills/
│   └── flows-put-lines-skill/
│       ├── SKILL.md              # AgentSkills-compliant skill definition
│       └── scripts/provision.py  # Script entry to run the skill
├── src/
│   ├── __init__.py          # Package exports
│   ├── main.py              # CLI entrypoint
│   ├── config.py            # Configuration loader
│   ├── gtfs_extractor.py    # GTFS processing
│   ├── flows_kpi.py         # PTV Flows KPI API client
│   ├── route_matcher.py     # Route matching for entities
│   └── map_generator.py     # HTML map visualization
├── claude_skill/
│   ├── __init__.py
│   ├── flows_put_lines_skill.py  # Claude skill implementation
│   └── README.md            # Skill documentation
├── pyproject.toml           # Package configuration
├── config.yaml              # Runtime configuration
└── README.md                # This file
```

### AgentSkills Skill
- Skill directory: [skills/flows-put-lines-skill/SKILL.md](skills/flows-put-lines-skill/SKILL.md)
- Quick run via script:

```bash
PTV_API_KEY=your_key \
FPL_GTFS_URL=https://example.com/gtfs.zip \
FPL_ROUTES=211,C2,62 \
FPL_MAX_ROUTES=3 \
FPL_APPLY=true \
python skills/flows-put-lines-skill/scripts/provision.py
```

The SKILL.md follows the AgentSkills specification (name, description, optional metadata,
progressive disclosure, and references). Use this when integrating with agent platforms
that support the AgentSkills format.
## Secrets
- Keep your PTV API key in the environment (`PTV_API_KEY`) or as a GitHub repository secret for CI/CD; do not commit keys to source control.
