# 📦 Understanding Outputs and Metadata

Once you run a method in CCP, the platform doesn’t just give you output files — it also generates structured **metadata** describing what happened, how, and with what.

This guide helps you understand:

- Where your results go
- What kind of metadata is available
- How to use that metadata for traceability and reproducibility

---

## 📂 1. Where Are Outputs Saved?

Every method writes output files to a shared folder inside the container:

`/ccp_data/`


Any file written there:

- Will be listed in the **execution result**
- Can be downloaded individually

💡 Tip: always write or move your results (images, CSVs, logs) to `/ccp_data/`.

---

## 🧾 2. What Are Metadata Files?

Alongside the actual output, CCP generates a series of metadata files. These help track:

- Which method was run
- With which parameters and inputs
- On which infrastructure
- By which user (if available)
- When and how the job ended

These files are included in every execution archive.

---

## 📄 3. Key Metadata Files Explained

| Filename                   | What it contains                             |
|----------------------------|----------------------------------------------|
| `metadata_method.json`     | Static info about the method: name, version, parameters |
| `metadata_request.json`    | The input values passed by the user          |
| `metadata_jobStatus.json`  | Execution state (e.g. finished, failed), timestamps |
| `metadata_infrastructure.json` | Where the job was run (engine, VM, etc.) |
| `metadata_provo.xml`       | Provenance file in [W3C PROV](https://www.w3.org/TR/prov-overview/) format |
| `outputs_stdout.txt`       | What your script printed to stdout           |
| `outputs_stderr.txt`       | What your script printed to stderr (warnings/errors) |

These are saved inside the **execution directory**, together with your actual outputs.

---

## 🔁 Example: Reproducibility

Let’s say you run a model training method and want to:

- Know which parameters were used
- Retrieve the training data file
- Share or repeat the same run later

You can do all of that using:

- `metadata_request.json` → to see what was passed
- `metadata_provo.xml` → to trace input/output relationships
- `outputs_ccp-entrypoint.sh` → to inspect the actual shell command used

---

## 📎 Best Practices

- Enable **archiving** if you want to keep full metadata + files together
- Use metadata to **document your experiments**
- Attach `metadata_provo.xml` when publishing or sharing results

---

## 🔗 Related Pages

- ▶️ [Executing a Method in CCP](02_executing_methods)
  
- ⏯️ [Understanding Executions](/03_execution/01_executions)

- 📜 [Execution `metadata_provo.xml`: Execution Provenance Metadata](/03_execution/05_metadata_provo)